Cognitive Developments in Economics

Cognitive Developments in Economics

Cognitive economics proposes an interdisciplinary approach to the study of human problem-solving, choice, decision-making and change, to explain economic transactions, the nature and evolution of organizations and institutions. This book contains some of the most cutting-edge research in economics. Leading scholars, including Ulrich Witt, Stanley Metcalfe, Massimo Egidi, Nicolai Foss, Richard Arena and Bart Nooteboom, cover such topics as: • • • •

economic methodology bounded rationality learning and cognition in firms institutional economics

Cognitive economics contributes to a large spectrum of economic fields such as consumer theory, economics of the firm, economics of innovation, evolutionary economics and experimental economics. This book is particularly pertinent given the recent award of the Nobel Prize to Daniel Kahneman – evidence that this is one of the true frontiers of economic theory today. Salvatore Rizzello is Professor of Economics and coordinator of the Centre for Cognitive Economics at the University of Eastern Piedmont, Italy.

Routledge frontiers of political economy

1 Equilibrium Versus Understanding Towards the rehumanization of economics within social theory Mark Addleson 2 Evolution, Order and Complexity Edited by Elias L. Khalil and Kenneth E. Boulding 3 Interactions in Political Economy Malvern after ten years Edited by Steven Pressman 4 The End of Economics Michael Perelman 5 Probability in Economics Omar F. Hamouda and Robin Rowley 6 Capital Controversy, Post Keynesian Economics and the History of Economics Essays in honour of Geoff Harcourt, volume one Edited by Philip Arestis, Gabriel Palma and Malcolm Sawyer

9 New Keynesian Economics/Post Keynesian Alternatives Edited by Roy J. Rotheim 10 The Representative Agent in Macroeconomics James E. Hartley 11 Borderlands of Economics Essays in honour of Daniel R. Fusfeld Edited by Nahid Aslanbeigui and Young Back Choi 12 Value, Distribution and Capital Essays in honour of Pierangelo Garegnani Edited by Gary Mongiovi and Fabio Petri 13 The Economics of Science Methodology and epistemology as if economics really mattered James R. Wible

7 Markets, Unemployment and Economic Policy Essays in honour of Geoff Harcourt, volume two Edited by Philip Arestis, Gabriel Palma and Malcolm Sawyer

14 Competitiveness, Localised Learning and Regional Development Specialisation and prosperity in small open economies Peter Maskell, Heikki Eskelinen, Ingjaldur Hannibalsson, Anders Malmberg and Eirik Vatne

8 Social Economy The logic of capitalist development Clark Everling

15 Labour Market Theory A constructive reassessment Ben J. Fine

16 Women and European Employment Jill Rubery, Mark Smith, Colette Fagan and Damian Grimshaw

26 Political Economy and the New Capitalism Edited by Jan Toporowski

17 Explorations in Economic Methodology From Lakatos to empirical philosophy of science Roger Backhouse

27 Growth Theory A philosophical perspective Patricia Northover

18 Subjectivity in Political Economy Essays on wanting and choosing David P. Levine 19 The Political Economy of Middle East Peace The impact of competing trade agendas Edited by J.W. Wright, Jnr 20 The Active Consumer Novelty and surprise in consumer choice Edited by Marina Bianchi 21 Subjectivism and Economic Analysis Essays in memory of Ludwig Lachmann Edited by Roger Koppl and Gary Mongiovi 22 Themes in Post-Keynesian Economics Essays in honour of Geoff Harcourt, volume three Edited by Peter Kriesler and Claudio Sardoni 23 The Dynamics of Technological Knowledge Cristiano Antonelli

28 The Political Economy of the Small Firm Edited by Charlie Dannreuther 29 Hahn and Economic Methodology Edited by Thomas Boylan and Paschal F. O’Gorman 30 Gender, Growth and Trade The miracle economies of the postwar years David Kucera 31 Normative Political Economy Subjective freedom, the market and the state David Levine 32 Economist with a Public Purpose Essays in honour of John Kenneth Galbraith Edited by Michael Keaney 33 The Theory of Unemployment Michel De Vroey 34 The Fundamental Institutions of Capitalism Ernesto Screpanti 35 Transcending Transaction The search for self-generating markets Alan Shipman

24 The Political Economy of Diet, Health and Food Policy Ben J. Fine

36 Power in Business and the State An historical analysis of its concentration Frank Bealey

25 The End of Finance Capital market inflation, financial derivatives and pension fund capitalism Jan Toporowski

37 Editing Economics Essays in honour of Mark Perlman Hank Lim, Ungsuh K. Park and Geoff Harcourt

38 Money, Macroeconomics and Keynes Essays in honour of Victoria Chick, volume one Philip Arestis, Meghnad Desai and Sheila Dow 39 Methodology, Microeconomics and Keynes Essays in honour of Victoria Chick, volume two Philip Arestis, Meghnad Desai and Sheila Dow 40 Market Drive and Governance Reexamining the rules for economic and commercial contest Ralf Boscheck 41 The Value of Marx Political economy for contemporary capitalism Alfredo Saad-Filho

42 Issues in Positive Political Economy S. Mansoob Murshed 43 The Enigma of Globalisation A journey to a new stage of capitalism Robert Went 44 The Market Equilibrium, stability, mythology S.N. Afriat 45 The Political Economy of Rule Evasion and Policy Reform Jim Leitzel 46 Unpaid Work and the Economy Edited by Antonella Picchio 47 Distributional Justice Theory and measurement Hilde Bojer 48 Cognitive Developments in Economics Edited by Salvatore Rizzello

Cognitive Developments in Economics

Edited by Salvatore Rizzello

First published 2003 by Routledge 11 New Fetter Lane, London EC4P 4EE Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 Routledge is an imprint of the Taylor & Francis Group This edition published in the Taylor & Francis e-Library, 2004. © 2003 editorial matter and selection, Salvatore Rizzello; individual chapters, the contributors All rights reserved. No part of this book may be reprinted or reproduced or utilized in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Cognitive developments in economics / [edited by] Salvatore Rizzello. p. cm – (Routledge frontiers of political economy ; 48) Includes bibliographical references and index. 1. Economics–Methodology. 2. Institutional economics. 3. Cognitive science. I. Rizzello, Salvatore, 1963– II. Series. HB131.C64 2003 330'.01–dc21 2003043149 ISBN 0-203-57990-9 Master e-book ISBN

ISBN 0-203-33950-9 (Adobe eReader Format) ISBN 0-415-30620-5 (Print Edition)

Contents

List of figures List of tables List of contributors Acknowledgements Introduction: towards a cognitive evolutionary economics

ix x xi xii 1

SALVATORE RIZZELLO

1 Discrepancies: competing theories and ideologies as cognitive traps: the scientific discovery as puzzle-solving

20

MASSIMO EGIDI

2 Cognitive science meets evolutionary theory: what implications does evolutionary psychology have for economic theorising?

53

JACK J. VROMEN

3 Towards a cognitive experimental economics

82

MARCO NOVARESE

4 Accounting for social knowledge in economic analysis: the relevance of Adam Smith’s framework

107

PIER LUIGI PORTA AND ROBERTO SCAZZIERI

5 The making of a behavioural economist: Herbert A. Simon and the early evolution of bounded rationality

133

MIE AUGIER

6 The rhetorical dimensions of bounded rationality: Herbert A. Simon and organizational economics NICOLAI J. FOSS

158

viii

Contents

7 Cognition, entrepreneurial conceptions and the theory of the firm

177

ULRICH WITT

8 Knowledge, understanding and the epistemology of innovative firms

193

J. STANLEY METCALFE

9 Managing exploitation and exploration

218

BART NOOTEBOOM

10 The epistemic foundations of social organizations: a game theoretical approach

243

CHRISTIAN SCHMIDT

11 Interactive learning and technological knowledge: the localised character of innovation processes

268

CRISTIANO ANTONELLI, JEAN-LUC GAFFARD AND MICHEL QUÉRÉ

12 Cognitive economies and the ‘nature of the district’

281

MARGHERITA TURVANI

13 Learning in economics: some Austrian insights

302

PIERRE GARROUSTE

14 Beliefs, knowledge and equilibrium: a different perspective on Hayek

316

RICHARD ARENA

15 Guesswork and knowledge in evolutionary economics: Veblen revisited

338

FRANCESCA LIDIA VIANO

16 Problem complexity and problem representation: some implications for the theory of economic institutions

371

LUIGI MARENGO

Index

389

Figures

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 1.11 1.12 3.1 3.2 9.1 9.2 10.1 10.2 10.3 16.1 16.2 16.A1 16.A2

The comparative efficiency of two strategies that solve the same problem Rubik cube: a sub-problem ‘Target the Two’: landscape of the efficiency of strategies ‘Target the Two’: routinized behaviours ‘Target the Two’: single-player routinization Minirubik Minirubik: the exchanging of tiles Minirubik: example of a condition–action instruction Minirubik: example of a starting configuration and the goal Minirubik: an array of elementary instructions Minirubik: ‘First A’ representation Minirubik: initial position which is suboptimally played with the ‘First A’ strategy The game played by one of the teams in the experiment ‘sum 10’ The game played by another team in the experiment ‘sum 10’ Cycle of exploitation and exploration Cycle of integration and disintegration Borel’s and Volterra’s problem Nash situation example Goal orientation Fitness landscape with independent components Fitness landscape with interdependent components Tower of Hanoi: the winning configuration Tower of Hanoi: two starting configurations

35 36 39 40 41 43 43 43 44 45 46 46 100 101 226 229 249 257 263 375 375 385 385

Tables

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 10.1 16.1 16.2 16.A1

Logical relationships between combinations and results An example of confirmation of wrong answers and change of right choice Distribution of the mean ratio of wrong answers confirmed by players Number of right answers ‘forgotten’ Percentage of rules developed by experimental players Detailed analysis of player behaviour System of rules developed by a player Analysis of the variance for the estimation of equation (1) Parameters estimation for equation (1) The Robinson Crusoe and social exchange economies A landscape with local optima A derived landscape with local optima States and transitions of the Tower of Hanoi

88 89 90 90 91 93 94 98 98 246 382 383 386

Contributors

Cristiano Antonelli, University of Torino, Italy Richard Arena, University of Sophia-Antipolis, Nice, France Mie Augier, Stanford University, USA Massimo Egidi, University of Trento, Italy Nicolai J. Foss, Copenhagen Business School, Denmark Jean-Luc Gaffard, University of Sophia-Antipolis, Nice, France Pierre Garrouste, ATOM, Paris, France Luigi Marengo, University of Teramo, Italy J. Stanley Metcalfe, University of Manchester, UK Bart Nooteboom, Erasmus University Rotterdam, the Netherlands Marco Novarese, University of Piemonte Orientale, Italy Pier Luigi Porta, University of Milano-Bicocca, Milano Michel Quéré, CNRS-IDEFI, Nice, France Roberto Scazzieri, University of Bologna, Italy Christian Schmidt, University of Paris-Dauphine, France Margherita Turvani, University of Venezia, Italy Francesca Lidia Viano, University of Perugia, Italy Jack J. Vromen, Erasmus University Rotterdam, the Netherlands Ulrich Witt, Max Planck Institute for Research into Economic Systems, Jena, Germany

Acknowledgements

This volume contains the invited lectures and selected papers presented at the International Workshop on Cognitive Economics held in Torino and Alessandria (Italy) during November 2000, organised by the Centre for Cognitive Economics of the University of Piemonte Orientale (Alessandria) and by the Fondazione Rosselli of Torino, as the fifth workshop of the Network Human Reasoning and Decision Making of the European Science Foundation. I wish to acknowledge help from the sponsor of the workshop: Regione Piemonte, Comune di Alessandria, Provincia di Alessandria, Società per l’insediamento e lo sviluppo universitario in Alessandria e Asti and the Fondazione CRT di Torino. My debts go also to the other members of the scientific committee of the workshop: Giuseppe Clerico, Massimo Egidi, Marco Novarese and Riccardo Viale and to an exceptional set of people that spurred me to edit this book. Among them, I must particularly acknowledge for their support Cristiano Antonelli, Richard Arena, Stan Metcalfe, Roberto Marchionatti, Roberto Panizza, Pietro Terna, Christian Schmidt and Ulrich Witt. Finally, I address very special thanks to Elena Pasquini for his outstanding revision of my English. S.R.

Introduction Towards a cognitive evolutionary economics Salvatore Rizzello

After the dissolution of the Neoclassical theory as the exclusive reference paradigm in economics, various approaches have come to the forefront in the research in the economic field. Cognitive economics stands out from them, being one of the most fertile. It is an interdisciplinary approach concerned with the following subjects: problem-solving, choice and change in the explanation of economic transactions, the nature and evolution of organizations and institutions in a context characterized by structural uncertainty, scarcity and incentives. The economic agents’ behaviour has psycho-neurobiological foundations and is analysed in the light of bounded information, bounded and procedural rationality and satisficing behaviour. Cognitive economics contributes to a large spectrum of economic fields, such as consumer theory, economics of the firm, economics of innovation, evolutionary economics, institutional economics and experimental economics. This book illustrates the most recent developments in this field. The aim of this introduction is twofold: supplying the reader with a general survey of the transversal subjects dealt with in the various chapters; sorting out the future perspectives which can be perceived in this field of research. For the first point, we will refer to the key words of this book – cognition, evolution, learning, uncertainty and path-dependence – and we will analyse them in detail. As concerns the second, this introduction will try to illustrate the ‘cognitive evolutionary approach’, which satisfactorily summarizes the future perspectives of cognitive economics, especially in the field of the analysis of endogenous change processes in dynamic economic systems. From the point of view of its historical evolution, cognitive economics is certainly linked to the cognitivist revolution of the 1950s (Rizzello 1999: ch. 8); still, the development of this discipline has its own history. Alfred Marshall can be considered one of the founders of this approach, thanks to his views on the profound connection between the structure of organizations and the workings of the mind and the role and structure of the brain (Raffaelli 1994). In the nineteenth century an important role in the

2

Salvatore Rizzello

emergence of this approach was also played by Carl Menger, especially by means of his studies on the spontaneous nature of norms and institutions and their role as social binding agents. Furthermore, following Menger, norms are now directly linked to the limits of the human mind in its handling of complex environmental variables.1 Also Thorstein Veblen is to be mentioned among those who were aware of the connection between mental mechanisms, evolution and the role of norms, as recently pointed out in a series of papers published on the Cambridge Journal of Economics (July 1998: Vol. 22, No. 4). The history of the twentieth century follows with its major representatives in this field: Hayek with his development of a model of mind aimed at explaining the role of bounded information and the nature of institutions in economic processes; Simon, who highlighted the connection between mental processes and the nature of human rationality and organizations; Boulding (1956), who shed light on the role of the image in the production of knowledge on the part of individuals, and on its relevance in decision-making and more generally in economic processes; Allais, Kahneman and Tversky and those researchers in the field of experimental economics who study individual and organizational learning processes and the processes of coordination of agents in a condition of uncertainty and bounded information, and who have stressed the limits of the expected utility theory and developed alternative theories. Moreover, in this field of research we will mention the recent contributions of the agent-based simulation approach in economics (Luna and Perrone 2001), which are being developed along with the development of computer technology and of systems using artificial agents. Such contributions are more and more often proposing very interesting exchanges of ideas with experimental economics. All these subjects will be dealt with in the following chapters. Coherent with its aims, this introduction is organized as follows: the first section deals with the connection between cognition and evolution. By integrating the cognitive and the evolutionary approaches to economics, the main purpose of this section is to demonstrate that the new self-organizing approach to explain the dynamics of change in economics is very relevant and compatible with cognitive economics. The next section concerns this relevance and presents economic systems as cybernetic ones. It introduces the relatively new concepts of exaptation and neurognosis, used in biology and anthropology, which give us new tools to explain the nature and role of learning in evolution. The third section outlines the compatibility of this approach with the nature and workings of the mind. In particular, it describes the affinity of these concepts with Hayek’s psychological foundations of the concept of evolution. As will emerge, all these processes are path-dependent. To this purpose, the fourth section describes path-dependence as resistance to change. This relatively new concept is compatible with Paul David’s and Brian Arthur’s idea of path-dependence, and it seems to be most appropri-

Introduction 3 ate in explaining the evolutionary dynamics of self-organizing systems. The penultimate section takes into consideration the relevance of this approach for organizational and institutional change. The final section offers some concluding remarks.

Cognition and evolution Cognition and evolution are two relevant features of economics. To support this point of view it is enough to consider that individual behaviour is the analytical unit to understand economic systems, such as markets, organizations and institutions. Moreover, the latter are dynamic systems, characterized by feedback, change, time irreversibility and development. When considering individual behaviour, it is very important to take into account what psychology and neurobiology teach us. It is also relevant to model choice by taking into consideration how the human mind works and which role is played by learning in decision-making. Furthermore, to describe the dynamics of change, innovation and development of economic systems it is advisable to assume an evolutionary approach to economics. Whereas the evolutionary approach is emerging as an almost consolidated new strand, the cognitive perspective is relatively new in economics. In the first instance, it is necessary to point out that both – cognition and evolution – are wide theoretical concepts which tend to resist satisfactory definition. In fact, although a growing number of economists use them, it is still possible to register different theoretical positions. Let us consider evolutionary economics. Half a century after its foundation (Alchian 1950), the evolutionary approach to economics presents a very large spectrum of application. Gradually this paradigm – which was traditionally confined to explain the dynamics of competition from a Darwinian and Schumpeterian point of view – has been extended to many other ambits, the most relevant of which are endogenous change, pathdependent dynamics, and organizational learning (Dosi 1991; Witt 1991; Dosi and Nelson 1994; Nelson 1995). It is not within the scope of this introduction to reconstruct the history of evolutionary economics, yet it may be appropriate to indicate briefly the relevant passages of the development of evolutionary economics.2 Nelson and Winter’s (1982) book represents both one of the most relevant contributions to the comprehension of the dynamics of change and a bridge that links different analytical traditions. In fact, on the one hand, following Alchian’s (1950) seminal work, it clusters the Darwinian theory on natural selection and competition and the Schumpeterian ideas on technological change and economic dynamics. On the other hand, though it is mainly confined to the explanation of exogenous dynamics, this book contains analytical insights which allow the integration of different tradition in the explanation of the dynamics of change.

4

Salvatore Rizzello

In the last decade of the twentieth century it was easy to find many contributions stressing the relevance of endogenous and path-dependent dynamics for economic change. Yet some problems had to be coped with in order to reach such integration. Next to Schumpeterian tradition, a new alternative approach arose. This can be synthesized as the Marshallian approach to economic change, stressing the relevance of endogenous change and self-organization analysis (Foster 1993, 1997; Hodgson 1997a; Foss 1997; Witt 1997).3 The latter rejects the use of biological analogies and gives prominence to the fact that the interpretation of Schumpeter’s ideas on evolution is mainly the fruit of a misconception (Andersen 1997; Hodgson 1997b; Foster 2000). The self-organizing approach is very relevant for evolutionary economics because economic systems are mainly time-irreversible and dissipative structures.4 Evolutionary analysis and cognitive economics should be integrated.5 Cognitive economics considers the relevance for economics of human cognitive aspects. In particular, it stresses that ‘economics implies choices. A choice is the result of psycho-neurobiological acts. The assumptions that are at the basis of economic theory, therefore, must be consistent with the mechanisms that guide the workings of the human mind’ (Rizzello 1999: xv). A relevant part of heterodox literature presents a very interesting approach that stresses the relevance of the mind’s workings in explaining economic behaviour.6 Although one may think that the cognitive approach to economics is very recent, we can find some relevant contributions in the past, starting from Marshall’s 1867 writings on the relevance of the mind to analyse organizations (Marshall 1867–8; [1871] 1961: ch. IX).7 Starting from Marshall’s early writings, this tradition has continued with the contributions of the Austrian school – Menger and Hayek in particular – on the subjectivist theory of perception, knowledge and cultural evolution (Menger 1883; Hayek 1937, 1952, 1963, 1973, 1979). The 1950s cognitive revolution integrated these theories with Simon’s works on rationality and Knight’s concept of structural uncertainty (Knight 1921; Simon 1957), and finally it enabled them to be integrated into the behavioural economics of the firm (March and Simon 1958), into the experimental economics approach to decision-making (Kahneman and Tversky 1979), and into the path-dependence literature on technological, organizational and institutional change (see, respectively, David 1985, 1997; Antonelli 1999; Denzau and North 1993; Witt 1993). A common element in this literature is that the microfoundations of economic behaviour are directly linked to the nature and role of the human mental mechanisms in charge of the production of knowledge and the emergence and use of rules, routines and their evolution. This point of view affects and tries to integrate both the neurobiological and the psychological human dimension, and learning is its analytical corner-

Introduction 5 stone; learning is also the main bridge between cognitive and evolutionary economics. Learning is crucial, in fact, in understanding how people choose, but also in explaining how routines and rules emerge in an organizational and institutional context, and the way in which the individual–environment feedback occurs.

Self-organization, neurognosis and exaptation The self-organization approach is emerging as a new promising branch of evolutionary economics, which differs in some respects from the most traditional models of evolution, applied to economic change. By proposing a new point of view on evolution, such an approach seems able to answer some open relevant questions in evolutionary economics that invest the nature, role and dynamics of economic change. Such an approach is typical of time-irreversibility and dissipative structures in which change is not linear, and uncertainty, creativity and novelty matter. The analytical foundations of time-irreversibility are based on the second law of thermodynamics (the principle of increasing entropy) that considers the capacity of the systems to acquire free energy and to promote structural evolution (Foster 1993; Hodgson 1995; Lesourne 1997; Witt 1997). Therefore, timeirreversible systems are non-equilibrium open thermodynamics ones and can be defined as dissipative structures which, in Prigoginean tradition, ‘are able to achieve a degree of “self-organization” (or autopoiesis) which enables them to export entropy and import free energy to maintain themselves’ (Foster 1993: 985).8 The evolutionary aspects of these systems rely on the fact that if these structures cannot reverse easily, they can only change by evolving. The self-organization approach presents a wide range of applicability that also includes socioeconomic systems, characterized by their informational rather than energetic nature. The most important units of analysis in economic systems are active agents and their intelligent nature, which produce knowledge (creativity) and informational flows. From a very general point of view, the main question to be answered concerns how change happens. Usually, standard evolutionary economics presents two distinct levels of analysis to explain change (i.e. an endogenous and an exogenous level). Most recently, evolutionary economists tried to integrate these levels (the first contributions in this direction are Silverberg et al. 1988; Lane 1993a, 1993b; Dosi and Kaniovsky 1994). The major tools used are directly borrowed from biology and consist mainly of evolution by means of natural selection (in a Darwinian or Lamarckian tradition). They use biological analogies to explain the relationship between agents (or firms) and environment, or to model the evolution of routines like genes.9 The point of view here presented considers the relevance of the internal

6

Salvatore Rizzello

dynamics of the evolution of systems, as human beings, organizations and institutions. Because of their entropic and cybernetic nature, it is advisable to study the dynamics of these systems by means of new analytical tools like neurognosis and exaptation, which emerged in biology and anthropology. Because of their informational nature, economic systems are entropic. They produce, use and waste information. Yet they are also cybernetic systems. The latter are characterized by their capacity to self-regulate and evolve in a mutable environment. This implies the presence of channels of communication, allowing the systems to react to changeability. Usually this happens by means of ‘negative feedback’.10 Many examples of this kind are easily found in living organisms, as well as in organizations and institutions. The reference here is not only to the capacity to regulate the temperature of the body, for example, but also to some interesting psychological cybernetics mechanisms, which sometimes allow us to ‘deny the validity of information which is too upsetting to our identity or to our existing images of the world’, as Boulding suggests (1992: 289; see also Boulding 1956).11 This aspect is particularly relevant. Human mind is in fact able to build images of the world. This characteristic of the mind allows human beings to adapt these images to changes. As the cognitive sciences suggest, mind is the product of brain activity. The human brain can be easily described as a cybernetic structure. To describe how it works, it may be appropriate to explore how the mechanisms of perception, learning and adaptation function. The traditional answer that biology gives us (evolution by means of selection) is not completely satisfying. Certainly it is a good tool to explain how organisms adapt themselves to the changing environment, but it is not able to explain the functioning of the polarity between environmental adaptation and the protection of their internal integrity. Individuals do not simply ‘adapt’ themselves to the changing environment – they resist these changes as long as possible by interpreting and selecting external data in a way that they deem to be the most appropriate for their cognitive maps. If we take into consideration that evolution does not occur only by means of an adaptation mechanism, but that in the process of change and development the cognitive innate structures prevail,12 we can find a better answer by using the new analytical tools mentioned above. Let us consider neurognosis. When an organism faces new information its capacity to give significance to this information depends on its previously stored experience and on its innate neurognostic structures.13 Experience, in fact, is the result of active interaction between associative structures (neurobiological) and sensorial data. ‘The neural networks comprising the cognized environment have their developmental origin in initial neurognostic structures that are generally present before, at, or just after birth. The initial organization of neurognostic structures is primarily determined by the genotype’ (Laughlin 1996: 365). Since our birth we are

Introduction 7 therefore structured to explore and model the world. The brain, at every moment, ‘imposes its relatively conservative order upon the experience it constructs’ (ibid.). The main characteristic of our brain is its capacity to evolve in a self-regulated way, including a degree of elasticity that allows it to explore, and interpret its world actively. Mental structures play a central role in the process of perception, and in that of giving significance and of constructing knowledge, and neurognosis seems to be able to offer a good tool to explore this dimension. The human brain and mind evolve by following a path that strongly depends on innate pre-existing structures. Because of this dependence on its previous experiences and its innate structures, this can be clearly described as a path-dependent process. But it is important to stress that path-dependence emerges here more as a mechanism of resistance to change, rather than as a simple influence of the previous path on the development of organisms. The human mind tends to preserve itself, as much as possible, from change.14 In the evolution of nervous systems a double mechanism works simultaneously: path-dependence and feedback. The brain collects external stimuli and the mind gives them significance. The brain associates the new stimuli by following its structures and it continuously tries to verify – by means of a feedback mechanism – the reliability of its classification. The mind gives significance to the stimuli it receives by using feedback and path-dependence. The question arising at this point is: which is the mechanism that can explain how brain and organisms evolve? We can find the answer by considering exaptation. The term ‘exaptation’ was coined by biologists to design the situations in which evolutionary systems discover new uses for old inventory (Varela 1979; Gould and Vbra 1982; Gould 1991). This happens when organisms become able to use, for a novel function, something which arose for some other reasons.15 Following Gould, major innovations in evolution are all the fruits of an exaptation process. ‘The human brain is, par excellence, the chief exemplar of exaptation’ (Gould 1991: 55). It continuously builds models of the world and of itself and, in doing so, new neuronal structures emerge in order to give significance to the sensorial data from old ones. Exaptation is not predictable because it is an act of co-optation for the new function of something which evolved for different reasons. Apparently exaptation is irreconcilable with neurognosis. The latter, in fact, stresses the relevance of innate structures in explaining evolution. The former does not consider innateness explicitly and it does not clearly underline the role of path-dependence. But the point of view here proposed supports these two analytical categories as complementary, by presenting Hayek’s idea of evolution, selection and change. Hayek’s surprising concept of evolution has psycho-neurobiological foundations. As will emerge, this is very close to the above-mentioned biological and anthropological literature.16

8

Salvatore Rizzello

Hayek’s psycho-neurobiological concept of evolution Although the interpretation of Hayek’s ideas on cultural evolution and group selection is still a controversial issue among his scholars (Caldwell 2000; Witt 2000), there is wide evidence that by means of this concept he refers to how learnt rules, group norms of conduct, habits, routines and institutions emerge and evolve. Furthermore, the role played by the human mind’s nature and its limits appears evident in Hayek’s conception of evolution of rules and institutions. Briefly, we can state that rules of conduct, habits and routines emerge from human limits in interpreting the very complex external world. Following rules and codifying them in institutions is an ‘economic way’ to act successfully. Rules and organic institutions, in fact, standardize the world and in so doing they simplify the ambit in which humans use their limited cognitive capacity. In the last decade a growing number of scientific contributions stressed the relevance of Hayek’s book on psycho-neurobiology, which for many years was neglected by scholars. The Sensory Order was conceived by Hayek in the 1920s but only published in 1952, thirty years later. The late acknowledgement of the relevance of this book permits us to revisit some central concepts of Hayek’s thought.17 As he stressed many times, The Sensory Order is a cornerstone to understand Hayek’s thought on knowledge, competition and cultural evolution (Hayek 1979: 199–200; Hayek 1994). It may be appropriate to illustrate here a brief summary of the model of the mind found in his book.18 In Hayek’s conception, the mind is a framework that orders perception through acts of interpretation. The human neuronal structure classifies external sensorial data by means of a process of association of classes of stimuli into classes of responses. The significance that we give to each perception depends upon the genetic characteristics of the individual and upon his/her previous activity of classifications of external stimuli (experience). The mind does not receive sensations in a passive way. On the contrary, it is an active tool that interacts with the external environment. The mind also continuously builds an image of itself and of the world and rebuilds them in a tacit and unconscious way. This image of itself and of the world is the framework that allows us to give significance, by means of personal and idiosyncratic interpretation, to external information. More important, this allows us to construct the knowledge – the product of an internal and subjectivist process – which we use to act. Being based on the person’s genetic traits and on his/her personal experiences, every person acts after performing a process of ‘interpretation’ of the external (objective) information, which is transformed into subjective knowledge which is unique and original. Through a learning process that takes place over the years, genetic traits and personal experiences continuously redesign the neuronal circuitry that represents

Introduction 9 our image of ourselves and of the external world, depending on both personal activities and the action of innate bioregulatory circuitry. This mechanism explains how brains assure the polarity between environmental adaptation and the protection of its internal integrity. As is emerging, neurognosis and exaptation are evident in Hayek’s conception of the human brain. The evolution of the mind – i.e. the evolution of our capacity to build and process images and symbols to generate knowledge – happens, in fact, by means of a balance between ontogenesis and phylogenesis. Starting from its native structures, the brain evolves by building new nervous circuitry. It results from the feedback with the relatively inelastic (but not completely rigid) nature of our a priori mental schemes which interpret external information in a path-dependent way. This interpretation is carried out by means of exaptation. In other words, previous neuronal structures built and developed to solve problems of interpretations of the external world effectively, reveal their capacity to co-opt new configurations and functions when the individual faces new problems.19 After this process, these new modified neuronal configurations, in turn, are ready to co-opt new unfamiliar external data, and so on. Some cerebral circuitry remains stable, and the brain builds up its balance between stable and unstable circuitry. When faced with a problematic new situation, individuals generally refer to previous successful schemas of action that permit an interpretation of new phenomena (Butos and Koppl 1997). Classification through individual association of stimuli leads to interpretative outcomes and action, which can vary greatly from one individual to another. At any given point in an individual’s life, a great part of his cerebral circuitry is personal and unique, since it reflects his genetic characteristics, and the history and events of that particular organism (Witt 1992; Vanberg 1994: ch. 6), which is also the result of the interactive process which takes place – by means of exaptation and feedback – with his/her cultural and social context (Hayek 1963). Therefore, Hayek emphasizes the importance of the connection between the evolution of the mind and the evolution of institutions. Even if it is beyond the scope of this introduction to consider the controversial debate on Hayek’s concept of evolution, it is certainly worth pointing out that this particular conception – so strongly linked to the role and function of the mind seen as a self-regulating mechanism – seems to be a useful tool to extend the analysis of the evolution of self-organizational structures. All these aspects of Hayek’s thought are, in fact, surprisingly coherent with the biological and anthropological micro-foundations of the analysis of self-regulating systems. This represents a good reason to make an attempt to integrate this ‘evolutionary’ literature with the cognitive tradition, and in particular with Hayek’s ideas. To this purpose, we need to take two more steps. Since path-dependence seems to play a crucial role for understanding these processes, the first step concerns the explanation

10 Salvatore Rizzello of how path-dependent analysis is coherent both with Hayek’s ideas and with the neurognostic biological tradition. The second one consists in illustrating the new perspectives of economics of the firm and institutional economics introduced by this approach.

Path-dependence in human cognition Path-dependence is a relevant analytical tool for economic theory. In the last two decades, a growing literature has dealt with this topic. It is not within the bounds of this introduction to propose a review of this interesting literature. But, in spite of the unconvincing criticisms by Leibowitz and Margolis (1995),20 it is possible to make an attempt to summarize, in a few sentences, the main points introduced by this literature into economics. Two basic ideas are central in path-dependence: (1) history matters in determining the dynamics of social and economic processes; (2) the passage from one state to another of an economic system depends on the previous paths. The first idea conveys the principle that historical events (even small ones) condition the system’s evolution with some rigidities that the economic action can modify only in part. The second one stresses the aspect that the outcome of a path-dependent process is not foreseeable at all. The final equilibrium reached by the system can be a suboptimal one. As David and Arthur demonstrated, the path of dynamic systems depends on a stochastic and self-reinforcing mechanism that usually conducts to ‘lockin’ the system into a trapping region, i.e. the basin of attraction that surrounds a locally (or globally) stable equilibrium.21 These kinds of equilibrium are stable but not (necessarily) optimal, and they often result to be multiple equilibria. A relevant aspect that is usually neglected in this literature is that the early mechanism of path-dependence is founded in the human brain (Rizzello 1997). The dynamics of economic systems depend, in fact, on the interaction among individuals and on their choices. The latter depend on the way agents acquire information and produce and use knowledge. This is a crucial process for economic theory, which Hayek clearly supported. Moreover, it is coherent with the idea that economic systems (from individual to institutional dimensions) are cybernetic and neurognostic. Not only that – if we assume the central role of the human mind and brain to explain economic processes, exaptation becomes a better tool to explain how these systems evolve. Let us explain this point in depth. Following Hayek’s insights and the more recent neurobiological and cognitive teachings,22 we can note that the human brain presents a particular pre-natal structure allowing it to interpret and give significance to the external world. This neuronal structure evolves by means of exaptation in classifying new external stimuli in its pre-existing nets. The actions which

Introduction 11 can solve problematic situations successfully are interpreted to satisfice (in Simon’s sense) choices. They depend on four aspects at least: the genetic neurobiological dimension of the individual; his/her previous experience; historical stochastic accidents; the feedback with the environment. It is easy to picture the brain as a neurognostic mechanism which tries to perceive external data by balancing its pre-existing neuronal structure, so as to interpret and use the new stimuli effectively. The interpretation of external stimuli generates an action. If this is successful – in satisficing terms – it reinforces the perception mechanism that reduces the neurobiological elasticity to interpret (to exapt) the same situation in a different way when it recurs. Or, in other words, it increases the resistance of our (neurognostic) brain to change. This mechanism of perception and feedback is coherent with the ‘lock-in’ idea. But, whereas path-dependence literature usually refers in general to this mechanism of economic systems, the point of view supported here is to extend the idea of the ‘trapping states’ to individual decision-making, starting from the brain’s processes of perception and the mind’s mechanisms of representation. The image of the external and internal world, which each individual constructs, depends both on his/her innate neurobiological perceiving structures and on the evolution of these structures that exapt themselves following previous experiences. The fact that the evolution of the brain is path-dependent means that it certainly depends on its history, but also on its neurognostic structure, which in turn resists continuous changes. The image of the self and the external world tends to preserve the previous one as long as possible. In a sense, neuronal structures have a conservative nature, but they are not completely rigid. As is emerging, when the brain processes information and produces knowledge it acts in a dimension characterized by structural uncertainty. It clusters four kinds of uncertainty that we can summarize as follows: human genetic features are stochastic; previous experiences are idiosyncratic; historical accidents are not forecasting; and successful feedback processes depend on the combination of the other three levels, and interindividual communication takes place thanks to a common institutional framework, education, language and the rules of conduct that individuals share.

Evolution and organizational and institutional change Up to now we have applied this model to decision-making processes, which combine Hayek’s view on perception and knowledge, the tradition on pathdependence, and the new biological conception of evolution. However, it might be possible to extend its applicability to organizational and institutional levels. The latter are the object of a wide literature on pathdependence and evolution. It might be fruitful to consider, in industrial and institutional economics, the dynamics of change as characterized by

12 Salvatore Rizzello exaptation, taking into account, for example, that innovation depends both on a firm’s history and on its resistance to change; or, similarly, the evolution of institutions as linked to history paths and also to cultural rigidities. On the industrial economic side, this idea could improve the models of the creation and control of demand by the firm, proposing a different and integrating point of view as compared to the literature that stresses exogenous change in particular. As Momigliano (1975) stressed twenty-five years ago, firms not only meet existing needs, they also create new ones. Research, innovations and inventions become more and more important within production processes. This view implies a concept of the competition process in which firms, instead of meeting the needs arising on the demand side, try to ‘frame’ the market ad to create their own demand artificially. Such strategies are strongly dependent on the firm’s specific characteristics and on its role in the market, its organizational structure, its history, previously adopted strategies, on how quickly it can get access to credit, on the process of learning in the use of scientific research and also on positive externalities due to economic policy. In other words, Momigliano proposes a (self-organizing) model of innovation and the development of dynamic structuralism (Antonelli 1995), stressing the relationship between a firm’s structure, the characteristics of the economic system it works within, and its performance. As regards the institutional side, the above proposal could further reinforce the models that consider the link between mind and institutions in explaining institutional change (Hayek 1942; Denzau and North 1994; Rizzello and Turvani 2000, 2002). More generally it could represent a step to explain further the relevance of endogenous dynamics in the evolution of economic systems by offering the new concept of evolution proposed by Hayek which is different from that of Nelson and Winter (Rizzello 2000). Stressing the role of evolution in the human mind and brain, and the link between mental mechanism and the emergence and nature of rules and institutions, Hayek gives us a model of evolution in which both ontogenesis and phylogenesis played a balanced role. In this light, the evolutionary cognitive approach could fill in the gap between the cognitive individual dimension and the holistic one. Let us consider this aspect very briefly. As suggested above, individual cognitive maps evolve by exaptation. But it is relevant to consider also that it mainly happens in a ‘cultural’ context. From a cognitive point of view, culture is an ensemble of representations shared by all the members of a group. This kind of representation is not ‘perfectly’ shared, because what individuals share are not mental facts but an ‘epidemiological’ distribution of casually linked mental and public facts in a human population (Sperber and Hirschfeld 1999: cxxii). Human cognitive dispositions allow individuals to adopt spontaneously cultural representations that are reinforced by previously acquired institutional constraints.23 Because of the neurognostic structure of their brains and their ability to exapt, humans continuously evolve both

Introduction 13 by resisting change and by adopting changes. Moreover, cognitive processes, involved in most relevant activity, do not take place just in one single mind but are distributed throughout many members. Since there is still a very big open problem in neoinstitutional literature, this could be a simple suggestion to address further developments.

Concluding remarks With reference to the key words of this book, one of the main purposes of this introduction was to stress the relevance of the integration of the cognitive approach to change and decision-making in evolutionary economics, with particular reference to self-organizing models. By introducing the concepts of neurognosis and exaptation it has tried to show that a unifying mechanism of evolution exists at every level of analysis, and it reinforces the idea that economic systems evolve in path-dependent terms. The introduction has not attempted to supply a wide discussion on all the aspects concerning evolutionary economics, as the reader will gain a good knowledge of them in the following chapters. On the contrary, it has focused on the microfoundations of evolutionary dynamics. By doing so, it has tried to show the affinity of this approach – which tries to combine neurognosis, exaptation and path-dependence – with the neuropsychological foundations of Hayek’s conception of evolution. Summing up, the self-organization approach to evolutionary economics seems to be particularly relevant in explaining endogenous change. The point of view here proposed analysed the cognitive aspects of this approach and some evolutionary implications. Following Hayek’s insights, it is time to go beyond the endogenous/exogenous dichotomy, taking into account the views which biology, anthropology and psychology propose on the same issues. In such an interdisciplinary perspective, cognitive evolutionary economics seems to offer good tools in order to improve the research into the dynamics of economic systems.

Notes 1 These subjects were further dealt with by Hayek (Rizzello 1999). 2 Hodgson (1995, 1999) offers a wide reconstruction of the history of evolutionary economics. 3 Some authors try to integrate these two traditions. In particular, Silverberg et al. (1988) maintained that competitive selection and the self-organizing approach could coexist in a neo-Schumpeterian model. But, as Foster pointed out (2000: 325), ‘they do not employ an economic self-organizing approach, in the sense of Schumpeter, but, rather, a more complex form of competitive selection’. Following Foster (1994, 2000), Schumpeter offers good insights to integrate competitive selection in a wider self-organizing approach. Schumpeter, in fact, conceives development and evolution as joint processes in the economic domain.

14 Salvatore Rizzello 4 Foster (1993) considers economic systems as dissipative structures, which usually have the peculiarity to be non-equilibrium open systems (see Mori and Kuramoto 1998). In Prigogine’s conception a dissipative structure is a thermodynamic system, ‘whose behaviour is determined by its boundary conditions, in contrast to what he defines as a dynamic system which is determined by its initial condition’ (Foster 1993: 985). Economic systems are also characterized by time irreversibility, which differentiates them from mechanical models. Foster (1993: 982) proposes to revisit Marshall’s thought. ‘Marshall was all too aware that evolutionary economics could not be possible until time irreversibility was properly understood.’ Furthermore, Foss (1997), Foster (1993: in particular p. 976) and Hodgson (1999) stress the relevance of Marshall’s ideas on time irreversibility, which contrast with the economic mechanical paradigm, whose introduction in economics is erroneously addressed to Marshall. More generally, Witt (1991: 87) indicated in time irreversibility and in endogenous change two structural aspects of evolution. 5 About the relevance of cognition for economics, the point of view here presented follows in particular the Austrian tradition on knowledge, creativity and novelty and it stresses the importance of these aspects for evolutionary economics, as Witt (1995, 1999) pointed out. 6 Rizzello (1999) offers, among other things, an overview of this literature. 7 As Raffaelli (1994) suggests, Marshall’s account of the economic system (especially in Book IV of the Principles) was significantly influenced by his early encounters with problems of knowledge. 8 Humberto Maturana coined the term ‘autopoiesis’ in about 1960. Maturana and Varela use the term to refer to the fundamental process of living systems. Autopoiesis is essentially the mechanism by which living systems continually produce themselves as autonomous unities. 9 The use of biological analogies in economics and the criticism of their use is a controversial issue. In the 1950s, Alchian’s (1950) paper primed a debate between Armen Alchian and Edith Penrose, who criticized the use of biological analogies (see Hodgson 1999 and Rizzello 2000). Likewise, Rizzello (2000) directed to Nelson and Winter’s 1982 book some remarks of Penrose’s criticism to Alchian. For other criticism about the use of biological analogies in evolutionary economics see Mirowski (1983), Witt (1996, 1999) and Foster (2000). 10 As Boulding (1992) pointed out, next to classical cybernetics, characterized by negative feedbacks, creodic processes and positive feedbacks play an important role in evolutionary processes. A creodic process is typical of a system that evolves following a blueprint or a design. As a good example of creodic processes, one can consider the construction of a building or the evolution of an egg. Positive feedbacks work in the opposite way of the negative ones. Far from moving systems into equilibrium, they increase the forces of disruption. Usually they concern the drastic and catastrophic changes, and in referring to economic systems they can be utilized to describe, for example, technological crisis in Schumpeterian terms. 11 In the literature this interesting aspect of human behaviour is referred to as cognitive dissonance, which describes how people are emotionally averse to cognitive inconsistencies and seek to reduce them. Economics also offers some application of cognitive dissonance (for a short description see Rizzello 1999: 80–81). 12 Reber (1993: 148–149) emphasizes that the existence and the relevance of some forms of nativism in the development of mind/brain is almost uniformly recognized, and that no one today defends a pure empiricism in the sense of Locke’s tabula rasa.

Introduction 15 13 The notion of neurognosis is linked to the idea that knowledge is constructed in cognitive frameworks. This aspect is not new at all in the literature on organization and management of ‘absorptive capacity’ (see Cohen and Levinthal 1990). From this perspective, firms have been theorized as ‘sensemaking systems’, a ‘focusing device’, ‘systems of shared meanings’ (Smircich 1983; Weick and Roberts 1993; Weick 1995; Choo 1998; Nooteboom 1999). 14 In this context it is interesting to point out that this idea of mental mechanisms, seen in path-dependent terms as resistance to change, is quite similar to the theory of personality, proposed by George Kelly (1995), which was based on the idea that interpretative systems are resistant to change and might be overwhelmed by major departures from familiar circumstances. 15 A famous example in this literature concerns penguins’ wings. Usually wings arose in birds in the context of flight, but it was an exaptation process that conferred advantages to penguins in swimming (Gould and Vrba 1982). 16 This problem is linked to another well-known problem in organizational literature, that of combining ‘exploitation and exploration’ (March 1991). 17 In two recent papers Caldwell (2000) and Horwitz (2000) underlined once more the relevance of The Sensory Order for the understanding of the Hayekian conception of evolution and for the spontaneous construction of a liberal order, respectively. Caldwell, in particular, describes the centrality of this book in the emergence of Hayek’s ideas on evolution, demonstrating the link between the human mind and the role of cultural evolution (on this issue see also Rizzello [2000] and Rizzello and Turvani [2000]). 18 A more comprehensive description can be found in Rizzello (1999). 19 Gould (1991) has stressed the relevance of exaptation for evolutionary psychology. 20 See also Paul David’s reaction to the criticism by Leibowitz and Margolis (David 1997). 21 ‘When a dynamical economic system enters such a region, it cannot escape except through the intervention of some external force, or shock, that alters its configuration or transforms the underlying structural relationships among the agents’ (David 1997: 34). 22 For a complete overview, see Wilson and Keil (1999). Liepert et al. (2000) offer, among other things, an empirical confirmation of the neuronal plasticity, of the continuous transformation and reorganization of the brain, and of its physiological sensibility to the experience. 23 The social cognitive learning theory (Bandura 1986) seems to offer good analytical tools in understanding the behavioural foundation of cultural evolution, as suggested by Witt (2000) and Rizzello and Turvani (2002).

References Alchian, A. (1950) ‘Uncertainty, Evolution and Economic Theory’, Journal of Political Economy, 58: 211–221. Andersen, E. (1997) ‘Neo and Post-Schumpeterian Contributions to Evolutionary Economics’, in J. Reijnders (ed.), Economics and Evolution, Cheltenham: Edward Elgar. Antonelli, C. (1995) ‘Dynamic Structuralism and Path Dependence: Industrial Economics in Italy’, Revue d’Econonomie Industrielle, 73: 65–90. —— (1999) The Microdynamics of Technological Knowledge, London: Routledge.

16 Salvatore Rizzello Bandura, A. (1986) Social Foundations of Thought and Action. A Social Cognitive Theory, Englewood Cliffs, N.J.: Prentice-Hall. Boulding, K. (1956) The Image. Knowledge in Life and Society, Ann Arbor: The University of Michigan Press. —— (1992) ‘Cybernetics in the Evolutionary Process’, in K. Boulding, Towards a New Economics. Critical Essays on Ecology, Distribution and Other Themes, Aldershot: Edward Elgar. Butos, W. and Koppl, R. (1997) ‘The Varieties of Subjectivism: Keynes and Hayek on Expectations’, History of Political Economy, 29(2) (Summer): 327–359. Caldwell, B. (2000) ‘The Emergence of Hayek’s Ideas on Cultural Evolution’, Review of Austrian Economics, 13: 5–22. Choo, C.W. (1998) The Knowing Organization, Oxford: Oxford University Press. Cohen, M.D. and Levinthal, D. (1990) ‘Absorptive Capacity: A New Perspective on Learning and Innovation’, Administrative Science Quarterly, 35: 128–152. David, P. (1985) ‘Clio and the Economics of QWERTY’, American Economic Review, 75: 332–337. —— (1997) ‘Path-Dependence and the Quest for Historical Economics: One More Chorus of the Ballad of QWERTY’, Discussion Paper in Economic and Social History, Oxford: University of Oxford Press. Denzau, A. and North, D. (1994) ‘Shared Mental Models: Ideologies and Institutions’, Kyklos, 47(1): 3–31. Dosi, G. (1991) ‘Some Thoughts on the Promises, Challenges and Dangers of an “Evolutionary Perspective” in Economics’, Journal of Evolutionary Economics, 1: 5–7. Dosi, G. and Kaniovski, Y. (1994) ‘On “Badly Behaved” Dynamics’, Journal of Evolutionary Economics, 4: 93–123. Dosi, G. and Nelson, R. (1994) ‘An Introduction to Evolutionary Theories in Economics’, Journal of Evolutionary Economics, 4(3): 153–172. Foss, N. (1997) ‘Evolutionary Economics and the Theory of the Firm: Assessments and Proposal for Research’, in J. Reijnders (ed.), Economics and Evolution, Cheltenham: Edward Elgar. Foster, J. (1993) ‘Economics and the Self-organization Approach: Alfred Marshall Revisited?’, Economic Journal, 103 (July): 975–991. —— (1994) ‘The Self-organizing Approach in Economics’, in P. Burley and J. Foster (eds), Economics and Thermodynamics, Boston, Dordrecht and London: Kluwer Academic Publishers. —— (1997) ‘The Analytical Foundations of Evolutionary Economics: from Biological Analogy to Economic Self-organization’, Structural Change and Economic Dynamics, 8: 427–451. —— (2000) ‘Competitive Selection, Self-organization and Joseph Schumpeter’, Journal of Evolutionary Economics, 10(3): 311–328. Gould, S.J. (1991) ‘Exaptation: a Crucial Tool for an Evolutionary Psychology’, Journal of Social Issues, 47(3): 43–65. Gould, S.J. and Vrba E.S. (1982) ‘Exaptation – a Missing Term in the Science of Form’, Paleobiology, 8: 4–15. Hayek, F.A. (1937) ‘Economics and Knowledge’, Economica, n.s. IV(13): 96–105. —— (1942) ‘Scientism and the Study of Society’, Economica, n.s. IX(35): 267–291. —— (1952) The Sensory Order. An Inquiry into the Foundations of Theoretical Psychology, London: Routledge & Kegan Paul.

Introduction 17 —— (1963) ‘Rules, Perception and Intelligibility’, Proceedings of the British Academy, XLVIII: 321–344. —— (1973) Law, Legislation and Liberty. A New Statement of the Liberal Principles of Justice and Political Economy, vol. I, Rules and Order, London: Routledge & Kegan Paul. —— (1979) Law, Legislation and Liberty. A New Statement of the Liberal Principles of Justice and Political Economy, vol. III, The Political Order of a Free People London: Routledge & Kegan Paul. —— (1994) Hayek on Hayek. An Autobiographical Dialogue, London: Routledge. Hodgson, G. (1995) ‘The Evolution of Evolutionary Economics’, Scottish Journal of Political Economy, 42(4): 153–172. —— (1997a) ‘Economics and the Return to Mecca: the Recognition of Novelty and Emergence’, Structural Change and Economic Dynamics, 8: 399–412. —— (1997b) ‘The Evolutionary and Non-Darwinian Economics of Joseph Schumpeter’, Journal of Evolutionary Economics, 7(2): 131–145. —— (1999) Evolution and Institutions. On Evolutionary Economics and the Evolution of Economics, Cheltenham: Edward Elgar. Horwitz, S. (2000) ‘From The Sensory Order to the Liberal Order: Hayek’s Nonrationalist Liberalism’, Review of Austrian Economics, 13: 23–40. Kahneman, D. and Tversky, A. (1979) ‘Prospect Theory: an Analysis of Decision under Risk’, Econometrica, 47: 263–291. Kelly, G. (1955) The Psychology of Personal Constructs, New York: Norton. Knight, F.H. (1921) Risk, Uncertainty and Profit, Boston: Houghton Mifflin Company. Lane, D.A. (1993a) ‘Artificial Worlds and Economics: Part I’, Journal of Evolutionary Economics, 3: 89–107. —— (1993b) ‘Artificial Worlds and Economics: Part II’, Journal of Evolutionary Economics, 3: 177–197. Laughlin, C. (1996) ‘The Properties of Neurognosis’, Journal of Social and Evolutionary Systems, 19(4): 363–380. Leibowitz, S.J. and Margolis, S.E. (1995) ‘Path Dependence, Lock-In and History’, Journal of Law, Economics and Organization, April: 205–226. Lesourne, J. (1997) ‘Early Signs of a Revolution in Microeconomics’, in K. Dopfer (ed.), Evolutionary Concepts in Economics, Dordrecht: Kluwer Academic Publishers. Liepert, J., Bauder, H., Miltner, W.H., Taub, E. and Weiller, C. (2000) ‘TreatmentInduced Cortical Reorganization After Stroke in Humans’, Stroke, 31 (June): 1210–1216. Luna, F. and Perrone, A. (eds) (2001) Agent-Based Methods in Economics and Finance: Simulations in Swarm, Dordrecht and London: Kluwer Academic Publishers. March, J.C. (1991) ‘Exploration and Exploitation in Organizational Learning’, Organization Science, 2(1): 71–87. March, J.C. and Simon, H.A. (1958) Organizations, New York: Wiley. Marshall, A. (1867–8) ‘Ye Machine’, reprinted in T. Raffaelli (1994): 116–132. —— ([1871] 1961) Principles of Economics (9th edn), London: Macmillan. Menger, C. (1883) Untersuchungen ueber die Methode der Socialwissenschaften der politischen Oekonomie insbesondere, Leipzig: Dunker Humblot. Mirowski, P. (1983) ‘An Evolutionary Theory of Economic Change: A Review Article’, Journal of Economic Issues, 17: 757–768.

18 Salvatore Rizzello Momigliano, F. (1975) Economia industriale e teoria dell’impresa, Bologna: Il Mulino. Mori, H. and Kuramoto, Y. (1998) Dissipative Structures and Chaos, Berlin: Springer. Nelson, R.R. (1995) ‘Recent Evolutionary Theorizing about Economic Change’, Journal of Economic Literature, 33(1): 48–90. Nelson, R.R. and Winter, S.G. (1982) An Evolutionary Theory of Economic Change, Cambridge, Mass.: Harvard University Press. Nooteboom, B. (1999) ‘Innovation, Learning and Industrial Organization’, Cambridge Journal of Economics, 23(2): 127–150. Raffaelli, T. (1994) ‘Alfred Marshall’s Early Philosophical Writings’, Research in the History of Economic Thought and Methodology: Archival Supplement, no. 4: 53–159. Reber, A. (1993) Implicit Learning and Knowledge. An Essay on the Cognitive Unconscious, New York and Oxford: Oxford University Press. Rizzello, S. (1997) ‘The Microfoundations of Path-dependency’, in L. Magnusson and J. Ottosson (eds), Evolutionary Economics and Path-dependence, Cheltenham: Edward Elgar. —— (1999) The Economics of the Mind, Aldershot: Edward Elgar. (Italian version (1997) L’Economia della mente, Roma: Laterza.) —— (2000) ‘Economic Change, Subjective Perception and Institutional Evolution’, Metroeconomica, 51(2): 127–150. Rizzello, S. and Turvani, M. (2000) ‘Institutions Meet Mind: The Way out of an Impasse’, Constitutional Political Economy, 11: 165–180. —— (2002) ‘Subjective Diversity and Social Learning: A Cognitive Perspective for Understanding Institutional Behavior’, Constitutional Political Economy, 13: 201–214. Silverberg, G., Dosi, G. and Orsenigo, L. (1988) ‘Innovation, Diversity and Diffusion. A Self-Organization Model’, Economic Journal, 98: 1032–1054. Simon, H.A. (1957) Models of Man, New York: Wiley. Smircich, L. (1983) ‘Concepts of Culture and Organizational Analysis’, Administrative Science Quarterly, 28: 339–385. Sperber, D. and Hirschfeld, L. (1999) ‘Culture, Cognition, and Evolution’, in R.A. Wilson and F.C. Keil (eds) MIT Encyclopedia of Cognitive Sciences, Cambridge, Mass.: MIT Press: cxi–cxxxii. Vanberg, V. (1994) Rules and Choice in Economics, London and New York: Routledge. Varela, F.J. (1979) Principles of Biological Autonomy, New York: Elsevier NorthHolland. Weick, K.E. (1995) Sensemaking in Organizations, Thousand Oaks, Calif.: Sage. Weick, K.E. and Roberts, K. (1993) ‘Collective Mind in Organizations: Heedful Interrelating on Flight Desk’, Administrative Science Quarterly, 38: 357–381. Wilson, R.A. and Keil, F.C. (eds) (1999) The MIT Encyclopedia of the Cognitive Sciences, Cambridge, Mass.: Cambridge University Press. Witt, U. (1991) ‘Reflection on the Present State of Evolutionary Economic Theory’, in G. Hodgson and E. Screpanti (eds), Rethinking Economics: Markets, Technology and Economic Evolution, Aldershot: Edward Elgar. —— (1992) ‘Evolutionary Concepts in Economics’, Eastern Economic Journal, 18(4): 405–419.

Introduction 19 —— (1993) ‘Path-Dependence in Institutional Change’, Papers on Economics & Evolution, #9306, Jena: Max Planck Institute for Research into Economic Systems. —— (1995) ‘Schumpeter vs. Hayek: Two Approaches to Evolutionary Economics’, in G. Meijer (ed.), New Perspectives on Austrian Economics, London and New York: Routledge. —— (1996) ‘A “Darwinian Revolution” in Economics’, Journal of Institutional and Theoretical Economics, 152: 708–715. —— (1997) ‘Self-organization and Economics – What is New?’, Structural Change and Economic Dynamics, 8: 489–507. —— (1999) ‘Evolutionary Economics and Evolutionary Biology’, in P. Koslowski (ed.), Sociobiology and Bioeconomics. The Theory of Evolution in Biological and Economic Theory, Berlin: Springer. —— (2000) ‘Social Cognitive Learning and Group Selection – A Game-theoretic Version of Hayek’s Societal Evolution’, Paper presented at the INEM–ASSA Session ‘Austrian Economics and Game Theory’, Boston, 8 January.

1

Discrepancies: competing theories and ideologies as cognitive traps The scientific discovery as puzzlesolving1 Massimo Egidi

Introduction In some contexts, like for example complex games and puzzles, the search for solutions to problems leads to the discovery of different procedures, none of which can be considered ‘the best’ one. In these contexts, in fact, the different solutions can be compared only in some specific well-known domain of application, while in other domains the comparison is incomplete or vague, or the domain may have imperfectly known boundaries; it is therefore impossible to put the different procedures in a precise preference order over the entire domain of applicability. Moreover, these procedures may be ‘locally stable’ because small local changes in the instructions that define them do not lead to any improvement, and therefore individuals that discover one solution may remain locked in it without trying to search for alternative solutions. The search for solutions to puzzles shows striking analogies with the search for new theoretical approaches that take place when a consolidated theory fails to explain new phenomena. The properties of search in puzzlesolving – multiplicity, local stability and incompleteness of solutions – also affect the new theoretical approaches (i.e. the new solutions to a scientific puzzle that emerge to challenge a dominant theory). Interestingly, the state of the art of new theoretical proposals in the field of decision-making interestingly illustrates this situation. After the numerous violations of traditional expected utility theory discovered since the experiments by Maurice Allais in the 1950s (Allais and Hagen 1979), new proposals, such as regret theory and others, have contended with it for the status of the ‘right’ theory. To date, comparisons among the competing theoretical proposals have singled out none of them as unequivocally preferable (Hey 1991). This situation, which has persisted for many years, prompts explanations which work in two directions: on the one hand, it suggests that epistemological elements relating to the existence and permanence of

Theories and ideologies as cognitive traps

21

competing theories should be rethought; on the other, it suggests that cognitive aspects of human thinking should be examined in order to explain why alternative solutions to a problem may persist and stabilize and thereby provide a cognitive foundation for decision theory. The chapter explores both questions. The first four sections examine some epistemological consequences of the coexistence of different theories, suggesting a revision of the notions of ‘error’ in these contexts, mainly with reference to Popper’s and Lakatos’s methodological approaches. The remaining sections analyse the process by which players in complex games construct solutions – by editing the problem, decomposing it into basic building blocks, and defining the categories that represent the problem. Experiments show that players may discover different solutions according to the training that they have received, and that they may remain locked in these solutions even though this proves to be suboptimal. This lock-in effect is explained in terms of ‘routinized thinking’, and it is shown to be due to imperfect categorization of the problem. In order to discover new solutions, individuals must redefine the categories with which they describe the problem; this requires a process of abstraction and specification involving the old categories, and allowing the creation of new basic categories. It will be shown that this process is constrained and driven by the emergence of unexpected exceptions, and therefore that, given the random emergence of anomalies, the reconstruction of new categories is intrinsically biased. The search for solutions is therefore described as an adaptive process driven by perceived errors and essentially based on prejudices and their revision. These results are closely related to Popper’s and Lakatos’s views of the creation of knowledge, and they also entail that a new definition must be given to rationality in evolutionary contexts, a definition that will be discussed in the conclusions.

Human decisions that deviate systematically from optimal behaviour In recent decades, extremely fruitful reconciliation has taken place between economics and psychology, inducing the former to accept more stringent criteria – compared to those of the past – of empirical validation. These criteria are based on recognition of the relevance of experiments which, especially in the field of individual decision-making, have led to a rethinking of the role of decision theory after a time in which this theory, and in particular expected utility theory, was largely ascribed a normative role as ‘logic of action’. As Langlois (1998) writes: Although we may trace this tendency to Menger . . . , it was probably Lionel Robbins’s Nature and Significance of Economic Science (1932) that fully ensconced in the minds of economists the idea that their

22 Massimo Egidi science is about the logic of means and ends rather than about the psychology of utility. (Langlois 1998: 57) This view of the economic discipline had a particularly critical shortcoming: it assumed that the majority of individuals always behave according to rational strategies, disregarding the limits to agents’ rationality. By introducing the notion of ‘bounded rationality’ at the beginning of the 1960s, H. Simon emphasized that it is unrealistic to attribute full rationality to decision-makers when the computation of their strategies is complex and requires great skill (which we cannot as a rule attribute to all individuals). The most notable and most successful attempt to overcome this difficulty was made by Milton Friedman in the 1950s, with his proposal of the ‘as if’ assumption. Friedman (1953) claimed that the large majority of individuals in economic institutions behave according to the fully rational strategies formulated by expected utility theory, even if they do not possess the necessary calculation abilities. They do so because competition induces them to behave ‘as if’ they know the best course of action. Individuals learn optimal behaviours by trial and error; they ‘discover’ increasingly efficient strategies because of the effect of competition, which favours those subjects whose behaviour comes closest to the optimal one. On this view it was implicitly assumed that individual deviations from optimal behaviour within a population were ‘errors’ distributed according to the Gaussian distribution. Because it was presumed that those adopting inefficient strategies would be progressively eliminated by economic competition, behaviours were supposed to concentrate around the optimal one. As economic theory attributed ever-greater importance to expectations, Friedman’s ‘as if’ assumption relegated observation of the mental processes underlying economic decisions to a very marginal position – despite the fact that the expectations that play a fundamental role in Friedman’s view are originated by agents on the basis of their mental representations of the economy. The most important challenge against this account was raised by March and Simon’s organizational studies, which sought to understand how human rationality effectively operates by conducting empirical observation of managers’ behaviours, expectations and opinions (see March and Simon 1958). Despite the strong emphasis on the importance of observing behaviours rather than mental processes, in the years that followed Simon developed methodologies with which to observe and simulate the mental processes involved in decision-making and, in particular, strategy-building. By gradually shifting the focus from real organizational contexts to ‘artificial’ environments like the game of chess, Simon proposed with Newell (1958, 1962),2 an analysis of strategic action which, on the one

Theories and ideologies as cognitive traps

23

hand, gave rise to the theory of ‘problem-solving’, and, on the other, served as a platform for the empirical observation of players’ decisions and thoughts. The game of chess was chosen for experimentation because it required a very high capacity for strategic calculation and could thus be used to establish the limits of human rationality and of computation in artificial programs. In the 1970s, Simon developed his Protocol Analysis to investigate the problem-solving activities of players engaged in a game. An experimenter employing this methodology records the symbolic mental activity of a chess player by asking him to describe his thoughts in detail as he constructs his strategies. Empirical research in the directions opened up by Simon has demonstrated that the ‘as if’ assumption is untenable because its main argument does not stand up to the facts. Indeed, it can be shown that in conditions of high uncertainty, players’ strategies are not distributed around a single optimal strategy. Rather, they are fully differentiated, so that it is impossible to identify the optimal strategy by means of competition: in fact, competition in tournaments does not elicit the best strategy. Analysis of chess, and of the way in which the players construct their strategies, thus introduces two well-known relevant aspects. The first, as is well known, is that in many circumstances it is not possible to calculate the optimal strategy, given the computational complexity of the problem.3 The second is that the strategies chosen by the players differ greatly: there is, for example, a wide variety of openings which cannot be compared in terms of optimality. A first significant consequence thus emerges: even if an optimal strategy does exist, we are at present unable to determine which opening comes ‘closest’ to the optimal strategy. Masters and skilled players choose from a wide variety of openings, without there being a preference order among them. Hence, even if we were to record the strategies used by the grandmasters who win international tournaments, we could not determine what the optimal strategy is. Consequently, one assumption implicit in Friedman’s position is no longer valid; namely, that real behaviours are distributed ‘normally’ around a behaviour taken as optimal by the theory. On the contrary, what we observe, at least in sufficiently complex games, are systematic and permanent discrepancies amongst the various strategies used by the players. This phenomenon is very pervasive and can be found in many other fields where human decisions are subjected to empirical verification. Systematic and permanent discrepancies from the behaviour predicted by the theory have been revealed by experiments on deduction, reasoning and choice in conditions of uncertainty. The experimental study of these activities is therefore important if we wish to understand how individuals develop their strategies and apply

24 Massimo Egidi them to real contexts. Cognitive psychology, and the study of learning in particular, have thus become frames of reference for the study of rational choice; for in order to understand how decisions are taken it is necessary to understand how human beings acquire information and knowledge, and how they use them to build up their strategies. The highly differentiated behaviour observed in many important experimental situations has given rise, since the 1970s, to further theoretical proposals for explaining human decision-making. Regret theory and other proposals were first attempts to establish decision-making theory on new and different bases. As we noted at the beginning, empirical comparisons of the predictions of these new theories have not convincingly shown that one of them has greater explanatory capacity than the others: each of them makes fairly accurate predictions in some experimental areas but fails to make good predictions in others (Hey 1991). Each of the new theories therefore has a limited domain of validity. Moreover, on the one hand there are overlaps between the domains in which the predictions of individual theories prove to be accurate, while on the other there are domains in which no single theory furnishes satisfactory predictions. Attempts to construct a new theory of decisions on the same epistemological criteria that defined expected utility theory, by slightly modifying some of its axioms, have failed to achieve clear and definitive results. The proposed generalization of expected utility theory, where some axioms are weakened or replaced, is still based on the epistemological assumption that decision theory is a ‘logic of action’. As said, this was the position defended from Robbins to Friedman, on the belief that it is possible to understand, and eventually to predict, real decisions, regardless of the psychological features of mental activity. I submit that these attempts have been unsuccessful because they ignore a profound characteristic of decision-making activity: the interdependence between the decision-making process and the mental representation of the elements that give rise to the decision. The progress made in understanding the main cognitive processes – induction, learning, categorization, framing, etc. – involved in human decision-making suggests that clearer light can be shed on the process by investigating cognition. Little modifications of the axioms of the expected utility theory seem unable to obtain the success desired, or to hold out the prospect of a future overall theory of decision-making, because they ignore the complexity of the underlying psychological phenomena. The cognitive processes involved in decision-making should not be ignored because in many respects they determine the decision itself. The foremost example of this connection has been provided by Kahneman and Tversky’s experiments on ‘framing effects’, the best-known of which, verified by an extremely wide range of experiments, shows that individuals are averse or favourable to the risk inherent in a decision according to how

Theories and ideologies as cognitive traps

25

this decision is presented to them. If it is presented in such a way that they codify it as a loss situation, they are favourable to risk, and, vice versa, they are averse to risk if they perceive it as a gain. A potentially successful research strategy is therefore one which acknowledges the importance of the mental representations of the elements on which individuals decide.4 In order to understand the individual decision-making process, therefore, it is essential to understand how the elements of the decision are codified and represented in a ‘mental model’, and how knowledge is organized by individuals and used in decisionmaking. The most promising point of departure is thus the field of problem-solving, reasoning, representation and categorization. This chapter will not attempt to address these topics, which are extremely broad in their scope. Rather, it will concentrate on a point that appears to have a close bearing on how the representation of problems can be characterized. It will examine the problem of ‘cognitive traps’, i.e. those situations in which different individuals faced with the same problem discover and adopt different solutions and persist in the use of those solutions even when they prove to be inefficient.

Competing solutions in problem-solving and competing theoretical approaches I have suggested that competing theories with regard to the same domain display many similarities with a situation that typically arises in problemsolving. Complex problems may be amenable to numerous alternative solutions, and comparison among them is incomplete or excessively demanding in computational terms, and is therefore ‘practically’ imprecise. As previously mentioned, this is the distinctive feature of the game of chess, which is an ideal artificial setting for the investigation of decisionmaking processes in complex situations, as well as for evaluating the epistemological implications of situations in which there are numerous different solutions to a problem which cannot be ordered in terms of their greater or lesser efficiency. When the ‘best course of action’ cannot be computed and cannot be tested, and when there is a variety of possible strategies that can be only ‘vaguely’ compared in terms of their efficiency, we are in the particularly interesting situation of what we may call ‘procedural uncertainty’ (Dosi and Egidi 1991). This, moreover, is the situation typical of high-level managerial decisions in evolutionary contexts.5 The multiplicity of solutions to a complex game – solutions that are supposedly difficult to discover and to compare in terms of efficiency – prompts three related questions: 1

The lock-in question: what are the elements that prevent or at least make it difficult for a player to escape from a suboptimal solution and

26 Massimo Egidi

2

3

search for more efficient procedures? This point, which links with the question of routinization, will be explored later (see pp. 42–48) with the support of experimental data. Assuming that alternative solutions have been discovered, what are the factors that make comparison among different solutions difficult or impossible? We shall see that there are two main possibilities: (a) the computational and statistical difficulty of comparing two solutions in the same domain; (b) the impossibility of exploring all domains of application, and in consequence the unexpected increase in exceptions, i.e. never previously explored points in the domain where a solution fails. What is the epistemological status of ‘errors’ and ‘biases’ when it is impossible to determine the best solution?

These three questions also arise when the more general issue of competing theories is addressed. In fact (point 2), when there are several competing theories – as happens in the field of decision-making – it may be difficult to determine which of them best explains all the facts under examination, and it is only the unexpected appearance of exceptions that lets individuals revise their theoretical approaches. Here again (point 1), either it may happen that, starting from a multiplicity of theoretical proposals, a transition process leads to convergence on a single (temporarily) dominant theory, or, on the contrary, the differences among approaches become permanent, which seems to be the typical characteristic of clashing ideologies. Finally (point 3), it is difficult to establish the significance of such terms as ‘error’ or ‘deviation’, given that a dominant theory to which to refer does not exist in this case. We start the discussion from point 3. What is meant by the word ‘error’? The question is intriguing because of its epistemological status. In many fields the very notion of ‘error’ as deviation from the ‘truth’ is contested. Politics is a typical domain in which there are frequently wide discrepancies among the strategies of decision-makers: democracy is an institution rooted in the principle that individuals usually have different and contrasting opinions on the same subject and that all opinions have the same legitimacy. A theory of rational political decision-making cannot assume that all citizens share the same view of society, and that the differentiation amongst political strategies originates only from a different ordering of preferences. The differentiation lies at a deeper level and involves non-shared principles and ‘visions’. Deviation is here considered to be a legitimate and natural difference of opinion on the same problem, so that no one is surprised that differences in visions may become permanent, also giving rise to opposed ‘ideologies’. The question of divergence between different solutions to the same problem, or between differing theories offering alternative explanations

Theories and ideologies as cognitive traps

27

for the same group of facts, is therefore particularly delicate when it is applied to social phenomena. Hayek and Popper confronted this problem, the crux of which is how the social division of knowledge occurs, and how either recognition of a single shared theory or the permanent opposition of theories comes about. Their position, to which I shall return in the following section, is that only if the respective proposals are subjected to a continuous critical process will the community be able to secure – temporarily – a certain amount of convergence in some areas of its beliefs and knowledge, and to discriminate, when possible, between different theoretical proposals. This raises a background question: what is the difference between an authoritarian and a pluralist conception of the sources of knowledge?

Deviation as error: an authoritarian conception of knowledge What are the origins of the authoritarian view of knowledge? One of its principal sources lies in Plato’s thought. For Plato the origin of our knowledge is divine; in his celebrated theory of anamnesis, set out in Meno, he claims that when we know we simply remember what our immortal soul already knew prior to our birth: knowledge is therefore re-cognition. Since this myth attributes man with the capacity to achieve the truth, Karl Popper (1960) considers it an optimistic account of the sources of knowledge. But he notes that Plato’s vision later grew pessimistic: Yet disappointment must have come to Plato; for in The Republic (and also in the Phaedrus) we find the beginnings of a pessimistic epistemology. In the famous story of prisoner in the cave he shows that the world of our experience is only a shadow, a reflection, of the real world. And he shows that even if one of the prisoners should escape from the cave and face the real world, he would have insuperable difficulties in seeing and understanding it . . . The difficulties in the way of an understanding of the real world are almost super-human, and only very few, if anybody at all, can attain to the divine state of understanding the real world – the divine state of true knowledge, of episte¯me¯ . . . Thus we find in Plato the first development from an optimistic to a pessimistic epistemology. Each of them form the basis of one of the two diametrically opposed philosophies of the state and of the society: on the one hand an anti-traditionalist, anti-authoritarian, revolutionary and Utopian rationalism of the Cartesian kind, and on the other hand an authoritarian traditionalism. (Popper 1960: 48)

28 Massimo Egidi According to Popper, it was the dispute between empiricism and rationalism that crucially freed philosophers from the need to acknowledge God as the authority guaranteeing the validity of human knowledge: The problem which I wish to examine . . . may perhaps be described as an aspect of the old quarrel between the British and the Continental schools of philosophy – the quarrel between the classical empiricism of Bacon, Locke, Berkeley, Hume and the classical rationalism or intellectualism of Descartes, Spinoza and Leibniz. In this quarrel the British school insisted that the ultimate source of all knowledge was observation, while the Continental school insisted that it was the intellectual intuition of clear and distinct ideas. (Popper 1960: 40) The Continental school assumed that truth could be discovered through the use of reason; Descartes based his epistemology on the optimistic idea that what we understand clearly may not be true, because otherwise God would be deceiving us. The Cartesian doctrine held that our intellect does not deceive us if we are able to formulate clear and distinct ideas, since intellect is created by God, the source of all knowledge. This is the doctrine of Veracitas Dei. The Cartesian position had immense emancipatory power in Western culture because it transferred to man, and thus to his own reason, the authority on which the validity of knowledge was based. God was still the final authority, but it was no longer necessary to consult the Scriptures in order to attain truth; instead, one could appeal to the human intellect. The British school attributed the source of knowledge to empirical observation. Bacon considered induction to be the key method with which to infer general rules and laws from observation of specific elements of experience. Here again the question of the certainty of knowledge was crucial: Bacon claimed in the Novum Organum that if individuals were to be able to ‘read the book of Nature’, they had to purify their minds from the prejudices and conjectures that could cast a veil over their minds and prevent them from achieving the truth. When our minds have been purged of false beliefs and prejudices, we are ready to read the book of nature, i.e. to induce general laws from the facts of experience. Bacon envisaged two methods of induction: the true method, ‘interpretatio naturae’, which I have illustrated thus far; and the false method, ‘anticipatio mentis’, which was a method based on prejudice. Popper notes: This latter method, rejected by Bacon, is in fact a method of interpretation, in the modern sense of the word. It is the method of conjecture or hypothesis (a method of which, incidentally, I happen to be a confirmed advocate). (Popper 1960: 52)

Theories and ideologies as cognitive traps

29

In some sense, Popper tells us that Bacon’s idea of purging the mind of prejudices is not a prerequisite for the achievement of knowledge but the true essence of the process of acquiring knowledge. As Thornton (1997) remarks: Scientific theories, for him, are not inductively inferred from experience, nor is scientific experimentation carried out with a view to verifying or finally establishing the truth of theories; rather, all knowledge is provisional, conjectural, hypothetical – we can never finally prove our scientific theories, we can merely (provisionally) confirm or (conclusively) refute them; hence at any given time we have to choose between the potentially infinite number of theories which will explain the set of phenomena under investigation. Faced with this choice, we can only eliminate those theories which are demonstrably false, and rationally choose between the remaining, non-falsified theories. Hence Popper’s emphasis on the importance of the critical spirit to science – for him critical thinking is the very essence of rationality. For it is only by critical thought that we can eliminate false theories, and determine which of the remaining theories is the best available one, in the sense of possessing the highest level of explanatory force and predictive power. This view of rationality is a fundamental aspect of Popper’s epistemology. In the next section of this chapter I shall compare this view, suggesting that knowledge is based on the critical capacity to revise and modify our conjectures, with experiments on human problem-solving which have shown that revising the solution to a problem and searching for a new one are based on imperfections in induction and categorization. In some sense, I shall undertake an exercise in ‘experimental epistemology’.

Analogies between problem-solving and building-up theories It is well known that when individuals solve a problem, they decompose it into in sub-problems on the basis of ‘heuristics’. In the next section (pp. 33–38) I shall show that – at least for some classes of problems – heuristics are generated on the ground of a representation of the problem. The revision of a given solution requires a different and new representation and categorization of the problem which display striking similarities with the way in which scientists revise theories as the consequence of emerging anomalies. Before exploring these similarities, which are rooted in the cognitive aspects of intellectual activity, I would emphasize one difference: the fact that the scientific enterprise is based on a social division of knowledge, while problem-solving is or may be an individual activity. The question is

30 Massimo Egidi therefore how the community verifies and validates new emerging ideas and theoretical approaches, given the wide division of knowledge and the limited domain of a single scientist’s expert knowledge. This question again elicits the contrast between the authoritarian and the critical role of scientific activity briefly discussed in the following sub-sections. Settling controversies within the scientific community One wonders whether the process of critical revision that, following Popper, should ensure serious comparison among theories really takes place in the scientific community. An observer able to participate directly in these activities would note many dissimilarities between the Popperian protocol and actual practice. These differences are due not only to a lack of completeness and rigour in practical procedures, but also to the nature of the problem. Who is it that guarantees the convergence of different opinions and assessments in the scientific community on a single, shared opinion? Following Popper, we can only hope that all the critical aspects of the theory debated in the scientific community have been raised and solved by reducing disagreements as far as possible. This is a modest requirement, but it is the only one that can be fulfilled, because more stringent tests have shown serious shortcomings. How to eliminate disagreements by reducing all scientific statements to a formal system in which all statements can be compared via pure computation was a central issue of scientific debate at the end of the nineteenth century. The solution proposed was to axiomatize the scientific disciplines, which meant reducing the scientific statements of a discipline, say physics, to logical derivations from a limited number of axioms. Some scholars suggested that this could be done by creating a language to describe scientific statements rigorously and non-ambiguously, so that the scientific community could communicate, evaluate problems and reach a shared opinion. Given that mathematics plays an essential role of in all scientific disciplines, it was the first candidate for formalization. This was the project undertaken by Giuseppe Peano, whose Formulario Matematico (1894) set out a meta-language for converting any discussion on mathematical issues into a formal calculus. By means of this cryptic but – for experts – ‘clear and distinct’ meta-language, pure mathematicians would supposedly acquire and share mathematical truths simply via calculus. But besides the difficulties to learn and use Peano’s meta-language, it was no guarantee that scientific debates could be resolved by means of pure calculus. This non-guarantee is a major property of formal systems discovered by Kurt Gödel, who, in his famous ‘On Formally Undecidable Propositions in Principia Mathematica and Related Systems’ (1931), demonstrated that within a formal system there exist questions that are

Theories and ideologies as cognitive traps

31

neither provable nor disprovable on the basis of the axioms that define that system. Thus, the scientific community contents itself with a ‘reasonably thorough’ but not fully formalized language – which in any case offers no protection against making errors. Since the 1930s the members of a scientific community have therefore abandoned the use of a meta-language, reinforcing the practice of the critical interaction to avoid discrepancies and errors. However, there is still a wide margin for unexpected errors to occur, as the following story illustrates. Around 1630 the French mathematician Pierre August de Fermat conjectured that the equation xn
yn  zn has no integer solutions for x, y and z (different from zero) when n  2. He wrote: ‘I have discovered a truly remarkable proof which this margin is too small to contain.’ The conjecture, called Fermat’s Last Theorem, resisted any attempt to prove (or falsify) it for four centuries. In 1993 Andrew Wiles submitted a highly complex proof of the conjecture to The Annals of Mathematics. Validating his demonstration took hundreds of mathematicians more than one year of work, during which many errors were discovered and corrected. But even if every reviewer has thus far agreed with the corrections, no one can guarantee that a new defect will not appear in the proof. O’Connor and Robertson (1997) give an interesting reconstruction of the facts, and in conclusion they write: No proof of the complexity of this can easily be guaranteed to be correct, so a very small doubt will remain for some time. However, when Taylor lectured at the British Mathematical Colloquium in Edinburgh in April 1995 he gave the impression that no real doubts remained over Fermat’s Last Theorem. This story is an interesting example of how human knowledge is constructed and acquired through a method that displays fallibility and opinion discrepancies in many ways. Discrepancies may be at least partially irreducible due to the complex activity required to converge towards a shared system of opinions; on the one hand, there is in fact an heavy individual mental effort required of each of the scientists involved in the process, in his limited domains of knowledge; on the other hand, there is the need to reduce disagreements among the members of a large group, who only partially and imperfectly share a common language and a common knowledge. Sociological aspects: how the authoritarian concept of knowledge filters into conventions and routines Since knowledge is imperfectly shared among experts, and the extension of the individual’s domains of knowledge is limited, in many conditions the

32 Massimo Egidi power of the critical, Galilean method, on which modern science is founded, is severely weakened. An overall application of the critical method only takes place in specialist contexts, and this seriously jeopardizes the development of knowledge. Indeed, in a highly segmented, hierarchical structure of knowledge like that of modern society, the ‘authoritarian’ nature of knowledge processes constantly reproduces itself. It resides in the ‘automatic’ way in which we learn the majority of our experiences, without subjecting them to critical discussion. We accept the authority of anyone deemed more competent than we are, and we often accept new notions and ideas without discussion or argument, even if we are not convinced by them, given our haste to acquire the skills that the modern world demands of us: we cannot critically analyse everything that is propounded to us. The huge division of mental labour that characterizes society counsels us not to ‘waste time’ on discussion of each and every new idea or notion: we are secure in the knowledge that what we learn is true – because it has been subjected to critical analysis by experts, who have done the critical work necessary to validate it. The contexts in which the principle of authority is rejected, and use is made of the critical method, are confined to relatively restricted areas of competence in which the experts are able to understand each other and thus be critical of each other’s work. Expert areas of competence are relatively narrow because acquiring thorough competence in a certain field of knowledge takes an extremely long time, precisely because it is based on the critical method. Outside these specific areas, most of the knowledge acquired is based on mixed forms of learning, which partly involve critical capacity but mostly the passive acquisition of notions seen as ‘consolidated’. This learning method is based on delegation. Essential for the functioning of all complex societies, delegation involves trust and acceptance of authority. It is necessary within the scientific community as well, given the extraordinary division of scientific work into the distinct disciplines that characterize modern science. As we learn and assimilate new ideas, we reproduce a permanent tension between implicit acceptance of the principle of authority (and of the division of mental work) and the Galilean principle based on critical capacity and experimentation. Thus, recourse to authority is not completely abandoned but becomes a hidden defect, one implicit in the scientific protocol that we use because, although we may exercise our critical capacity, our knowledge and the cognitive domain in which we operate are incomplete: we are thus forced to trust others and to import their errors into our system of knowledge.

A ‘pluralistic’ approach: theories on comparison How can different competing theories be compared in an authentically pluralistic view of the science? Popper’s approach is based upon the idea

Theories and ideologies as cognitive traps

33

of a logical asymmetry between verification and falsification. Thornton (1997) comments: For Popper . . . it is logically impossible to conclusively verify a universal proposition by reference to experience (as Hume saw clearly), but a single counter-instance conclusively falsifies the corresponding universal law. In a word, an exception, far from ‘proving’ a rule, conclusively refutes it. In line with remarks by the physicist Pierre Duhem (1906) – see Motterlini (1999) – the question can be set in a subtly different manner: in general an exception, instead of definitively refuting a general proposition, simply reduces its domain of applicability and points up that the categorization upon which the proposition is based must be redefined. It is slightly misleading to define scientific propositions or rules as ‘universal’ ones. What does ‘universal’ mean with regard to a proposition defined by a theory? It means that the proposition holds true when applied to all contingencies defined by the theory. The domain of validity of a proposition is limited by the theory itself. It is therefore convenient to introduce the notion of the ‘domain’ of applicability of a proposition or a rule. We can discriminate among competing theories when they are applicable to the same domain of data and empirical facts, and when comparison can be made among the different predictions or explanations that they yield on the basis of the facts pertaining to that domain. The argument can be outlined by starting from an analogous question: comparison between different solutions of a problem. Note that the requirements for comparison between theories also hold for human problem-solving activity: science is such an activity, so we can use the structure of problem-solving to understand how to compare different theories. Categories and abstractions in the description of problems Abstraction and classification are crucial elements in the process of building mental representations of problems; we will discuss them in this section, avoiding formalizations as much as possible. When a solution to a problem is defined in a given context, it is usually defined on a ‘domain’ that comprises more than one simple element. For example, the definition of ‘winning configuration’ in chess implies a large (and unknown) number of different configurations, all of which have the same property: the king must be under definitive attack, i.e. unable to avoid attack in one move. Finding a solution to a problem – for example, finding a winning strategy in chess – means discovering a procedure to achieve one element in a set of final winning positions whatever strategy the opponent chooses. A

34 Massimo Egidi similar and simpler definition holds for puzzles, the Rubik cube for example: here, finding a solution requires finding a procedure with which to achieve one element in the set of final winning configurations starting from the initial configuration. In the case of the Rubik cube, the final winning configuration may be one specific configuration: for example, the one in which every face (side) consists of tiles of the same colour. Or it may be defined as a class of configurations: for example, configurations in which every face is composed half by tiles of one colour and half by ones of another. The level of abstraction at which the problem is defined has an important relation with the level of abstraction at which a procedure to solve the problem may be defined. It is important, in fact, to understand under what conditions it is possible to define procedures that apply to the full domain of the problem; indeed, one of the main aspects of the ‘art’ of a programmer is the ability to construct a procedure whose degree of abstraction fits perfectly with the domains of the problem. Some definitions will help clarify the matter. A puzzle is a game in which an individual must achieve a given goal, usually by making changes to an initial configuration according to a system of rules and constraints. The rules state what actions may be made for each state of the game, and what their effects will be. A puzzle is therefore defined by the configurations, and by the rules that operate on the configurations. An important type of problem in the world of puzzles consists in achieving a given configuration (or a configuration that belongs to a class characterized by some property). A problem can therefore be defined by the rules of the puzzle and a pair of sets: the set of initial I and the set of final F configurations. A solution is a procedure (or a program, i.e. a set of condition–action rules coherent with the rules of the puzzle) that enables some element of F to be achieved starting from I. The question of representation immediately arises in these contexts. Assuming that we discover a strategy S(x, y) that solves the problem, S is a program that enables the player to achieve the configuration y 僆 F starting from x 僆 I. It may happen that S holds for only one pair x, y 僆 FxI, for many pairs, or for all pairs x, y 僆 FxI. The art of a good programmer consists in constructing S in such a way that the ‘domain of applicability’ of S is perfectly coincident with the two sets I, F that define the problem. Without loss of generality we can assume that every state of the game, excluding the goal configuration(s), may be a starting state. A case in point is the Rubik cube, a game in which the player must be able to achieve a final configuration whatever the starting configuration may be. There are millions of configurations in this game, and a program consisting of one instruction for each configuration would be accordingly enormous. A desirable program is one composed by a relatively small set of abstract instructions each of which is applicable to a group of configuration. The creation of a program is then limited by two important desirable properties: on the

Theories and ideologies as cognitive traps

35

one hand, the simplicity of representation: the program should be composed by as low as possible number of instructions. On the other, the efficiency: the number of steps to execute a program should be as small as possible. We will show in the following example that, at least in puzzles, there is a trade-off between simplicity and efficiency. This trade-off defines the constraints on which the discovery of new solutions proceeds. Assume that, given a problem I, F we discover a strategy S(x, y) that solves the problem for all x 僆 F, y 僆 F. With each pair x, y 僆 IxF we can associate a payoff, a measure of the efficiency of the problem-solving strategy. A very elementary measure is a (monotonically decreasing) function of the number of steps executed by applying the procedure: the higher the number of steps, the less efficient the procedure. Figure 1.1 depicts two strategies S and S* which solve the same problem with different degrees of efficiency: in some sub-areas of the set, S is more efficient than (preferable to) S*, (S 傻 S*), while in the complementary areas S* is vice versa preferable to S. In these conditions, which frequently occur in games and puzzles, there are two different strategies applying to different sub-domains. Players must pay the price of greater mental effort to learn one more strategy if they want to achieve optimality in execution of the procedure. Of course this situation can be generalized, in the sense that it may

6 4 1

2 0 0 1 0

1 1

Figure 1.1 The comparative efficiency of two strategies that solve the same problem

36 Massimo Egidi happen that many different strategies, S. S, . . . Sn can be defined, each of which is optimal only in one limited part of the domain of applicability. Returning to the Rubik cube, it is evident that an optimal solution exists, i.e. a path of minimal length connecting every initial configuration with the goal configuration. This optimal program can be described at ‘ground level’, i.e. by detailing an instruction for every configuration, and of course this description would imply an enormous number of instructions. Despite this obvious disadvantage, however, if a program is described with this ‘ground’ representation, its efficiency can be improved very simply. In fact, it can be shown (see Egidi 2002: appendix 3) that – despite a positive level of interaction (epistasis) – the optimal program can be found by simple ‘mutations’; i.e. by modifying every instruction sequentially until the optimal set of instructions is discovered. Of course this is not the way individuals proceed when constructing a program to play Rubik, or to solve similar puzzles. They try to compact the representation, i.e. to find rules that are applicable to classes of configurations. At the ground level an instruction consists of a configuration of the game and the action to be taken; at a more abstract level, we can identify classes of configurations with the same role in the game, i.e. to which we want to apply the same action. Assume that we want to construct a program compactly, i.e. by identifying classes of configurations to which appropriate actions apply. These classes can be called ‘building blocks’ constructed by abstraction or codification from the game properties. In the case of the Rubik cube, for example, in order to achieve the final position in which every face has tiles of the same colour, a player may try first to put the top corners in their right places. Given the disposition of the colours of one of the top corners, the player tries to put the second one in a position coherent with the first corner (see Figure 1.2). It is clear that the directions to move the

Figure 1.2 Rubik cube: a sub-problem

Theories and ideologies as cognitive traps

37

second top corner to its right position disregard all the positions of the other tiles. This means that players consider an enormous number of configurations – all those with the second top corner in the same position – to be equivalent. They will define the rules to apply only looking at the position of the second top corner, and will therefore consider all the configurations with the first and the second top corners in the same position as a single building block. The players will consequently perform the same action for every configuration of the same building block. A strategy S (or program) can therefore be described compactly as a list of building blocks, to each of which is attached an action. Thus a complete program consists of a list of relatively few instructions defined by the building blocks. Of course, the definition of the building blocks relates to the division of the original problem into sub-problems. A given set of building blocks describes the problem in its parts with some degree of abstraction: it is therefore the basic component of a representation. As we shall see in detail on pp. 44–48, constructing a basic system of building blocks enormously simplifies the representation of the problem, and enormously reduces the number of instructions. In so doing, it may introduce hidden errors, i.e. inefficiencies in the program that solves the game due to the way in which the problem has been decomposed into building blocks. To see briefly how a wrong decomposition introduces errors, assume that we know the optimal program described at the ground level: we have the list of the best actions to be performed for every configuration of the system. Assume that the best action for configuration xi is action ai, and the best action for configuration xj is action aj; and assume that ai  aj. For a given decomposition of the problem, the two configurations xi xj may belong to the same building block. In this case, the same action must be applied to both of them, and it therefore will be impossible to achieve the optimal solution. When a player modifies one instruction in a ‘compacted’ description of the program, he changes the action to be performed in relation to a given building block. The change is therefore applied to an entire set of configurations. Therefore, by introducing building blocks, we restrict the set of possible elementary modifications (mutations) that can be applied to a program. Hence, as we have shown, some decompositions of the problem necessarily lead to descriptions that do not incorporate the optimal solution. This implies that for any given problem there are many ‘wrong’ decompositions. These decompositions are in some sense the result of an excessive abstraction, or extrapolation. This explains how it is possible for many suboptimal strategies S, S, . . . to coexist: these strategies are simplified descriptions of the problem based on ‘wrong’ decompositions of the problem. As we shall see in more detail in the next sections, these strategies are locally stable and suboptimal for

38 Massimo Egidi some configurations of the domain. Therefore players that learn and adopt one of them may remain trapped in this representation. Cognitive traps at individual and team level I have shown that suboptimal strategies S, S, . . . in puzzles originate from ‘wrong’ decompositions of the problem, and that only changes in representations enable players to achieve the optimal solution in the full domain of applicability. Moreover, the players may not perceive the errors (suboptimalities) introduced by the decomposition that they discover. The domain from which they induce a decomposition may in fact be restricted to configurations for which the decomposition is optimal, as we have seen in Figure 1.1. These properties of problem-solving can be experimentally explored. The experiments now briefly described illustrate biases in human behaviours on the basis of the theoretical approach previously outlined. In the experiments described, groups of players were exposed to different sets of configurations for a training period. For each set of configurations Di it was possible to discover a simplified strategy which was optimal in that limited domain. Each group of players learnt the simplified strategy in the particular domain to which they were exposed, and remained locked in its specific strategy, using it beyond the domain of optimality. The first example of this kind of experiment was proposed by Luchins (1942) and Luchins and Luchins (1950): individuals exposed to simple mathematical problems admitting different solutions, S and S* tended to use the strategy that they learnt first (in a context in which it was efficient) even in sub-areas in which a better strategy could be found. These experiments suggest that the automaticity with which players repeat the same sequences of actions, solving a problem with a procedure that they have learnt in a particular domain even in conditions in which that procedure is clearly suboptimal, can be explained in terms of automaticity in their mental processes. Luchins and Luchins have argued in fact that routinized behaviours are based on routinized thinking – the socalled ‘Einstellung effect’ – or the automatic use of ‘chunks’ which enables individuals to save on mental effort (Weisberg 1980) but which at the same time induces them to cling to solutions for problems even when they prove to be suboptimal. These findings have close analogies with the properties of problemsolving in team contexts explored by Cohen and Bacdayan (1994) and Egidi and Narduzzo (1997) on the basis of experiment using the game ‘Target the Two’. These experiments suggest that the ‘Einstellung effect’ holds even with regard to team decision-making: when solving a repetitive task, for example by repetitively playing the same game, groups of players adopt routinized behaviours and persist in their use with remarkable stability even when they are clearly suboptimal. The routinization of behaviour may be considered the outcome of a

Theories and ideologies as cognitive traps

39

process of mental effort-saving that originates in the process of discovering and representing a strategy. During this process, individuals make systematic use of default classifications to reduce the short-term memory load and the complexity of symbolic manipulation. The result is the construction of an imperfect mental representation of the problem that nevertheless has the advantage of being simple and yielding ‘satisficing’ decisions. This view is illustrated by the previous example in which we supposed that if many strategies S, S, . . . , Sn apply with different degrees of efficiency to subdomains of a given problem, a trade-off will arise between the simplicity of the problem’s representation and its efficiency. The existence of this trade-off has been experimentally confirmed by some of the results obtained using the game ‘Target the Two’, which I now briefly describe. ‘Target the Two’ is a card game in which the two players must cooperate in order to achieve the final result. Each pair receives a reward proportional to the efficiency of their play: that is, the fewer the moves made by a pair to achieve the result, the higher its reward. Tournaments are organized in which pairs of players compete against each other. In each round of the tournament the cards are distributed randomly, and the players must learn how to co-ordinate themselves in the most efficient manner, but without communicating verbally. There are two suboptimal strategies, which I shall call A and B, each of which is optimal with respect to a restricted domain of initial configurations. For games which begin with initial configurations belonging to a certain set , strategy A dominates strategy B; while for games that begin with initial configurations belonging to a certain set , strategy B dominates strategy A. The two domains  and  have a part in common: that is, there are initial configurations with respect to which the two strategies are equally efficient. Figure 1.3 shows the number of moves required by each strategy for

9 8 7 6 5 4 3 2 1 0 1

5

9

13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 113 117

Figure 1.3 ‘Target the Two’: landscape of the efficiency of strategies

40 Massimo Egidi every configuration of the game. Arranged along the horizontal axis are the different game configurations (there are 124 structurally different configurations), and along the vertical one the number of moves required to achieve the result using strategy A and strategy B. Immediately apparent are the three domains in which A dominates B (from configuration 1 to 40), in which B dominates A (from 80 to 124), and in which A and B are equivalent (from configuration 40 to 80). An experiment conducted by the present writer with A. Narduzzo (1997) showed that pairs of players may become trapped in a suboptimal strategy, without learning the optimal one, even in conditions where the optimal strategy is easy to discover. Two groups of players – Ga and Gb – were formed, and each of them participated in a tournament consisting of two parts. In the first part, group Ga was exposed to hands in which strategy A was dominant and also easy to learn. Likewise Group Gb was given hands that could be easily played with strategy B, which was dominant. In the second part of the tournament both groups were given the same configurations, chosen at random. The results showed a persistent difference in behaviour: the players in group Ga used strategy A much more frequently than did players in group Gb, even when the strategy was dominated by the other one, and vice versa. Figure 1.4 shows this persistence of behaviour. The horizontal axis shows the runs (after the training period) in which both groups were given the same game configurations. The vertical axis shows the percentage of pairs which used strategy A in group Ga and group Gb, respectively. (On the horizontal axis are the different runs.) In each of the two groups there was a rather high percentage of players who consistently used a single strategy, the one learnt during the game training phase, so that the figures continued to differ throughout the rest of the tournament.

100 90

Ga

80 70 60 50 40

Gb

30 20 10 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Figure 1.4 ‘Target the Two’: routinized behaviours

Theories and ideologies as cognitive traps

41

With regard to the behaviour of individual pairs, a relevant percentage of pairs in group Ga only used strategy A in all game conditions, and the same behaviour was displayed by group Gb (Egidi and Narduzzo 1997: 699). Consequently, we may call the behaviours of these players ‘routinized’ because they invariably used the same strategy, without learning the alternative strategy even though it was more appropriate. This experiment raises the question of whether the systematic bias was caused by an interaction effect – for example, elements due to the difficulty of cooperating – or whether it was the result of individual routinized thinking. In order to clarify the matter, I conducted an experiment identical to the previous one in which the sample was similarly structured and the same sequence of games was played, but with the difference that only single players, instead of pairs, played the game. The coordination problem, and implicit communication between the players (they were not allowed to communicate explicitly), were thus eliminated. In this experiment, too, there were two groups of players – Ga and Gb – who were initially given fifteen card sequences in which strategy A and strategy B were respectively dominant, and then both groups were exposed to the same sequence of randomly selected hands. The results were quite clear (Figure 1.5). In this game, too, players in group Ga continued to prefer strategy A even when it was dominated by strategy B, and vice versa the players in group Gb. The ‘lock-in’ effect therefore affected single individuals also. The findings of Luchins and Luchins are thus confirmed in this particular context as well. However, routinization was less marked than in the case of cooperating pairs, and the number of routinized individuals was much lower and differently distributed than in the previous experiment with pairs. At least in 90 80 Ga

70 60 50 40

Gb

30 20 10 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Figure 1.5 ‘Target the Two’: single-player routinization

42 Massimo Egidi this context, therefore, the coordination process reinforces ‘deviations’ from the Olympic rationality that characterizes individual behaviours. In a context of tacit knowledge in particular, the difficulty of coordination is largely responsible for the persistence of cognitive and behavioural biases in a team. Although generalizing this result would require a much larger body of empirical data, we nevertheless have interesting evidence of the considerable extent to which difficulties in coordination reinforce the barriers that individuals encounter when trying to get out of a cognitive trap. It is interesting to note that the difficulties of team problem-solving seem analogous to those observed in the above story about Fermat’s Last Theorem. Building blocks, local stability, suboptimality of representations The extrapolation evidenced by the experiments on ‘Target the Two’ is a general feature of the problem–representation process which gives rise to systematic imperfections and biases. The imperfections originate from the process of constructing the categories which represent the building blocks of the problem’s solution. To find a solution, in fact, individuals normally try to decompose a problem into parts to be solved separately. Every decomposition is based on a categorization of the problem, so that different decompositions may be characterized by different abstraction levels of the categories (some abstraction levels may ignore some of the interdependencies among the sub-problems). In analogy with the example of many strategies S, S, . . . , Sn, raising the level of abstraction with which a sub-problem is represented means extending the domain of validity of the categories and their relations even beyond the field in which it has been experienced by the individual, with the consequence that domains in which the solution is inefficient may be unintentionally included. The onset of errors in the mental representation of a problem may therefore be the ‘natural’ effect of attempts to simplify the categorization and identification of the building blocks of a problem. Moreover, experiments on individual and team behaviour provide a basis on which the persistency of biases can be explained. Persistency of biases can be interpreted as the existence and stability of suboptimal solutions to problems due to the difficulty of redefining the sub-problems that constitute the elementary building blocks of the problem’s representation. These elementary building blocks are based on systems of categories that, as will be shown in the next section, focus the players’ attention and drive the construction of their mental models. The division of knowledge derives from the way in which individuals categorize problems during the training or learning phase. It should be stressed that a given problem may be decomposed in a large variety of different ways which also give rise to different levels of abstraction in the categorization of sub-problems. Every decomposition pattern results

Theories and ideologies as cognitive traps D C

43

B A

Figure 1.6 Minirubik

from a different manner of codifying information at different levels of abstraction. In Egidi (2002) the optimal decomposition of a problem – the discovery of a strategy for playing the game Minirubik – is compared with other decompositions which are simpler and easier to learn but suboptimal. We describe again the example of Minirubik, to show the properties of local stability and suboptimality of the solutions. Minirubik is a sort of Rubik square. The player has a square consisting of four differently coloured tiles denoted by the letters A, B, C, D (Figure 1.6). The tiles can be exchanged horizontally or vertically, as shown in Figure 1.7, and players must exchange them until they have achieved a final configuration. Players are rewarded according to the number of moves that they make to achieve the goal: the higher the number of the moves, the lower the payoff. With this simplified representation, a strategy can be represented as a list of condition–action instructions of the type exemplified in Figure 1.8. D C

B A

UP

B C

D A

D C

B A

DOWN

D A

B C

D C

B A

RIGHT

D C

A B

D C

B A

LEFT

C D

B A

Figure 1.7 Minirubik: the exchanging of tiles Condition D A

B C

Action Up

Figure 1.8 Minirubik: example of a condition–action instruction

44 Massimo Egidi To simplify the description, a configuration can be written as a sequence of four letters (or colours), rather than as a square of four letters (or colours), by applying the following rule: start from the upper-left corner of the square and list the elements of the square, moving clockwise. With this rule the square in Figure 1.6, for example, is transformed into the list DBCA and the instructions for Figure 1.8 can be written as DBCA → Up. Following Holland (1975), the symbol # is used synonymously with ‘don’t care’, which means that a configuration like A##C is an abstract configuration representing the set of configurations in which A is in the first position (upper-left corner) and C in the last one (lower-left corner). The symbol # allows us to represent sets of configurations, i.e. to represent configurations at some level of abstraction. Suppose that the initial and the final configurations are given, respectively, by the two strings BDAC and ABCD, as in Figure 1.9. A program with which to move from this particular starting configuration to the final goal configuration consists of the following instructions: 1 2 3

Move A from the initial position clockwise to the final position. If B it is not yet in the final position, move it to the upper-right corner leaving A in its position. If C and D are not yet in the required final positions, exchange them.

It is clear that we have constructed this set of instructions by following a heuristic, or in other words by following criteria for the decomposition of the original problem into sub-problems involving related levels of abstraction and categorization of the problem. These criteria are based on the idea of focusing on the position of one tile at a time (first we move A, then B and finally C). This implies the following two points: 1

2

Categorization: we have implicitly adopted a ‘categorization’ of the problem: the elements of our reasoning are the categories A###, #A##, ##A#, ###A, A#B#, . . . and we mentally manipulate them. Interdependence: the instructions proposed are based on the conjecture that it will be possible to solve the problem by considering the movement of each tile, disregarding the effects that the change of position of one tile has on the others and therefore detecting rules that apply to the simple categories defined above. Starting configuration B C

D A

Goal configuration A D

B C

Figure 1.9 Minirubik: example of a starting configuration and the goal

Theories and ideologies as cognitive traps

45

By decomposing each of the three above instructions into elementary actions, we can rewrite it as an array of elementary instructions (in the form of conditions–actions) as is illustrated in Figure 1.10. We may wonder whether it is possible to arrange these instructions into a more general format, maintaining the solution criteria that we have adopted (i.e. focusing on the positions of one tile at time) and successfully applying them to all starting configurations of the game. The answer is positive: by means of simple reasoning (see Egidi 2002: 140) it is possible to extend the previous instructions to a broader domain; namely, the set of all initial configurations. We obtain the array of Figure 1.11, which applies to every initial configuration and enables players to achieve the goal configuration. The first specific system of instructions (Figure 1.10) drawn up to solve the initial, specific problem is a sub-set of the new system of instructions, which applies to every initial condition. The system of abstract rules in Figure 1.11 is therefore complete, because the abstractions allow us to cluster and classify all specific rules of the system into a few sub-sets: it is a compact representation of the rules of action. I have called this representation ‘First A’ because the ‘heuristic’ of this system is to move first A, then B and finally C into their goal positions. We thus have simplicity in the representation of the procedure with which to solve the entire class of problems. Unfortunately, however, we do not have full efficiency; in fact, for some initial conditions, the procedure is suboptimal (see Egidi 2002: appendix 2), as the example in Figure 1.12 shows. Conditions

Actions

# #

# A

Right

# #

A #

Up

A #

# B

Right

A D

B C

Down

Figure 1.10 Minirubik: an array of elementary instructions

46 Massimo Egidi Conditions

Actions

# #

# A

Right

# #

A #

Up

A #

# B

Right

A D

B C

Down

A B

# #

Down

# A

# #

Left

Figure 1.11 Minirubik: ‘First A’ representation D B

C A

A D

B C

Figure 1.12 Minirubik: initial position which is suboptimally played with the ‘First A’ strategy

The optimal sequence here, in fact, is DCBA → Right → DACB → Left → BACD → Up → ABCD, whereas the rules prescribed by our procedure generate the sequence DCBA Right → DACB → Left → BACD → Up → ADCB → Down → ADBC → Right → ABDC → Down → ABCD, which is obviously made up of more steps. The simplified procedure S(x, y) that we have defined is optimal only in a sub-domain of the initial conditions. Therefore, when adopting this procedure, players have the advantage of a representation which is very simple and abstract and complete, but at the price of inefficiencies because the number of moves required to achieve the goal is higher than the optimal number. Moreover, the ‘First A’ representation is a compact list of instructions that is suboptimal and locally stable: 1

Local stability. The reader may like to perform the exercise, modifying the actions corresponding to the list of conditions in Figure 1.11, and

Theories and ideologies as cognitive traps

2

47

verifying that the efficiency of the programs worsens. This means that, given the representation of the problem defined by the building blocks ###A, ##A#, #A##, A##B, A#B#, ABDC, the instructions in Figure 1.11 are a local optimum. When the instructions in the figure are modified, the program worsens; i.e. the number of moves required to achieve the goal increases. Suboptimality. As shown in correspondence to some configurations (for example DCBA), the program in Figure 1.11 is suboptimal; i.e. there exists a path to the goal that is shorter than the path defined by Figure 1.11. We have seen that (local stability) every list of instructions obtained by modifying the figure worsens the situation. Therefore, however we modify the instructions in the list – changing the actions to be performed in relation to the conditions described in abstract in Figure 1.11 – we will never achieve the (ground) optimal program.

This is therefore a clear example of the trade-off between simplicity of representation and the efficiency of the program: the programs generated by Figure 1.11 show inefficiencies in correspondence to some configurations, and a player adopting this representation cannot avoid making these errors (inefficiencies), even if he tries to modify some of the actions. The only way to improve the efficiency of the program is to change the representation. Preliminary experiments confirm that a large number of players exposed during the initial runs of the game to configurations optimally solved by the strategy illustrated in Figure 1.12 will adopt this strategy, thereby committing systematic errors perfectly in line with the figure’s instructions. Therefore biases emerge from induction: players make a default classification of the configurations to which they want to apply a rule (for example, ##A# → Right). Put otherwise, they conjecture a rule that is supposed to apply to the entire domain defined by a category (##A#), without perceiving that the rule has, on the contrary, a more limited domain of application. Perception of this limitation can only come with experience, because the rule has been conjectured by a process of induction whereby a player extends a rule discovered in a particular context to the entire domain. It is evident that any division of a problem into sub-problems is based on a structure conjectured in accordance with certain decomposition principles. The classifications and abstractions elicited by these principles may embody hidden errors which cannot be perceived without direct inspection. Finding these hidden errors, in fact, would require detailed examination of all the existing rules: an extremely time-consuming task which would nullify all the advantages of a concise representation. Consequently,

48 Massimo Egidi individuals, precisely because of the inductive nature of their search, do not actively seek to find exceptions to the rules that they have established to solve the problem. We can see from the Minirubik example (see Egidi 2002: appendix 2) that, in order to correct the errors hidden in abstract rules, it is necessary to be guided by emerging exceptions. Any attempt to discover those errors actively would require a fully detailed description of the game configurations that nullified the ‘parsimony’ of the induction-based inference and pre-empted the effort to express the strategy simply. The locally optimal (globally inefficient) solutions in which individuals may remain trapped while analysing a problem are therefore created by the limits to their capacity to falsify the rules that they have conjectured in all relevant domains and to discover hidden errors. Precisely because they cannot discover exceptions, it is highly unlikely that they will actively redefine the basic categories on which they have constructed the solution, so they may remain trapped in the given representation.

Conclusions In the latter sections I have described experiments which shed light on the features of the human search for solutions in complex artificial environments such as complex games and puzzles. These characteristics are strikingly similar to the features of the human activity of theory creation and modification. I shall therefore conduct further comparison between the two intellectual domains of problem-solving and epistemology. In problem-solving, two basic elements drive the search for solutions: (a) induction – based on conjectures – by which individuals extend specific cases to wider domains, and (b) specification, the opposite process elicited by emerging errors. Players in complex games do not represent all possible configurations in their minds. Rather, they usually proceed by generalization on the basis of examples – which means that they seek to induce general rules from specific experiences, as happens for example with children naturally learning a new language. Induction is closely related to default classification (Holland et al. 1988). When a player extends the domain of the validity of a rule discovered in a specific domain as widely as possible, he classifies by extension all conditions that match the specific example by assuming that they are the right conditions for application of the given rule. This procedure enables players to create a complete – albeit highly suboptimal – initial strategy, in which the rules of action for a large number of configurations are defined ‘by induction’ from examples. The extrapolation can also be viewed as a default classification. When a new configuration occurs, eliciting a new rule as an exception, in general it gives rise to a new rule that is more specific than the default

Theories and ideologies as cognitive traps

49

one. A problem arises in extending the new specific rule, i.e. in extrapolating to larger domains, because this extrapolation will conflict with the previous system of rules. The problem is that adding one exception to a system of abstract rules does not yield a new, compact representation of the strategy, because if individuals transform a specific exception into a new rule with some degree of generality, the new rule may conflict with many other pre-existing rules of the strategy. It follows that individuals may prefer to maintain the old system, perhaps adding a few exceptions, rather than devising new abstract rules, because the mental effort required to redefine a sub-problem system is greater than that required to memorize an exception. If the number of ‘exceptions’ grows too large, and if they systematically occur during the game, the players cannot simply continue to memorize new exceptions; they must instead restructure the space of the rules, re-codifying information. In other words, they must change the representation – a change which may be highly discontinuous because it generally entails destructuring the division of problems and redesigning the problem with new building blocks. (Changing the representation may be particularly difficult if it requires redefinition of too many basic sub-problems.) This is, of course, an extremely onerous mental task, so that it is likely that the new example will be treated as an anomaly, without prompting re-categorization of the problem. Suboptimal solutions are therefore stable under the emergence of exceptions because changing the representation while maintaining a limited number of general rules requires radical modifications to the building blocks – the elementary sub-problems – of which a strategy is composed. I have suggested that the properties of problem-solving illustrated in this chapter can be extended by analogy to the world of competing theories. When we compare the explanatory capacities of two competing theories (for example, expected utility and regret theories), we do so in the area where the domains of the two theories overlap: that is, we conduct the comparison by referring to the ‘facts’ to which both theories apply. Comparison between two theories in a domain in which both are applicable should normally be feasible. The only serious problem that may arise concerns the difficulty of determining which theory is best solely on the basis of statistics – as happens, for example, in certain comparisons between expected utility and regret theory (Hey 1991: ch. 5), when the statistical tests are of insufficient power to yield clear-cut results. More interestingly, it may happen that the core statements of a theory have domains to which they have never been applied: the emergence of anomalies, i.e. unexpected examples in the theory’s domain of applicability that contradict it, giving rise to a dynamic of change analogous to the dynamic of problem-solving. As illustrated earlier, this process of revision requires a

50 Massimo Egidi new decomposition based on a new categorization of the elementary building blocks of the problem (the theory). I have suggested that the failure (to date) of attempts to find a satisfactory new theory of decisionmaking, on the basis of slight modifications to the classical axioms of expected utility theory, has been due to the extreme simplification of the basic axioms, which do not take serious account of the main features of human thinking. Turning to epistemology, insofar as the modification of a theory is similar to problem-solving, the features displayed by the search for new solutions in problem-solving entail critical revision of Popper’s idea of falsifiability in the light of Lakatos’s position: any anomaly, or element of falsification, reduces a rule’s domain of applicability, but it does not necessarily allow redefinition of all the sets of interrelated abstractions (a hierarchical system of categories) that made up the previous solution. This may come about as a result of the cumulative effect of the random emergence of many anomalies. In this case, anomalies drive the process of recategorization and induce individuals to adopt new solutions which are once again locally stable, though imperfect. The experiments described in this chapter may aid understanding of why theories and ideologies persist over long periods of time with remarkable stability when they have been largely falsified, and of what processes induce individuals to discard previous theoretical approaches. Redefinition of the categories constituting the building blocks of a solution (a theoretical approach) requires the complex process – what Popper calls ‘critical thinking’ – that I have tried to describe. Prejudices and erroneous simplifications are therefore natural and necessary for the creation of new solutions. Rationality emerges essentially as the capacity to get rid of our prejudices.6

Notes 1 I am grateful to Matteo Motterlini for his brilliant and helpful remarks on the draft version of this chapter. The usual disclaimers hold. 2 See also Newell et al. (1962). 3 Even though the existence of a best strategy can be demonstrated mathematically. 4 Of course, it is necessary to abandon the ingenuous idea that decision-making is based only on the capacity of individuals to order their preferences and to select the preferred option, taking any possible connections into account. 5 This is the condition in which Simon’s idea of ‘bounded rationality’ achieves full significance, because the limits of rationality emerge substantially, so that the human process of problem-solving is described not as an approximation to classical Olympian rationality but in terms of the properties of mental processes based on the division of knowledge and categorization. 6 In 1994, Popper defined rationality as the attitude of eliminating errors in a critical way.

Theories and ideologies as cognitive traps

51

References Allais, M. and Hagen, O. (eds) (1979) Expected Utility Hypothesis and the Allais Paradox: Contemporary Discussions of Decisions Under Uncertainty with Allais Rejoinder, Dordrecht: D. Reidel. Cohen, M.D. and Bacdayan, P. (1994) ‘Organizational Routines are Stored as Procedural Memory: Evidence from a Laboratory Study’, Organization Science, 5(4): 554–568. Cohen, M.D., Burkhart, R., Dosi, G., Egidi, M., Marengo, L., Warglien, M. and Winter, S. (1996) ‘Routines and Other Recurring Action Patterns of Organizations: Contemporary Research Issues’, Industrial and Corporate Change, 5(3): 653–698. Cyert, R.M., Simon, H.A. and Trow, D.B. (1956) ‘Observation of a Business Decision’, Journal of Business, no. 29: 237–248. Denzau, A.T. and North, D.C. (1994) ‘Shared Mental Models: Ideologies and Institutions’, Kyklos, 47(1): 3–31. Dosi, G. and Egidi, M. (1991) ‘Substantive and Procedural Uncertainty: An Exploration of Economic Behaviors in Changing Environments’, Journal of Evolutionary Economics, 1(2) (April): 145–168. Duhem, P. (1906) La Théorie physique. Son objet et sa structure, Paris: Chevalier and Rivière. Egidi, M. (2002) ‘Biases in Organizational Behavior’, in M. Augier and J.J. March (eds), The Economics of Choice, Change and Organization: Essays in Memory of Richard M. Cyert, Aldershot: Edward Elgar. Egidi, M. and Narduzzo, A. (1997) ‘The Emergence of Path Dependent Behaviors in Cooperative Contexts’, International Journal of Industrial Organization, 15(6) (October): 677–709. Ericsson, K.A. and Simon, H.A. (1984) Protocol Analysis: Verbal Reports as Data, Cambridge, Mass.: The MIT Press. —— (1985) ‘Protocol analysis’, in T.A. Van Dijk (ed.), Handbook of Discourse Analysis: Vol. 2, Dimensions of Discourse (ch. 14). New York: Academic Press. Faltings, G. (1995) ‘The Proof of Fermat’s Last Theorem by R. Taylor and A. Wiles’, Notices of the AMS, 42(7): 743–746. Friedman, M. (1953) ‘The Methodology of Positive Economics’, in Essays in Positive Economics, Chicago: University of Chicago Press. Gödel, K. (1931) ‘Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I’ (On Formally Undecidable Propositions in Principia Mathematica and Related Systems), Monatshefte für Mathematik und Physik, 38: 173–198. Hey, J. (1991) Experiments in Economics, London: Basil Blackwell. Holland, J.H. (1975) Adaptation in Natural and Artificial Systems. Ann Arbor: University of Michigan Press. Holland, J.H., Holyoak, K.J., Nisbett, R.E. and Thagard, P.R. (1988) Induction – Processes of Inference, Learning, and Discovery, Cambridge, Mass.: The MIT Press. Kahneman, D. and Tversky, A. (1979) ‘Prospect Theory: An Analysis of Decisions Under Risk’, Econometrica, 47: 313–327. —— (2000) Choices, Values and Frames, Cambridge: Cambridge University Press.

52 Massimo Egidi Lakatos, I. (1974) ‘Popper on Demarcation and Induction’, in P.A. Schlipp (ed.), The Philosophy of Karl Popper, La Salle, Ill.: Open Court. Langlois, R. (1998) ‘Rule-following, Expertise, and Rationality: A New Behavioral Economics?’, in K. Dennis (ed.), Rationality in Economics: Alternative Perspectives, Dordrecht: Kluwer Academic Publishers. Luchins, A.S. (1942) ‘Mechanization in Problem-Solving’, Psychological Monograph, 54: 1–95. Luchins, A.S. and Luchins, E.H. (1950) ‘New Experimental Attempts in Preventing Mechanization in Problem-Solving’, The Journal of General Psychology, 42: 279–291. March, J.G. and Simon, H.A. (1958) Organizations, New York: Wiley. Motterlini, M. (ed.) (1999) For and Against Method. Including Lakatos’s Lectures on Method and the Lakatos–Feyerabend Correspondence, Chicago: University of Chicago Press. Newell, A. and Simon, H.A. (1962) ‘Computer Simulation of Human Thinking and Problem Solving’, in M. Greenberger (ed.), Management and the Computer of the Future, New York: Wiley. Newell, A., Shaw, J.C. and Simon, H.A. (1958) ‘Chess-playing Programs and the Problem of Complexity’, IBM Journal of Research and Development, 2: 320–335. —— (1962) ‘The Processes of Creative Thinking’, in H.E. Gruber, G. Terrell and M. Wertheimer (eds), Contemporary Approaches to Creative Thinking, New York: Atherton Press. O’Connor, J.J. and Robertson, E.F. (1997) Fermat’s Last Theorem: http://wwwgroups.dcs.st-and.ac.uk/~history/HistTopics/Fermat’s_last_theorem.html#75. Popper, K. (1959) The Logic of Scientific Discovery (translation of Logik der Forschung), London: Hutchinson. —— (1960) Philosophical Lecture ‘On the Sources of Knowledge and of Ignorance’, Proceedings of the British Academy 1960, London: Oxford University Press. —— (1963) Conjectures and Refutations: The Growth of Scientific Knowledge, London: Routledge. —— (1994) ‘The Self, Rationality and Freedom’, in K. Popper, Knowledge and the Body–Mind Problem. In Defence of Interaction, London and New York: Routledge. Simon, H.A. (1963) ‘Problem Solving Machines’, International Science and Technology, 3: 48–62. —— (1979) ‘Rational Decision Making in Business Organization’, American Economic Review, 69: 493–513. —— (2002) ‘Science Seeks Parsimony, Not Simplicity: Searching for Pattern in Phenomena’, in A. Zellner, H.A. Keuzenkamp and M. McAleer (eds), Simplicity, Inference and Modeling: Keeping It Sophisticatedly Simple (ch. 3), Cambridge: Cambridge University Press. Simon, H.A. and Newell, A. (1972) Human Problem Solving, Englewood Cliffs, Calif.: Prentice-Hall. Thornton, S. (1997) ‘Karl Popper’, The Stanford Encyclopedia of Philosophy (Summer 2002 edition), URL: http://plato.stanford.edu/entries/popper/. Weisberg, R. (1980) Memory: Thought and Behavior, New York: Oxford University Press.

2

Cognitive science meets evolutionary theory What implications does evolutionary psychology have for economic theorising? Jack J. Vromen

Introduction Cognitive science meets evolutionary theory – that is how evolutionary psychology (henceforth: EP) often is characterised. And indeed, as we shall see, one of EP’s main objectives is to point out the evolutionary raison d’être of information-processing modules in the human mind. EP not only sets out to explain why we have the modules that we have, but also aims to facilitate the identification of hitherto unknown modules. In this chapter I try to explore what potential contributions EP can make to economics. As a starting-point I take two recent attempts made by economists to connect evolutionary theory in general, and EP in particular, with economics. First Oliver Williamson’s (1998) argument is addressed that EP supports the ‘transaction cost triple’ for describing human agents: opportunism, bounded rationality and farsighted contracting. The second attempt to relate evolutionary theory to economic theory is made in the so-called indirect evolutionary approach. This approach aims to explain what basic preferences evolutionary processes have endowed us with. The indirect evolutionary approach assumes that the insights of evolutionary theory can readily be accommodated within the analytical framework of standard economic theory. After a brief discussion of EP’s essential features an effort is made to find out which of the two attempts gets the main thrust of EP right.1 It is argued that after having had a first, superficial glance it seems that each of the two attempts is partly right and partly wrong. But if we dig deeper it turns out that at a more fundamental theoretical level they both got it wrong. It will be pointed out that EP’s basic presuppositions are incompatible with presuppositions underlying both Williamson’s transaction cost economics and the indirect evolutionary approach. The chapter concludes with a few speculations about alternative ways to link EP with economics.

54 Jack J. Vromen

Williamson on EP’s vindication of the transaction cost triple Williamson (1998) claims that the transaction cost triple for describing human agents – bounded rationality, farsighted contracting, and opportunism – finds support in the EP literature. Let us have a closer look at the argument that Williamson develops to back up this claim. What exactly is Williamson’s ‘transaction cost triple’? For reasons that will soon become clear it is instructive to distinguish between the meaning of the elements and their respective roles in the transaction cost economics (henceforth: TCE) framework. For Williamson ‘bounded rationality’ essentially is a negative concept. It basically says that in contrast to ‘perfect rationality’ agents do not have boundless or limitless capacities to acquire, store and process information, do not have boundless computational capacities, and so on. This is not to say that in problem-solving and decision-making agents cannot try to find and implement optimal solutions or try to make the best possible decisions. They can try to do all this. It is just that in their attempts to do so the capacities of agents are limited in various ways. Agents are intendedly rational, but only limitedly so, as Williamson has it. In Williamson (1998) it is argued in particular, first, that boundedly rational agents are prone to make judgement errors and, second, that boundedly rational agents employ decision-making heuristics. ‘Farsighted contracting’ refers to the capacity of agents, their bounded rationality notwithstanding, to foresee general consequences of different types of contract. Agents capable of farsighted contracting do not suffer from myopia. Because of this capacity to look ahead in time agents know when to expect harmful consequences from existing contract forms. They can devise or choose contract forms that avoid such harmful consequences. Williamson, finally, sees ‘opportunism’ as an extension of the standard assumption of self-interested behaviour in economic theory. To be more precise, he defines it as self-interest seeking with guile. We get a better view on how Williamson puts together these notions when we consider the role they play in the TCE framework. The main implication of bounded rationality in the TCE framework is that complete contracting is not a feasible option. Boundedly rational agents lack the capacity to foresee all possible future contingencies. Allowance should be made in contracts for flexible responses to unforeseen future events and developments. That is why we have incomplete contracts only. If only agents were not opportunistic, this would not pose a serious problem. But if agents (or at least a few of them) are opportunistic, as typically is assumed in TCE, then contract enforcement and contract compliance problems can be expected to arise. These problems cannot be solved just by contractants making promises to abide to the contract. As Williamson puts it, ‘mere promise, unsupported by credible commitments, is not selfenforcing’ (Williamson 1998: 3). Here, finally, the assumption of farsighted contracting comes in. Agents capable of farsighted contracting can foresee

Cognitive science meets evolutionary theory 55 that mere promises have unfortunate consequences in the form of conflicts. Relying on mere promises may stand in the way of fully exploiting all potential mutual gains. Agents will sooner or later realise that the missing credible commitments can be provided by governance structures (such as hierarchical ‘hands-on’ governance). To sum up then, Williamson argues that EP vindicates all three elements of his ‘transaction cost triple’ separately and subscribes to the way in which the three elements are linked with one another in his TCE. Not only is EP taken to endorse the view that opportunism and bounded rationality, understood in the way Williamson does, are two essential behavioural characteristics of human beings, EP is also taken to underwrite Williamson’s view that individuals can overcome contract compliance problems resulting from them by finding suitable governance structures.

The Indirect Evolutionary Approach (IEA) on the evolution of basic preferences Ben-Ner and Putterman (1998, 2000) argue that the promise EP holds out for economics is that it can identify our preference profiles in a nonarbitrary way. Zywicki (2000) voices the same expectation. EP can explain why we have the preferences that we have. Both Ben-Ner and Putterman (2000) and Zywicki (2000) note that EP suggests that humans are naturally endowed with altruistic inclinations (or, more generally, with otherregarding preferences). But they are quick to add that the altruism involved (of the Tit-for-Tat variety, for example) is of a reciprocal or conditional type. This means, they go on to argue, that the altruism EPists talk about really is some sort of enlightened egoism: individuals are only willing to give others a helping hand (or to co-operate with them) that reciprocate this favour. Only then is helping others out conducive to their own self-interest. Thus they take EP to be supporting the assumption often made in economic analyses that individuals are led by self-interest (if not selfishness). Gintis (2000) disagrees. Gintis grants that there are forms of reciprocal altruism that really are instances of self-interested co-operation. But EP indicates that there are also forms of ‘strong’ reciprocity to be observed;2 forms that cannot be reduced to self-interest. The image of human nature and of the human behavioural repertoire that EP comes up with considerably goes beyond Homo economicus, Gintis argues. Gintis introduces three new evolutionary models of human behaviour. The first one is that of Homo reciprocans: Homo reciprocans exhibits what may be called strong reciprocity, by which we mean a propensity to cooperate and share with others similarly disposed, even at personal cost, and a willingness to punish those

56 Jack J. Vromen who violate cooperative and other social norms, even when punishing is personally costly, and even when there are no plausible future rewards or benefits from so behaving.3 (Gintis 2000: 288) The other two models are that of Homo egualis4 – exhibiting an urge to reduce inequality, a weaker one when on top and a stronger one when on the bottom (Gintis 2000: 284) – and that of Homo parochius – exhibiting a tendency to divide the world into insiders and outsiders, and a tendency to favour insiders and to disfavour outsiders. Gintis’s view that evolution has endowed Homo sapiens with a motivational and behavioural repertoire that far exceeds that of the Homo economicus model seems to be shared by proponents of the so-called ‘Indirect Evolutionary Approach’ (IEA). Consider for example Güth and Yaari (1991), the paper in which the IEA was introduced.5 Güth and Yaari (1991) set out to show how a genetically determined preference for reciprocal behaviour could possibly have survived ruthless selection processes. In order to do this, Güth and Yaari consider a simple game in extensive form with two players. They confine their attention to an analysis of two different ‘gene types’ (as players): one player is assumed to be genetically disposed to reciprocate harm with harm, whereas the other player is assumed to be genetically disposed to yield to a harming opponent. Assuming in addition complete information and individual rationality, Güth and Yaari show that the reciprocating type is the only Evolutionarily Stable Strategy in the game. The interesting thing about this result is that, appearances notwithstanding, a preference to reciprocate is reproductively successful, not despite but because of its determined reciprocating nature. People genetically endowed with a preference to reciprocate can avoid being exploited by others – precisely because they are determined, regardless of the personal costs involved in doing so – to harm others if these others harm them (and the others know this).6 Thus like Gintis, Güth and Yaari also argue against the view that processes of biological evolution could possibly have led only to conditional or weak forms of reciprocity. It is entirely possible that a genetically encoded preference for unconditional or strong reciprocity, that guides behaviour also if that comes at a personal cost, has evolved. What sets the IEA apart from direct evolutionary approaches in economics is that evolution is believed to bear on human behaviour indirectly, via preferences, rather than directly.7 In the direct evolutionary approach of evolutionary game theory it is assumed that the behaviour of individuals is selected more or less directly in evolutionary processes. Environmental factors and variables, for example, or conscious thought are assumed not to intervene in processes in which what is selected (genes, for example) translate into overt behaviour. By contrast,

Cognitive science meets evolutionary theory 57 the guiding idea in the IEA is that while our present preferences bear witness of our evolutionary past as a species, our evolutionary past as a species does not fully determine how we behave nowadays. Our capacity to act rationally allows us to act flexibly and in accordance with our expectations about the future. Even though we may be called prisoners of our preferences, the IEA ‘does not neglect individuals’ ability to adjust their behavior in different environments when studying a selection process’ (Huck 1997: 777). The IEA’s basic logic, Huck writes, is that ‘preferences guide behavior, behavior determines fitness, and fitness drives the evolution of preferences’ (Huck 1997: 773). We have a sequence here of successive stages of interaction. In each stage preferences directly affect behaviour. But fitness differences ultimately determine what sorts of preferences tend to dominate the population and hence also indirectly what sort of behaviour survives in the long run.8 If preferences and fitness differences would coincide completely right from the start, then results like Güth and Yaari’s (1991) cannot be obtained. What gives the IEA its bite is that it investigates what happens if preferences differ from fitness differences. In a game like the trust game, analysed in Güth and Kliemt (1998), trust could never evolve if preferences were identical to fitness differences. It is only if we introduce players with different preference profiles, among them trustworthy players – players who do not take much pleasure in exploiting the trust put in them by others – and ‘sadist’ players – who take much pleasure in doing so, that trust can get off the ground. The IEA seems to be developed mainly to construct simple models capable of covering ‘the bewildering variety of evidence’ (Fehr and Schmidt 1999: 818) obtained in experimental game theory. In games such as the ultimatum and dictator games, for example, experiments consistently violate the prediction of perfect-equilibrium analysis (namely, that the proposer keeps virtually all the money to be divided). But in simple auction-market game experiments, play of the predicted subgame-perfect equilibrium is robustly observed after a few repetitions. Proponents of the indirect evolutionary approach typically reject positing different motives for the different sorts of games. Instead they argue that the same motivations hold across different games. A concern for relative payoff is included in the motivation function (Bolton and Ockenfels 2000), for example, or self-centred inequity aversion is assumed (Fehr and Schmidt 1999). What makes people’s behaviour different in different sorts of games, they argue, is not that people are led by different (mixes of) motivations, but by the different opportunities the different sorts of games offer for strategic manoeuvring.

58 Jack J. Vromen

The bones of contention Williamson and the IEA seem to draw quite different, if not opposite conclusions from evolutionary theory. Whereas Williamson argues that EP supports his view that individual human beings are opportunistic, proponents of the IEA maintain that dispositions to engage in unconditional reciprocal and trustworthy behaviour are part of the encompassing preference profiles of individuals. And while Williamson argues that his assumption that individuals are boundedly rational (yet capable of farsighted contracting) is sustained, the IEA sticks to the assumption that individuals can and do act fully rationally on the basis of their preference profiles. Who is right? Who is closer to what evolutionary theory in general, and EP in particular have to say about these issues, Williamson or proponents of the IEA? Let us not rush to overhasty answers, but have a closer look at EP first.

Evolutionary psychology (EP) in a nutshell What is EP? In Vromen (1999) I argue that the following two claims put forward by EP can be seen as its essential features (see also Kennair 2002): 1

2

The human mind and its workings can best be understood in terms of a repertoire of separate modules. These modules are psychological mechanisms that are domain-specific (or special-purpose), contentspecialised and context-sensitive (this is sometimes called the Swiss army knife model of the human mind). Each psychological mechanism is an adaptation to a particular subset of evolutionary pressures prevailing in the Pleistocene era’s huntergatherer society (this is sometimes phrased as: our skulls house a Stone Age mind).

The second claim states that each psychological mechanism evolved as a solution to some particular adaptive problem in the distant past. There is no general presumption that our psychological mechanisms are well adapted to the circumstances prevailing in our present society.9 If our present society differs markedly from the hunter-gatherer’s society in which our ancestors lived in relevant respects, then it may well be the case that our ‘stone age’ mechanisms prevent us from arriving at optimal solutions to problems we are faced with now. To understand what is meant by the first claim better, it is instructive to turn to the pioneering work of Cosmides and Tooby that is perhaps still the most exemplary work in EP to date.10 A discussion of Cosmides and Tooby’s research probably also best brings out the gist of what is going on in EP. The issue preoccupying Cosmides and Tooby (1992) is that of cooperation.11 Individuals who in early times of mankind have been relat-

Cognitive science meets evolutionary theory 59 ively successful in engaging in mutually advantageous forms of co-operation, Cosmides and Tooby argue, have had an evolutionary advantage over those who have been less successful. Cosmides and Tooby rely on game-theoretic treatments as to what type of co-operative behaviour might possibly have evolved. In their view these treatments have culminated in Axelrod’s (1984) computer tournaments showing that Tit-for-Tat is a superior decision rule for a wide class of repeated PDs. If all individuals in a population play Tit-for-Tat, Cosmides and Tooby argue, then they can all be said to abide to a social contract. For they all can be said then to subscribe to the unwritten rule not to give in to the temptation to defect. If all play Tit-for-Tat, no one cheats on the social contract. Or, as Cosmides and Tooby put it, everyone incurs the cost (by not defecting) to reap the benefits of the social contract. Cosmides and Tooby spell out the requirements that are to be met for Tit-for-Tat to be feasible (or computable) for individual agents. They put these requirements in terms of algorithms (that are mostly related to numerical estimates agents make and to comparisons of these estimates) and decision rules (that link comparisons to particular types of actions to be taken). In the case of Tit-for-Tat they list fourteen algorithms and decision rules in total. The algorithm that captures Cosmides and Tooby’s particular attention is the one enabling agents to detect cheaters. For Titfor-Tat to work, agents must be able to register who is cheating. Otherwise they would not be able to retaliate selectively against those who have defected on the last occasion. Thus Tit-for-Tat entails the hypothesis that agents master the cheaters’ detection algorithm. Cosmides and Tooby subsequently report the empirical work that they themselves and others conducted to test this hypothesis. The Wason Selection Task is of particular importance here. Peter Wason was interested in the issue whether everyday learning corresponds to Popperian hypothesis testing – that is, whether people look for empirical evidence contradicting a hypothesis in their everyday learning. In order to find this out, Wason presented a formal representation of a conditional hypothesis to persons (‘If P then Q’) and confronted them with a choice between four cards. (The cards are said to contain all information about a case under consideration.) Each card has on the one side either P or not-P and on the other side Q or not-Q. Of each card only one side is showed to the persons. The first card has P, the second has not-P, the third has Q and the fourth has not-Q on top. The persons are asked what cards they have to check (to inspect their reverse) in order to find out whether the conditional hypothesis is violated. It should not be too difficult for people (especially for those who had some training in first-order logic), Wason thought, to figure out the right answer: the first and fourth cards have to be examined. For only a card with P on the one side and not-Q on the other can possibly contradict the hypothesis. But to his surprise less than 25 per cent of the respondents

60 Jack J. Vromen gave the right answer. Subsequent experiments in which persons were provided with real-to-life examples (rather than a formal representation of them) showed that the percentage of respondents giving the right answer is rather sensitive to the type of propositional content given to the conditional hypothesis. Changes in this percentage due to changes in the type of propositional content given to the hypothesis are called content effects. It turned out that the percentage raises to the highest level (75 per cent) if the terms in which the hypothesis is put is recognisably related to a ‘standard social contract’ problem. Thus, if the conditional hypothesis is presented as ‘If you take the benefit, then you pay the cost’, most respondents give the right answer and inspect the first and fourth card. Cosmides and Tooby take this as prima facie empirical support for their hypothesis that people indeed follow a cheater detection algorithm when faced with recognisable social contract problems.12 General-purpose rules, such as modus ponens and modus tollens in first-order logic, are not flawlessly applied by people across all domains. Rather, people follow specialpurpose rules that have evolved to solve specific problems in specific domains such as social interaction. These rules are content-specialised (or content-dependent) rather than content-independent: they are activated only by particular content domains (Cosmides and Tooby 1992: 166). Only in particular content domains certain special representational formats (in terms of costs incurred and benefits reaped, for example) do suggest themselves that direct the attention of the individuals involved to certain relevant factors and actions. In social contract problems, for example, people are particularly keen to spot people who reap benefits without incurring the corresponding costs. What we see here is that EP need not bring in evolutionary theory and evolutionary insights just as an add-on to (or rationale for) already and independently identified modules and mechanisms. Evolutionary theory and evolutionary insights may also provide a heuristic for identifying modules and mechanisms. This is what Cosmides and Tooby do: on the basis of evolutionary game-theoretic analyses Cosmides and Tooby hypothesise that humans avail of a specific cheater detection algorithm. Buss (1995: 12) calls this the function-to-form direction in which evolutionary analysis of psychological mechanisms can (and does) proceed. In a function-to-form analysis one starts by identifying a function (or adaptive problem, or ‘a lock’) and then searches for the form (or solution, or a key) that might fit. Alternatively, one could start with already and independently identified forms (mechanisms) and look for functions that these forms could serve. This is the form-to-function direction in which evolutionary theory can be brought to bear on the study of psychological mechanisms. Buss argues that EP proceeds in both directions. Cosmides and Tooby conclude that there is a special-purpose, contentspecialised algorithm for detecting cheaters in social contract problems. The cheater detection algorithm is just one of the algorithms, or modules,

Cognitive science meets evolutionary theory 61 our human mind is equipped with, EPists argue. Many more have evolved in the evolution of our species. Each of them evolved in a special domain of life, ‘designed’ by natural selection to solve a particular evolutionary problem within that domain.13 One of the distinguishing features of EP within evolutionary theory, it can be argued, is its fine-grained partitioning of the ultimate desideratum in biological evolution, relative reproductive success, into particular evolutionary pressures. This is what ‘specialpurpose’ and ‘domain-specificity’ are all about: they refer to the special purpose for which and the specific domain in which our ‘Darwinian’ algorithms originally evolved way back in time. ‘Content-specialised’ refers to the special content that informational input now has to have in order to activate some particular algorithm (or modules). A module is triggered only if stimuli with the right propositional content are provided. What about the third property EP assigns to our mind’s modules? What does it mean to say that our modules are ‘context-sensitive’? It basically means that our modules produce different representational and behavioural outputs in different contexts. Modules do this if they have a conditional form. People with such conditional modules who are exposed to different local circumstances will exhibit different representational and behavioural patterns.14 In this way, proponents of EP try to explain how cultural diversity comes about even if people across different cultures share the same repertoire of modules. After having elucidated the meaning of the three essential properties that EP ascribes to the mind’s modules, one may still be left wondering what modules (or, for that matter, psychological mechanisms) are. What stuff are they made of? Are they spatially separable brain circuits? No, strictly speaking they need not coincide with separate brain circuits.15 In EP’s scheme of things, mechanisms are functionally defined. In line with its overall evolutionary problem-solving orientation, mechanisms are identified by the tasks they perform. On Cosmides and Tooby’s own selfunderstanding EP engages in task analyses. What EP has to offer is a computational theory in David Marr’s terms (Cosmides and Tooby 1992: 178, 205). According to Marr, ‘an algorithm is likely to be understood more readily by understanding the nature of the problem solved than by examining the mechanism (and the hardware) in which it is embodied’ (Marr 1982: 27). Similarly, Cosmides and Tooby aim at the cognitive level of explanation: ‘the cognitive level of explanation describes psychological mechanisms in functional terms, as programs that process information’ (Cosmides and Tooby 1987: 283–284). Insofar as Cosmides and Tooby have something to say at all about brains and neurosciences, they do so in cognitive terms: ‘Because the function of the brain is informational in nature, its precise functional organization can be described accurately only in a language that is capable of expressing its informational functions – that is, in cognitive terms, rather than in cellular, anatomical, or chemical terms’ (Tooby and Cosmides 1995: 1190).

62 Jack J. Vromen EP believes that our psychological mechanisms dispose us to think, learn and behave in certain pre-formatted or pre-programmed ways. It is stressed in EP that we are often not aware of this. But this does not mean that EP holds that engagement in deliberate and purposeful thought and action is denied by our mechanisms. Buss (1995) argues that EP is fully consistent with the idea that individual human beings are (or at least can be) involved in goal-directed behaviour.16 But Buss also remarks that when we engage in goal-directed behaviour, underlying psychological mechanisms may affect the goals that we pursue and the ways in which we try to accomplish our goals without us knowing this. EP emphatically does not rule out the possibility that a lot of our behaviour is learnt rather than genetically instructed (Symons 1979: 9). EP does not want to be drawn into traditional nature versus nurture debates (Cosmides and Tooby 1987, 1992: 209). EPists rather want to transcend such traditional dichotomies. What EPists emphasise is that any learning effort always requires an evolved capacity to do so and that learning efforts always are directed or guided by psychological mechanisms to some degree. EP is quite controversial. It has been criticised from many different angles. Indeed, it can be argued that almost any aspect of EP is contested. EP’s alleged adaptationist overtones are criticised (Gould 1991, 1997), its hypothesis that there is ‘massive modularity’ is contested on both theoretical and empirical grounds (Samuels 1998; Fodor 2000), its lack of sufficiently carefully conducted empirical research and tests is criticised (Lloyd 1999; Downes 2001), its failure to connect with frontiers in the brain and neurosciences is criticised (Buller 2000). There are also criticisms that culture (and especially cultural transmission) is given short shrift by EP (Tomasello 1999), that EP neglects the role of ontogenetic development (Stotz and Griffiths 2001; Griffiths 2002), and that the claims made in EP transcend what can be substantiated by science (Dupré 2001). But, of course, the mere fact that EP is accused by many does not yet establish that it is ‘guilty’. The present author, who does not feel fully competent to adjudicate these debates, got the impression that many of the accusations made are heavily ideologically charged (which, of course, also does not by itself prove the accusations wrong). Furthermore, in order to get a more balanced picture it must also be acknowledged that EP seems to gain momentum in psychology (Kennair 2002). At any rate, it is too early to tell what the fate of EP in Academia will be.

The bones of contention I: motives Armed with at least a preliminary understanding of EP, we are better prepared now to answer the question who is closer to the general thrust of EP, Williamson or the IEA? Let us consider the motivational or affective aspect of behaviour first. What sorts of motivations, preferences, ‘passions’, feelings and the like are we equipped with, according to evolution-

Cognitive science meets evolutionary theory 63 ary theory in general, and EP in particular? This for sure is not the type of question that can be answered in a simple and straightforward way. What can be argued, however, is that several arguments and assertions put forward by both Williamson and proponents of the IEA are flawed and misguided. Williamson approvingly quotes Dawkins in order to find support for his view that opportunistic agents can find farsighted solutions for the problems they have to cope with: ‘One unique feature of man . . . is his capacity for conscious foresight.’ Indeed, Williamson goes on to argue, it is the ‘capacity to simulate the future in imagination . . . [that saves] us from the worst excesses of the blind replicators’ (Dawkins 1976: 200). Williamson takes Dawkins to be arguing that the worst excesses have their origins in our selfishness. There is no doubt that proponents of EP share the view that we can escape the tyranny of our selfish genes if we want to (Symons 1979). We are not the passive slaves or obedient executors of the genes that we carry. Williamson seems to attribute two features to Dawkins’s ‘selfish gene’ perspective, however, that neither Dawkins himself nor proponents of EP would accept. First there is the idea that unless counteracted by other forces such as farsightedness, our selfish genes would inflict myopic excesses upon us. The second feature is that these excesses have their origins in our selfishness. Dawkins does not assume that selfish genes necessarily lead to myopic behaviour. If we are organisms capable of farsighted behaviour and contracting, as Williamson assumes, then there must be selfish genes that make this possible. ‘Selfish genes’ are a prerequisite of farsighted behaviour rather than one of the forces competing with farsighted contracting in determining behaviour. The point Dawkins is trying to make is not that our selfish genes are blind, myopic forces that can be put on a par with other capabilities or forces that determine our behaviour. Genes do not affect our behaviour directly. Genes affect our behaviour only indirectly. Genes code for certain capabilities. These capabilities can develop into mechanisms that affect behaviour directly. The point Dawkins wants to make is that we can use our capability for farsighted behaviour to interfere in processes of biological evolution if we do not like the course biological evolution would be running without interference of this capacity. Dawkins does not argue that organisms hosting selfish genes must themselves be or behave selfish too. Dawkins calls genes selfish not because he ascribes some hidden agenda to them that they pursue secretly, inducing the organisms that they inhabit to behave accordingly. The notion of a selfish gene Dawkins coined as a convenient shorthand for expressing that no matter what are the immediate, proximate causes determining biological evolution’s course, what ultimately counts in biological evolution is relative reproductive success of genes in a population’s gene pool.17 The fact that in biological evolution a trait can only evolve if the ‘selfish genes’ of the vehicles (or interactors) having the trait are

64 Jack J. Vromen propagated in subsequent generations, does not imply that any such trait itself (or the vehicles having the trait) must be selfish too. Assuming that it does amounts to conflating two different levels of biological organisation. Selfish genes can make for genuine non-selfish organisms at a higher level of organisation. To acknowledge that such organisms are products of biological evolution (and, in a sense, vehicles of selfish genes) does not imply that these organisms cannot but be selfish ‘deep down below’. Dawkins makes it perfectly clear that he believes that biological evolution can bring about genuine altruism. In certain evolutionary settings the genes of genuine altruists are reproductively more successful than non-altruist organisms.18 Thus Williamson is wrong in arguing that evolutionary theory in general, and EP in particular, supports his assumption that individuals are opportunistic. As EP predicts, and as experimental evidence seems to substantiate, individuals go to great length (and even at considerable personal costs) in detecting and punishing those who threaten the prevailing social contract, for example. One could object here that ‘the iron logic’ of evolutionary theory demands that this eventually should bring the detectors – punishers (or at least their ‘selfish genes’) personal benefits. But this does not imply that the detectors – punishers (let alone their ‘selfish genes’) detect and punish because they expect to benefit from doing so. Evolutionary theory leaves room for the possibility that individuals who are committed to co-operate (or at least to reciprocate co-operation by others) for non-instrumental reasons have an evolutionary edge over others, not despite, but precisely because of their commitment (see, for example, Frank 1988). In this respect proponents of the IEA seem to have the better arguments. Proponents of the IEA do acknowledge our ability to engage in genuinely altruistic behaviour. What about the cognitive aspect of behaviour? What does evolutionary theory in general and EP in particular have to say about rationality?

The bones of contention II: rationality There does not seem to be a consensus in evolutionary theory whether or not human behaviour falls short of the ideal of perfect rationality. Some argue that natural selection has endowed us with a general-purpose device that enables us to find perfectly rational, optimal solutions to problems across the board. In economics, for example Aumann (1997) and Robson (2001) hold this view. This is emphatically not how EPists look at the relation between rationality and evolution, however. As noted above, one of the characteristic features of EP is precisely its argument that natural selection must have equipped human beings with a multitude of specialpurpose psychological mechanisms. There has never been such a thing as a general adaptive problem, it is argued. In ancestral environments there only have been special adaptive problems, and special problems call for

Cognitive science meets evolutionary theory 65 special solutions: a multitude of special-purpose devices make for more efficient solutions for these problems than one general-purpose device. Now how rational are these alleged special-purpose psychological mechanisms? Such mechanisms may well have outcompeted rival mechanisms in the distant past, as EPists argue, but do they still yield optimal solutions for the problems we face today? Not necessarily, as EPists themselves acknowledge. It depends. If our present environment is sufficiently similar to the ancestral environment in which our mechanisms emerged (often called the EEA: the environment of evolutionary adaptedness), then we may expect the mechanisms to still perform well (Kennair 2002: 24). But if that is not the case, then the mechanisms may prompt us to go and search in the wrong directions. Sometimes our evolved psychological mechanisms enable us to perform even better than rational (Cosmides and Tooby 1994). This is because the mechanisms often act as shortcuts. The mechanisms often generate behaviour quite quickly and effortlessly. They thus save on time and energy. A general-purpose device by contrast may engender behaviour only after having conducted a time- and energy-consuming search process. In general, whereas a general-purpose device is slow but ‘clean’, a specialpurpose device is quick but dirty (see also LeDoux 1996). This means that having a system of special-purpose mechanisms can also have its drawbacks. The system only yields good solutions if the problems still resemble the problems that gave rise to its emergence. Only then does the dirtiness of the system not bring about poor performance. It is perfectly possible that because of its ‘cleanness’ (or precision, or accuracy) a generalpurpose mechanism, though slower, does better in unprecedented environments. But if EPists are right, this option is simply not available to us. We simply do not possess a general-purpose mechanism. Let us get back to Williamson and the IEA now. Williamson argues that EP supports TCE’s assumption of bounded rationality. More specifically, he argues that EP endorses the view that in decision-making agents are guided by search heuristics and that because of this they are at least sometimes prone to make judgement errors. Williamson here refers to work in the so-called ‘Heuristics and Biases Program’ in experimental psychology. In their experiments, people like Daniel Kahneman, Amos Tversky and Richard Thaler have collected many biases, puzzles, and effects that they take to be anomalies for standard expected utility theory. It seems that Williamson is prepared to go along with the conclusions drawn in the ‘Heuristics and Biases Program’. On the other hand, Williamson takes sides with Gigerenzer (1996) in his critique of this program. Gigerenzer accuses Kahneman and Tversky of paying insufficient attention to the fact that some biases and puzzles may disappear if the problems subjects in experiments are confronted with are put in different terms. For example, if problems of choices under uncertainty are put to subjects in point estimates they appear to be poor statisticians. If

66 Jack J. Vromen the same problems are put to them in frequentist terms, however, they perform much better. Gigerenzer (1996) and especially Cosmides and Tooby (1996) argue that this remarkable difference in performance is due to the fact that problems of choice under uncertainty that our hominid ancestors had to deal with presented themselves in frequentist terms and not in terms of point estimates. A specialised frequentist mechanism evolved, they go on to argue, that is subsequently activated only if the right sort of informational input is provided. The conclusion Williamson seems to draw from this is that if there is something that is to be called irrational in Kahneman and Tversky’s experiments it is not the behaviour of the subjects in Kahneman and Tversky’s experiments but Kahneman and Tversky’s experimental design. As Williamson sees it, the problem with Kahneman and Tversky’s experimental design is that it is not informed by evolutionary considerations. Apparently Williamson believes that experimental results are more reliable if circumstances are approximately rebuilt in the experimental design that our hominid ancestors were faced with and in which our psychological mechanisms took shape. But this of course begs the question. It begs the question whether our present-day conditions and environmental circumstances ‘in the wild’ (that is, outside laboratory-like experimental settings) resemble the EEA closely enough to warrant the expectation that our evolved mechanisms bring about rational behaviour. After all, what is the point of attempts to recreate ancestral conditions in experiments if current conditions in actual life are markedly different from these conditions? If a lot of information in relevant real-to-life choice situations now comes to us in terms of point estimates, we are poor statisticians in these actual choice situations. It does not help to argue then that we once were reasonably good statisticians in the then prevailing conditions and that we can still be good statisticians if conditions are suitably altered.19 At any rate, Williamson’s take on the Gigerenzer–Kahneman and Tversky debate seems to reflect his overall attitude towards the issue of rationality. Although Williamson assumes bounded rationality, he does not believe that the bounds on and limitations in decision-making and problem-solving are very tight and restrictive. Individuals are assumed not to be able to foresee all possible future contingencies (hence complete contracting is impossible), but they are able to foresee possible hazards and take precautionary measures to prevent them from occurring (in the form of farsighted contracting). While Williamson accepts bounded rationality, he does not accept what Simon appeared to take as an integral part or as a corollary of it: satisficing. It seems that Williamson always regarded accepting satisficing in economic theorising as one step (or even a few steps) too far on the slippery slope to irrationality. All this shows that Williamson does not only cling to full or perfect rationality as some sort of self-evident benchmark in economic theorising, he also believes that the

Cognitive science meets evolutionary theory 67 degree of rationality to which actual individuals are capable in contracting falls short of this benchmark by only a tiny little bit. Like Williamson, Gigerenzer and his ABC group (the Center for Adaptive Behavior and Cognition; Gigerenzer et al. 1999; Todd and Gigerenzer 2000; Gigerenzer and Selten 2001) also accept bounded rationality. But unlike Williamson, Gigerenzer argues that a further theoretical elaboration of the notion of bounded rationality only can take either of the following forms: either models of satisficing are to be developed further (as Selten has been doing), or our ‘adaptive toolbox’ of fast and frugal heuristics is to be explored further (as Gigerenzer himself and his ABC group are doing). Both of them take a classic text of Simon as their starting-point.20 Models of satisficing primarily draw on Simon (1955), in which Simon concentrated on the limitations of actual decision-making organisms. Simon (1956) views decision-making from a complementary perspective. Here the focus is on the structure of the organism’s environment. Simon points out that the action-guiding principles of an organism need not be complex or sophisticated for the organism to satisfy its basic needs. The organism can succeed in doing this if it exploits clues in the environment to orient its search efforts. Gigerenzer’s ABC group’s notion of fast and frugal heuristics builds upon this idea: ‘fast and frugal heuristics employ a minimum of time, knowledge, and computation to make adaptive choices in real environments . . . the purest form of bounded rationality is to be found in fast and frugal heuristics, which employ limited search through objects (in satisficing) or cues and exploit environmental structures to yield adaptive decisions’ (Todd and Gigerenzer 2000: 731). Note how similar the ideas of Gigerenzer’s ABC group are to EP in this respect. Like EP’s psychological mechanisms the ABC group’s fast and frugal heuristics short-circuit decision-making, enabling individuals to make decisions with little sacrifice of time and energy. Indeed, the similarity between the two groups goes further than this. When proposing to replace traditional definitions of rationality by the new conception of ecological rationality, Todd and Gigerenzer draw on Cosmides and Tooby. The guiding idea behind ‘ecological rationality’ is that there are many environmental structures, each calling for a particular structure of heuristics. This also brings Todd and Gigerenzer to the notion of the adaptive toolbox: ‘the collection of specialized cognitive mechanisms that evolution has built into the human mind for specific domains of inference and reasoning, including fast and frugal heuristics’ (Todd and Gigerenzer 2000: 740). Their obvious disagreements notwithstanding, both Williamson and proponents of the heuristics and biases program accept full or perfect rationality as the benchmark.21 Williamson seems to believe that the degree to which actual individuals behave rationally pretty close approximates full or perfect rationality. Kahneman, Tversky and their pedigree claim that actual individuals behave much more irrationally. Despite these differences, all of them seem to share the view that the distance or

68 Jack J. Vromen deviation from full or perfect rationality indicates the degree to which individuals behave irrationally. In this view bounded rationality always involves some degree of irrationality. Todd and Gigerenzer disagree: ‘Bounded rationality is neither limited optimality nor irrationality’ (Todd and Gigerenzer 2000: 739). Instead, for them bounded rationality signifies a qualitatively different notion of rationality, ecological rationality, in which the focus is on the match between different specific environmental structures and the structure of the fast and frugal heuristics in the adaptive toolbox. What about the IEA? We saw that typically in the IEA full or perfect rationality is assumed. The idea is that when our basic preferences evolved long ago our ancestors acted rationally. That is, they acted in accordance with their preferences at that point in time and with the then prevailing circumstances. Thus proponents of the IEA apparently believe that a long time ago, before our current basic preferences evolved, our ancestors were already equipped with a general capability to behave optimally, given their preferences and circumstances at that time. This seems to be a highly improbable historical scenario. If there is a chronological order here, the general consensus in evolutionary theorising seems to be that it is more probable that our current capabilities to act rationally, whatever their precise nature and character, evolved only after our basic preferences were established. Proponents of the IEA have turned out to be sensitive to this criticism. Güth and Kliemt accept that ‘ “rationality” itself is the outcome of natural selection’ (Güth and Kliemt 1998: 379). Güth and Kliemt want to show that this insight can be accommodated within the IEA. Given this ambition, one would perhaps expect Güth and Kliemt to engage in an attempt to point out how rationality could have evolved before our basic preferences took their definitive shape. What they set out to do instead, however, is to demonstrate that there can be a two-tiered dynamic process in which short cycles, in which people learned how to adapt their behaviour to their then prevailing preferences, are embedded in long cycles, in which our basic preferences evolved. Güth and Kliemt argue that if the time and costs involved in the short cycles are negligible, the short cycles do not disturb the long cycles of preference formation that proponents of the IEA envision. On the basis of this ‘argument’, Güth and Kliemt can maintain that the IEA does not need to assume perfect rationality. People need not be endowed with a general capability to display optimal behaviour instantaneously for the general historical scenario depicted in the IEA about the evolution of our basic preferences to be sustained. What Güth and Kliemt tacitly assume, however, is that when our current basic preferences evolved, people were already endowed with a general capacity to learn relatively easy (or effortlessly) in a relatively short period of time how to adjust their behaviour to their preferences. This assumption runs counter

Cognitive science meets evolutionary theory 69 to one of the most essential ingredients of EP. EP holds that we do not avail ourselves of a general problem-solving or a general all-domain learning device. We only possess an array of special problem-solving and domain-specific learning devices. In some problem situations these devices enable us to arrive at good or even optimal solutions in a relatively short, costless and reliable learning process. In other problem situations, however, in situations in which we encounter unprecedented problems, for example, our devices may rather stand in the way of finding good solutions. Let us recapitulate the findings about the cognitive aspects of human behaviour that we have gathered so far. Like Williamson, proponents of the IEA retain perfect rationality as a useful benchmark. In this respect, both camps are equidistant from what especially Gigerenzer and his ABC group recommend: reject the demon-like notion of perfect rationality as a benchmark and replace it with a more realistic one. Unlike Williamson, however, proponents of the IEA seem to be blind to the possibility that at least sometimes our evolved search heuristics disable us to find optimal solutions for present-day problems. Sometimes we may be able to arrive at an optimal solution only after having gone through a time- and energyconsuming search or learning process. Sometimes we may even be unable to ever find an optimal solution. In this regard, Williamson seems to be closer to EP than the IEA.

Does EP call for a radically different view on human behaviour? Our results so far have been mixed. It was argued that, with respect to the motivational or affective component of behaviour, proponents of the IEA seem to be closer to EP than Williamson. There are situations in which individuals are led by genuinely altruistic (and hence non-self-interested and non-opportunistic) motivations. But with respect to the cognitive component EP seems to vindicate Williamson’s rather than the IEA’s assumptions about rationality. There are situations in which individuals are unable to arrive at perfectly rational choices. EP suggests that the degree to which individuals are able to behave rationally depends on the psychological mechanisms (and associated search heuristics) available and on the environmental circumstances that activate them. Yet it can be argued that the real message that EP entails for economic theory is less reconciliatory with either Williamson or proponents of the IEA than indicated so far. EP, it can be argued, suggests an altogether different view on how individual behaviour is generated, and how environmental circumstances affect this, than the view that Williamson and the IEA endorse. Beneath the disagreements between Williamson’s and the IEA’s view (as outlined above) there seems to be a basic agreement about how

70 Jack J. Vromen individual behaviour is generated. Williamson and proponents of the IEA disagree on whether appearances of altruistic behaviour are illusory or real and on whether individuals arrive at perfectly rational choices. But underlying these disagreements there is an agreement about the major constituents of individual behaviour and their basic properties. Both Williamson and proponents of the IEA assume that preference profiles and some degree to which individuals are able to act rationally are the major constituents of individual behaviour. Furthermore, they all assume that these two sorts of constituents are separable elements or components in decision-making processes. Finally, they all assume not only that different individuals have (roughly) the same preference profiles and the same degree of rationality, but also that the individuals’ preference profiles and degree of rationality are invariant across various environmental circumstances. Differences in behaviour between different individuals are accounted for in terms of different opportunities and different prices and costs. The same holds for differences in behaviour that the same individual displays in different environmental circumstances. Here too the differences are accounted for in terms of different configurations of opportunities and costs. EP seems to suggest a view that is markedly different from this. The only point of convergence, it seems, is that EP also assumes that when it comes to understanding how individual behaviour is generated, differences between individuals can be glanced over. EP focuses on psychological mechanisms that individuals have in common with each other. In this sense EP rehabilitates the notion that human nature is the same everywhere. But this does not mean that EP supports the notion of a unitary self. Quite to the contrary (see also Field 2001: 58). Unlike Williamson’s TCE and the IEA, EP does not subscribe to the view that the major behaviour-generating constituents are invariant across various environmental circumstances. As was already argued above, EP holds that the degree to which individuals are able to arrive at rational choices varies from the one sort of environmental circumstances to the other. If situated in the one sort of environmental circumstances, the psychological mechanisms that we are equipped with facilitate quick and optimal responses. If other sorts of circumstances obtain, our mechanisms enable us to arrive at optimal responses, but only after having gone through a laborious learning process. In yet other sorts of circumstances, we may never be able to find optimal responses. The sorts of environmental circumstances that EPists have in mind are quite unlike those that the economic theorists discussed above have in mind. According to the economic theorists, the essential dimensions of environmental circumstances relate to opportunities and costs. The general idea is that given their fixed preference profiles and fixed capability to act rationally, individual behaviour responds flexibly to changes in configurations of opportunities and costs. The image that EP suggests is

Cognitive science meets evolutionary theory 71 quite different. For EPists, the essential dimensions of environmental circumstances do not relate to present opportunities and costs but to the specific psychological mechanisms that they activate. The general idea here is not that the same psychological mechanisms are at work no matter what environmental circumstances prevail, but rather that different mechanisms are at work, depending on what sort of circumstances prevail. On the basic view of the economic theorists, it can be argued, prevailing opportunities and costs determine the particular sort of environmental circumstances that individuals are confronted with. Each different configuration of opportunities and costs makes for a different sort of environmental circumstances. By contrast, the partitioning of environmental circumstances into different sorts in EP is based on the specific psychological mechanism that is activated. Environmental circumstances belong to the same sort if they activate the same mechanism; they belong to another sort if they activate different mechanisms. EPists argue that environmental circumstances of the same sort are all in some way reminiscent of the original evolutionary problem that once gave rise to the mechanism in question. Yet, as Sperber (1996) emphasises, the environmental circumstances that now activate some mechanism may also be very different in many respects from the ones in the Pleistocene era responsible for the evolution of that mechanism. In particular, the ways in which environmental circumstances are now organised and categorised culturally may be quite unlike how environmental circumstances presented themselves to our ancestors.

An untidy bunch of input–output flow charts Sperber argues that the issue at stake can be put in terms of input conditions. Each mechanism or module has its own specific input conditions. Only if information is provided that meets its input conditions will a mechanism be activated. Given that EP adopts a computational approach towards psychological mechanisms, an input–output flow chart indeed seems to be well suited to visualise the workings of a psychological mechanism. The input to some mechanism is some piece of information satisfying its input conditions. Its output again consists of a piece of information or of overt behaviour. In between, which can be called the functioning of the mechanism proper, we have information-processing. Although EPists do not descend to the algorithmic level, and hence does not go into the details of the algorithms involved (Marr 1982: 24–27; Kennair 2002: 23), it is clear that EPists conceive of information-processing in terms of algorithms. In particular, they assume that inference rules and decision rules take care of the production of output, given some input. Now ‘inference rules’ and ‘decision rules’ may suggest that for EP the workings of a psychological mechanism are an altogether cognitive affair in which affective elements have no part to play. But this impression is

72 Jack J. Vromen misleading.22 EP maintains that affective and cognitive components are inextricably intertwined in the operation of psychological mechanisms. Economists who believe that psychological mechanisms can be treated as basic preferences may be wrong (for reasons that we are currently discussing), but they are on to something. Psychological mechanisms do relate to what we want, what we like, what we find agreeable and beautiful and the like. But the manner in which our desires and tastes affect the ways in which we process information (and, ultimately, our behaviour) differs from how economists see the relation between (basic) preferences and behaviour. Our desires and tastes are not homogenised in some preference ordering from which some goal can be derived that we pursue under all conceivable circumstances.23 Our desires, tastes, passions and instincts are more like predispositions (or inclinations, or propensities). They are not all active all of the time. Typically only one (or at most a few of them) manifests itself at a time. If manifested, they pull our learning efforts (including the inferences that we make), and our behaviour most of the time, unconsciously in certain directions. They colour our perceptions and bias what we do on the basis of them. The image of an input–output flow chart can also be fruitfully put to use to sum up the profound differences between Williamson’s TCE and the IEA on the one hand, and EP on the other. The input–output chart representing how Williamson and proponents of the IEA believe individual behaviour is generated is relatively simple (but quite demanding of humans). There is only one general-purpose, domain-unspecific algorithm linking informational input to behavioural output. On the input side, all information is assumed to satisfy the input conditions for the algorithm. The algorithm pays attention to opportunities and costs as the crucial variables in the environment. The algorithm (doing the information processing) itself reflects the interaction of two distinguishable constituents: a preference profile, as the affective or motivational constituent, and a general capacity to act rationally (at least to some degree) on the basis of the specific informational input provided and the preference profile, as the cognitive constituent. Given the specific informational input provided (some configuration of opportunities and costs), the algorithm directly produces behaviour as output. Differences in output are thus entirely due to the differences in environmental circumstances. It is always the same algorithm that is operating over and over again. Compare this with the input–output flow chart (or, rather, flow charts) in EP. There are many special-purpose, domain-specific algorithms.24 Each algorithm has its own input conditions. An algorithm is only triggered if some piece of information satisfies its input conditions. What aspects or variables of the environment an algorithm pays attention to depend on the algorithm in question. Differences in algorithms reflect different mixtures of affective and cognitive elements. It is hard, if not impossible, to disentangle these elements. Algorithms need not produce behaviour as output. They may also produce

Cognitive science meets evolutionary theory 73 new pieces of information that meet the input conditions of other algorithms. Thus there may exist sequences of algorithms in which the output of the one algorithm serves as input to another algorithm. We may conclude that the input–output flow charts of EP are considerably more complex and less neat than the economists’ flow chart. This is so for two reasons. First of all, instead of one algorithm having no input conditions, we have many algorithms all having their own input conditions. Second, instead of one algorithm leading linea recta to behaviour, we typically have sequences of interlocked algorithms. This means that we need to know a good deal more about the algorithms, their workings, their respective input conditions and the ways in which they are interlocked in order to tell what output is produced if some input is provided.

What difference does it make? At this juncture one may wonder what difference it makes whether the neat and simple flow chart of the economists or the messy and complex flow charts of EP are approximately right. What implications do any of these have for the types of individual behaviour that we can expect under various environmental circumstances? If the charts have the same implications, economists may not be terribly interested in replacing their own neat and simple chart by the messy and complex charts of EP. There are good reasons, both theoretical and empirical ones, however, to expect that the charts have different implications. The theoretical reason stems from the fact that the economists and EP envision different keys for the partitioning of sorts of environmental circumstances. For the economists each configuration of opportunities and costs constitutes a sort of its own. For EP there are as many sorts of environmental circumstances as there are psychological mechanisms. There is no a priori reason to assume that the sorts of environmental circumstances that economists envision coincide with those envisioned by EP. In fact, there are good reasons to assume that the two do not coincide since EPists argue that the partitioning of current sorts of environmental circumstances is reminiscent of the partitioning of evolutionary problems that our ancestors faced. This assumption seems to be supported by experimental findings. As Cosmides’s version of the Wason selection task shows, a different framing of what appears to be the same chunk of information may lead to a very different performance. If we look at this from the perspective of the economists’ flow chart this is a miracle. After all, the difference in framing has no consequences for the opportunities and costs that individuals are faced with. And, as we have seen, a difference in performance can only be forthcoming in the economists’ scheme of things if there are changes in the configuration of opportunities and costs. EP is tailor-made to explain findings

74 Jack J. Vromen like Cosmides’s. EP explains why output may be very sensitive to seemingly minor, if not negligible modifications of inputs: a slightly modified input may satisfy the input conditions of a different mechanism. In other words, a modification of informational input that seems to be negligible from the perspective of the economists’ partitioning may from the perspective of EP’s partitioning imply the transition to a new sort of environmental circumstances. Conversely, significant changes in the configuration of opportunities and costs (significant, that is, from the perspective of economic theory) need not lead to significant changes in behaviour. Experimental findings in co-operative and bargaining games suggest that subjects are willing to reciprocate co-operative behaviour of others even at a considerable personal cost. Reciprocation of co-operative behaviour is only intelligible in the economists’ scheme of things if environmental circumstances are such that there are (expectations of) personal benefits involved. It is a miracle if environmental circumstances are such that there are personal costs involved. EP suggests that in cases like this the relevant environmental circumstances did not change significantly. They still belong to the same sort: the same mechanism remains activated. This may have consequences also for policy issues. What appears to be a rational policy from the perspective of economic theory may not be rational at all from the perspective of EP. In order to achieve some social goal, for example, economists may count on the price effect and advise a suitable adjustment of the incentive structure. EP may help explain, however, why this sometimes yields adverse effects. EP may help explain, for example, when and why crowding out of intrinsic by extrinsic motivation (or the justification effect; see Frey 1997) can be expected. Empirical evidence indicates that sometimes adverse effects are produced when some authority manipulates the incentive structure in order to enhance contributions of individuals (to produce more public goods, for example). An explanation of this in line with the overall thrust of EP’s framework is that because of the changes in the incentive structure (and presumably also because of the way in which this is presented) another, more self-serving mechanism is activated than the less selfserving mechanism (based on a sense of civic duty, for example) activated so far.

Concluding remarks EP, we have seen, allows for the possibility that individual human beings are led by non-opportunistic considerations. In this respect, the IEA seems closer to EP than Williamson. But in another respect Williamson seems to be closer to EP than the IEA: EP suggests that at least sometimes our stone-age psychological mechanisms produce behaviour that is ill-adapted to current circumstances. But we have also seen that it is doubtful whether

Cognitive science meets evolutionary theory 75 EP allows for such a conceptual distinction between motivational and affective aspects of behaviour on the one hand and cognitive aspects on the other. EP seems to imply a view on human behaviour in which motivational (and emotional) and cognitive aspects are inextricably intertwined. One of the exciting features of EP is that it holds out the hope of transcending traditional dichotomies. Instead of assuming that behaviour is based on either emotion or cognition, EP assumes that behaviour always reflects both. Likewise, instead of assuming that either nature or nurture must have the greater effect on human behaviour, EP assumes that both are always involved in human behaviour in a way that makes it impossible (or at least silly) to assign relative weights to either of them. Strong cultural or environmental influences on human behaviour presuppose rather than defy ‘nature’: they presuppose a strong innate susceptability in human beings to these sorts of influences. This also implies that the dichotomy innate versus learned is misconceived as well. Any learning effort presupposes an innate capacity for making the effort and, furthermore, will be guided by pre-formed heuristics.25 It was also argued that EP might entail revolutionary implications for economic theorising. EP’s view that our psychological mechanisms are domain-specific (or special-purpose) and content-specialised seems to imply an understanding of human behaviour that is radically different, not just from both Williamson and the IEA but from almost any kind of economic theorising that we have witnessed so far. It can be argued that EP supports the notion of the fragmented or multiple self that has received only scant attention in economics (Elster 1979; Schelling 1984; Kuran 1995). EP conceives of the human mind in terms of a multitude of psychological mechanisms, each of which is activated only under suitable circumstances. What is more, EP suggests that the relevant dimensions in environmental circumstances that determine which mechanism is activated may be quite different from the dimensions standard economic theory focuses on (to wit, personal benefits and costs). Thus, EP may be able to explain why individual behaviour sometimes is much less sensitive to changes in opportunities and costs than standard economic theory predicts. It may well be the case that many economists find such implications of EP too radical. They may not be ready yet to abandon their beloved theoretical framework. They may also find some theoretical justification for their resistance in the fact that EP is quite controversial and that it is still in its infancy.26 But if EP makes some headway, and if puzzling experimental evidence and anomalous empirical data accumulate, the time may be ripe for a wholesale reconsideration of the economic theory’s edifice.

76 Jack J. Vromen

Notes 1 For other interesting investigations of the relation between EP and economics, see Hodgson (1998a, 1998b) and Twomey (1998), Hoffman et al. (1998) and Friedman (2001). See also Frank (1988, 1999) for interesting attempts to infuse insights from evolutionary theory in general and EP in particular into economic theory. 2 According to Gintis, this is borne out in the experimental psychology literature. 3 See also Fehr and Gächter (2002) on the related notion of altruistic punishment. 4 The equality sought can pertain to chances, opportunities and treatment of the individuals involved. If so, Gintis’s Homo egualis is similar to what Ben-Ner and Putterman call process-regarding preferences. 5 See also Güth and Kliemt (1994, 1998), Güth (1995), Huck (1997), Fehr and Schmidt (1999), Bolton and Ockenfels (2000), Bolton (2000). 6 The crux of the matter here, it can be argued, is that ‘subjective’ valuations (determining what the players do) are different from ‘objective’ results (determining what type has survival value). 7 Güth and Kliemt argue that what makes the IEA indirect is that preferences influence evolutionary success indirectly (Güth and Kliemt 1998: 384). This suggests that in a direct evolutionary approach, such as evolutionary game theory, preferences influence evolutionary success directly. This suggestion is misleading, I think. In evolutionary game theory either preferences are not taken to be part of evolutionary processes at all (and then they do not influence evolutionary success at all), or they are taken into account, but then they must be at least part of what ‘evolutionary success’ means. In either way, it is misleading to say that preferences influence evolutionary success directly. This is why I opt for a somewhat different interpretation of ‘indirect’. 8 See also ‘while choices are motivated by the preferences, which may include an internal commitment to norms, objective evolutionary success of preference types depends on “external incentives” ’ (Güth and Ockenfels 2000: 336). 9 Some hold that EP is sociobiology dressed up with a new label. It can be argued, however, that there is at least one significant difference (Vromen 2001). Whereas sociobiology assumes that evolutionary forces directly impinge on manifest behaviour, EP assumes that evolutionary forces in a distant past influence our behaviour indirectly via psychological mechanisms that are adapted to the then prevailing circumstances (and not necessarily also to circumstances prevailing now). 10 It was Leda Cosmides who started this line of research in her Harvard University Ph.D. dissertation (see Cosmides 1989). 11 Cosmides and Tooby take co-operation to be an evolutionary problem falling under the more general rubric of social interaction. Other problems falling under the same rubric are those of mate selection, threat and coercion. Besides the rubric of social interaction Cosmides and Tooby also distinguish between the rubrics of food acquisition and disease avoidance, for example. 12 Cosmides and Tooby do not rest content with this result. They consider several rival explanations of the experimental findings at length. Several modified versions of the experiment are run in order to tell which explanation gets it right. On the basis of the results obtained the rival explanations are rejected. 13 In fact one can say that in EP’s scheme of things, domains are identified by the evolutionary problems to be solved. 14 EP’s notion of culture accompanying this is ‘evoked culture’. 15 In philosophical jargon this point is sometimes put in terms of multiple realis-

Cognitive science meets evolutionary theory 77

16 17 18

19

20 21

22

23

24 25 26

ability: one and the same algorithm can be instantiated in different material substrata. For a more elaborate discussion of this issue, see Vromen (2001). For a further discussion of this, see also Vromen (1995: ch. 7, sect. 5). Furthermore, as Sober and Wilson (1998) convincingly point out, even if it were true that biological altruism cannot evolve, there might still be psychological altruism. Even if there were only selection processes at the level of genes, ruling out the possibility of biological altruism, that is, people might still be willing to sacrifice some of their belongings out of a concern with the wellbeing of others. For a more elaborate discussion of Sober and Wilson’s notion of group selection, see Vromen (2002). For normative purposes this information may be very relevant, of course. In general, if we think we know in what conditions our performance would be improved, we could try to change conditions in that direction. This line of thought is perfectly compatible with EP (see Buss 1999: 18–19). Simon himself once wrote that ‘I would send these two publications to another galaxy’ (Simon 1982: 204). This benchmark of perfect rationality seems similar to what Simon (1978) calls substantive rationality: it focuses on how well the behaviour displayed meets some optimality or efficiency standards. By contrast, EP and the ABC group seem to invoke a notion of procedural rationality. Here the focus is on the process (and not on the outcomes) of decision-making and problemsolving. This need not be misleading if ‘cognitive’ is taken to include possible affective components (as often is done in cognitive science). For economists (and many others as well) cognitive and affective components are assumed to exclude each other. It should be noted here, however, that ‘preferences’ sometimes are interpreted as mere behavioural dispositions (see Sugden 1991). On this ‘minimalist’ interpretation preferences seem to be quite similar to what EPists call modules (or mechanisms, or epigenetic rules). The economists discussed in this chapter do not seem to subscribe to this minimal interpretation, however. It would be more appropriate to talk of sets of algorithms, as a psychological mechanism typically involves several algorithms. To avoid cumbersome phrases, however, I will simply talk of algorithms. It is tempting to add the mind–body dichotomy here. But this may be a bit premature, as EP does not have terribly much to say about the ways in which the mind’s mechanisms are instantiated in brain processes. For a more elaborate related discussion about the vicissitudes of crossdisciplinary learning, see Vromen (2003).

References Aumann, R. (1997) ‘On the State of the Art in Game Theory: An Interview with Robert Aumann’ (taken by Eric van Damme), in W. Albers, W. Güth, P. Hammerstein, B. Moldovonu and E. van Damme (eds), Understanding Strategic Interaction: Essays in Honor of Reinhard Selten, Berlin: Springer Verlag, 8–34. Axelrod, R. (1984) The Evolution of Co-operation, London: Penguin Books. Ben-Ner, A. and Putterman, L. (1998) ‘Values and Institutions in Economic Analysis’, in A. Ben-Ner and L. Putterman, Economics, Values and Organization, Cambridge: Cambridge University Press. —— (2000) ‘On Some Implications of Evolutionary Psychology for the Study of

78 Jack J. Vromen Preferences and Institutions’, Journal of Economic Behavior & Organization, 43: 91–99. Bolton, G.E. (2000) ‘Motivation and the Games People Play’, in Leonard D. Katz (ed.), Evolutionary Origins of Morality, Thorverton: Imprint Academics, 285–291. Bolton, G.E. and Ockenfels, A. (2000) ‘ERC: A Theory of Equity, Reciprocity, and Competition’, The American Economic Review, 90(1): 166–193. Buller, D.J. (2000) ‘A Guided Tour of Evolutionary Psychology’, in Marco Nami and Massimo Marraffa (eds), A Field Guide to the Philosophy of Mind, Department of Philosophy, University of Rome (http://www.uniroma3.it/kant/field/) Buss, D.M. (1995) ‘Evolutionary Psychology: A New Paradigm for Psychological Science’ Psychological Inquiry, 6: 1–30. —— (1999) Evolutionary Psychology: The New Science of the Mind, Boston: Allyn & Bacon. Caporael, L.R. and Brewer, M.B. (eds) (1991) Issues in Evolutionary Psychology. (Special Issue of the Journal of Social Issues, 47(3).) Cosmides, L. (1989) ‘The Logic of Social Exchange: Has Natural Selection Shaped How Humans Reason?’, Cognition, 31: 187–276. Cosmides, L. and J. Tooby (1987) ‘From Evolution to Behavior: Evolutionary Psychology as the Missing Link’, in J. Dupré (ed.), The Latest on the Best, Cambridge, Mass.: The MIT Press, 277–306. —— (1992) ‘Cognitive Adaptations for Social Exchange’, in J.H. Barkow, L. Cosmides and J. Tooby (eds), The Adapted Mind, Oxford: Oxford University Press, 163–228. —— (1994) ‘Better than Rational: Evolutionary Psychology and the Invisible Hand’, American Economic Review, 84: 327–332. —— (1996) ‘Are Humans Good Intuitive Statisticians After All? Rethinking Some Conclusions from the Literature on Judgment Under Uncertainty’, Cognition, 58: 1–73. Dawkins, R. (1976) The Selfish Gene, Oxford: Oxford University Press. Downes, S.M. (2001) ‘Some Recent Developments in Evolutionary Approaches to the Study of Human Cognition and Behavior’, Biology and Philosophy, 16(5): 575–595. Dupré, J. (2001) Human Nature and the Limits of Science, Oxford: Clarendon Press. Elster, J. (1979) Ulysses and the Sirens: Studies in Rationality and Irrationality, Cambridge: Cambridge University Press. Fehr, E. and Schmidt, K. (1999) ‘A Theory of Fairness, Competition, and Cooperation’, Quarterly Journal of Economics, 114: 817–868. Fehr, E. and Gächter S. (2002) ‘Altruistic Punishment in Humans’, Nature, 415: 137–140. Field, A.J. (2001) Altruistically Inclined? The Behavioral Sciences, Evolutionary Theory, and the Origins of Reciprocity, Ann Arbor: The University of Michigan Press. Fodor, J. (2000) The Mind Doesn’t Work That Way, Cambridge, Mass.: The MIT Press. Frank, R.H. (1988) Passions within Reason, New York: W.W. Norton & Company. —— (1999) Luxury Fever: Why Money Fails to Satisfy in an Era of Excess, New York: Free Press.

Cognitive science meets evolutionary theory 79 Frey, B.S. (1997) Not Just for the Money: An Economic Theory of Personal Motivation, Cheltenham: Edward Elgar. Friedman, D. (2001) ‘Economics and Evolutionary Psychology’, Indret, Working paper no. 48: http://www.indret.com/eng/artdet.php?ed53&Idioma eng&IdArticulo392) Gigerenzer, G. (1996) ‘On Narrow Norms and Vague Heuristics: A Reply to Kahneman and Tversky’, Psychological Review, 103: 592–596. Gigerenzer, G., Todd, P.M. and the ABC Research Group (1999) Simple Heuristics That Make Us Smart, New York/Oxford: Oxford University Press. Gigerenzer, G. and Selten, R. (eds) (2001) Models of Bounded Rationality, Cambridge, Mass.: The MIT Press. Gintis, H. (2000) Game Theory Evolving, Princeton, N.J.: Princeton University Press. Gould, S.J. (1991), ‘Exaptation: A Crucial Tool for an Evolutionary Psychology’, in Caporael and Brewer (1991): 43–65. —— (1997) ‘Darwinian Fundamentalism’, The New York Review of Books, 44: 34–37. Griffiths, P.E. (2002) ‘Evo-Devo Meets the Mind: Towards a Developmental Evolutionary Psychology’, in R. Brandon and R. Sansom (eds), Integrating Evolution and Development, Cambridge: Cambridge University Press. Güth, W. (1995) ‘An Evolutionary Approach to Explaining Cooperative Behavior by Reciprocal Incentives’, International Journal of Game Theory, 24(4): 323–344. Güth, W. and Kliemt, H. (1994) ‘Competition or Cooperation: On the Evolutionary Economics of Trust, Exploitation and Moral Attitudes’, Metroeconomica, 45(2): 155–187. —— (1998) ‘The Indirect Evolutionary Approach: Bridging the Gap Between Rationality and Adaptation’, Rationality and Society, 10(3): 377–399. Güth, W. and Ockenfels, A. (2000) ‘Evolutionary Norm Enforcement’, Journal of Institutional and Theoretical Economics, 156(2): 335–347. Güth, W. and Yaari, M.E. (1991) ‘Explaining Reciprocal Behavior in Simple Strategic Games: An Evolutionary Approach’, in U. Witt (ed.), Explaining Process and Change: Approaches to Evolutionary Economics, Ann Arbor: University of Michigan Press, 23–34. Hodgson, G. (1998a) ‘The Approach of Institutional Economics’, Journal of Economic Literature, 36: 166–192. —— (1998b) ‘Thorstein Veblen’s Evolutionary Economics’, Cambridge Journal of Economics, 22: 415–431. Hoffman, E., McCabe, K.A. and Smith, V.L. (1998) ‘Behavioral Foundations of Reciprocity: Experimental Economics and Evolutionary Psychology’, Economic Inquiry, 36: 335–352. Huck, S. (1997) ‘Institutions and Preferences: An Evolutionary Perspective’, Journal of Institutional and Theoretical Economics, 153(4): 771–779. LeDoux, J. (1996) The Emotional Brain: The Mysterious Underpinnings of Emotional Life, New York: Simon & Schuster. Kennair, L. and Ottesen, E. (2002) ‘Evolutionary Psychology: An Emerging Integrative Perspective Within the Science and Practice of Psychology’, Human Nature Review, 2: 17–61. Kuran, T. (1995) Private Truths, Public Lies, Cambridge, Mass.: Harvard University Press.

80 Jack J. Vromen Lloyd, E. (1999) ‘Evolutionary Psychology: The Burdens of Proof’, Biology and Philosophy, 14: 211–233. Marr, D. (1982) Vision: A Computational Investigation into the Human Representation and Processing of Visual Information, San Francisco: Freeman. Robson, A.J. (2001) ‘The Biological Basis of Economic Behavior’, Journal of Economic Literature, 39: 11–33. Samuels, R. (1998) ‘Evolutionary Psychology and the Massive Modularity Hypothesis’, British Journal for the Philosophy of Science, 49: 575–602. Schelling, T. (1984) Choice and Consequence, Cambridge, Mass.: Harvard University Press. Simon, H.A. (1955) ‘A Behavioral Model of Rational Choice’, Quarterly Journal of Economics, 69: 99–118. —— (1956) ‘Rational Choice and the Structure of the Environment’, Psychological Review, 63(2): 129–138. —— (1978) ‘Rationality as a Process and as Product of Thought’, American Economic Review, 68: 1–16. —— (1982) Models of Bounded Rationality: Vol. 2, Behavioral Economics and Business Organization, Cambridge, Mass.: The MIT Press. Sober, E. and Wilson, D.S. (1998) Unto Others: The Evolution and Psychology of Unselfish Behavior, Cambridge, Mass.: Harvard University Press. Sperber, D. (1996) Explaining Culture: A Naturalistic Approach, Oxford: Basil Blackwell. Stotz, K.C. and Griffiths, P.E. (2001) ‘Dancing in the Dark: Evolutionary Psychology and the Argument from Design’, in S. Schwer and M. Rauscher (eds), Evolutionary Psychology: Alternative Approaches, Dordrecht: Kluwer Academic Publishers. Sugden, R. (1991) ‘Rational Choice: A Survey of Contributions from Economics and Philosophy’, Economic Journal, 101: 751–785. Symons, D. (1979) The Evolution of Human Sexuality, New York and Oxford: Oxford University Press. Todd, P.M. and Gigerenzer, G. (2000) ‘Précis of Simple Heuristics that Make us Smart’, Behavioral and Brain Sciences, 23: 727–780. Tomasello, M. (1999) The Cultural Origins of Human Cognition, Cambridge, Mass.: Harvard University Press. Tooby, J. and Cosmides, L. (1995) Mapping the Evolved Functional Organization of the Mind and Brain, in M.S. Gazzaniga (ed.), The Cognitive Neurosciences, Cambridge, Mass.: The MIT Press, 1185–1197. Twomey, P. (1998) ‘Reviving Veblenian Economic Psychology’, Cambridge Journal of Economics, 22: 433–448. Vromen, J. (1995) Economic Evolution: An Enquiry into the Foundations of New Institutional Economics, London: Routledge. —— (1999) ‘Evolutionary Psychology and Economic Theory’, Paper presented at the annual ISNIE Conference, Washington, D.C., 17–19 September. —— (2001) ‘The Human Agent in Evolutionary Economics’, in J. Laurent and J. Nightingale (eds), Darwinism and Evolutionary Economics, Cheltenham: Edward Elgar. —— (2002) ‘Stone Age Minds and Group Selection – What Difference Do They Make?’, Constitutional Political Economy, 13(2): 173–195. —— (2003) ‘What Can Be Learnt from “Serious Biology and Psychology”?’ (To

Cognitive science meets evolutionary theory 81 appear in Proceedings of Annual EAEPE Conference, 2001 in Berlin, edited by Klaus Nielsen and Stavros Iaonnides.) Williamson, O.E. (1998) ‘Human Actors and Economic Organization’, Paper presented at ISNIE Conference, Paris. Wilson, E.O. (1998) Consilience: The Unity of Knowledge, New York: Alfred A. Knopf. Zywicki, T. (2000) ‘Evolutionary Biology and the Social Sciences’, Human Studies Review, 13(1): http://libertyguide.org/libertyguide/hsr/hsr.php/36.html

3

Towards a cognitive experimental economics Marco Novarese

Introduction For the last few years, economic literature has shown a new stream of analysis, at the crossroads of the heterodox tradition and cognitive sciences. A new approach is born, called cognitive economics, since attention is focused on the study of individual and organizational learning, seen as a key factor in shaping social phenomena (Rizzello 1999). This new stream of analysis aims at refunding economic methodologies, starting from more realistic hypotheses than those at the basis of the mainstream. Therefore, empirical observation of learning and decision-making is necessary. Experimental economics, as the main way to gather data on these aspects, is essential for the development of such an approach. At the moment, though, there is still a lack of methodological discussion on how to perform experiments. This chapter aims at analysing the relation between experimentalism and the cognitive approach in economics. As will be shown, cognitivism implies and requires methodological and procedural peculiarities – as compared with the mainstream – in the realization of experiments and in the analysis of the relative data. These differences depend, first of all, on the fact that, though it is deeprooted in social sciences, this can be considered a new-born discipline with obvious different necessities from those of an older science. At the moment, in fact, empirical analysis should here be directed mainly towards the gathering of empirical evidence, instead of comparing alternative models to find the best one. The importance of this descriptive interest is also related to the goal of understanding reality, and not just building models with predictive capacity. One of the main questions suggested by the cognitive literature, a question which raises many theoretical and empirical problems, is the heterogeneity of human agents. Ceteris paribus hypotheses cannot be simply posed and accepted, but should be tested and verified. Differences among subjects should, in fact, be studied and understood. But this goal requires new instruments and new kinds of analysis. For example, individuals,

Towards a cognitive experimental economics

83

because of their differences, can react in different ways to the same experimental treatment. A macro analysis could then lead to conclude that a variable is not relevant, because positive and negative effects can compensate each other at a mean level. But the ‘mean agent’ might make no sense. Some new devices and tools are available to carry on the study. New kinds of experiments, psychological tests and interviews have been recently developed and tested. Also the interaction between experimentalism and computational economics and its simulations can be very useful. Simulations with artificial agents is a growing and promising area of research; but the heterogeneity of human agents makes it difficult, at the moment, to perform such models without looking at the real world. The risk would be, again, that of a science disconnected from reality. The following discussion is necessarily of a preliminary nature. At the moment there are many questions asked and few definite answers, but some early results are encouraging. This chapter also aims at raising a necessary methodological discussion on the general experimental methods. Some of the problems which emerge while conducting experiments based on a cognitive approach can, in fact, be relevant in more general terms. This work is organized in the following way. The first section proposes some methodological and historical notes on experimental economics and singles out a specific newborn cognitive stream of studies. Some peculiarities seem to characterize papers that can be related to this approach. Then a series of methodological and procedural problems and results are discussed through the presentation of two experiments conducted at the Centre for Cognitive Economics in Alessandria. The first one is designed to study individual learning. It helps us demonstrate that a detailed analysis of each player’s individual data is necessary and useful. The second one is directed to investigate team learning and to show two possible tools: psychological tests and interviews.1 The penultimate section proposes some considerations on the relationship between simulations with artificial agents and experiments. Conclusions follow.

Experimental economics and cognitive economics: some historical and methodological notes A historical perspective Usually three main areas of research are supposed to characterize the development of experimental economics (see, for example, Roth 1993): • • •

experiments on markets; experiments on human reasoning and decision-making (mainly focused on von Neumann and Morgenstern’s theory of expected utility); experiments on the theory of games.

84 Marco Novarese Novarese and Rizzello (1999) individuate a fourth, very recent and – at the moment – less developed, stream: the one focused on individual and organizational learning and, more generally on cognitive economics. The main papers concerning this are by Cohen and Backdayan (1991), Egidi and Narduzzo (1997) and Novarese and Rizzello (2003). This classification is obviously arbitrary, as is common in this type of case. Consider two examples, showing how a different classification could be used: 1

2

The theory of games has been used to model market mechanisms or individual decision-making and learning (i.e. the issues that define the other areas). Cognitive economics is strictly related to important experiments classified in the stream on decision-making (see Rizzello 1999), such as Allais (1953), or Kahneman and Tversky (1979, 1992).

Yet the proposed classification is useful to evidence origins and specificity of the different disciplinary areas that gave birth and favoured the development of experimental economics, influencing its methodological evolution. In this sense, for example, the theory of games had a fundamental role in stimulating the realization of experiments and the debate on them during the 1950s, thanks to the Rand Association (see Innocenti 1995). Similarly, then, it seems useful to separate the fourth stream from the others, because of its novelty, in terms of methodological and procedural aims and problems, and also because of the hypothesis under investigation. A methodological perspective Even if sometimes the mainstream has considered it as a minor discipline, experimental economics has often been, directly or indirectly, put at the service of the neoclassical approach, that has also affected its methodological development. Different examples could be proposed. Consider the ones presented here; others will follow. John Hey (1991), one of the most important scholars in this field, states that experimentalism allows the testing of economic models under ceteris paribus hypotheses. Then economics can test its models just as sciences such as physics and chemistry do. Hey and Darnanoni (1988) show that, even if subjects use rules of thumb to take decisions in an experiment on consumer behaviour, they make the optimal choice ‘as if’ they were able to maximize. The theoretical model under analysis seems perfectly able to survive the test of the real world; experiments are used to test this fact.2 In introducing one of the first papers of the new approach, Egidi and Narduzzo (1997) underline a different point: experiments make it possible

Towards a cognitive experimental economics

85

to gather empirical evidence (for example, on learning processes) on aspects otherwise almost impossible to observe in the field. One of the main difficulties is the tacitness and opacity of mental mechanisms for the individual. For this reason, there are problems in the use of standard psychological experimental tools (like protocol analysis, for example), and different tools are required. Since the cognitive approach aims to be close to reality, experimental contests are here usually more complex. As Hey (1991) suggests, in fact, experimental economics usually deals with simple situations because that is the requirement of the theories. Take the studies of Allais (1953) and Kahneman and Tversky (1979, 1992). They are interested in criticizing the theory of expected utility. This model represents the main point of reference in their experimental setting, but it also becomes the cornerstone on which the performance of their alternative model (the prospect theory) is built and compared. Individuals are called upon to face different kinds of (implicit or explicit) lotteries and similar situations, of which they are generally perfectly informed on all relevant probabilities and on the numerical data necessary to choose. In the real world, these situations – that according to Knight (1921) are characterized by risk, as opposed to uncertainty – are still very uncommon. Novarese and Rizzello’s or Egidi and Cohen’s experiments are very different, as individuals build their knowledge in a context characterized by uncertainty and, even if well-defined hypotheses are tested, there is no model to which the laboratory framework is designed. Other differences inevitably follow. In what I call ‘traditional experiments’, attention is usually focused on the possible differences among treatment groups. In the other stream, even if there are treatments, there is also a strong descriptive stress. As subjects face complex tasks (and not just simple hypothetical market or game sets), and as the experimenter aims at building models of the real learning processes and decision-making (and not just to compare different theories), players’ behaviour is seen as interesting in itself and is also analysed beyond treatment groups. These practices are coherent with the recalled idea of experimental economics as a means to gather data on aspects otherwise difficult to study. The researcher is interested in understanding why subjects behave in the way they do, so attention is also focused on individual data and not just on macro phenomena. Probably, these differences also reflect the different stages in the development of the two approaches and the resulting differences in their needs: on the one hand there is the mainstream, with long years of development, on the other hand there is a newborn discipline. The following sections show some examples of analyses, results and problems related to the new ‘cognitive experimental economics’. It is useful to anticipate the main methodological problems faced and their major effects. These problems are suggested both by the empirical

86 Marco Novarese practice and by the cognitive science itself. While they are probably more important in relation to the new approach, they have also a more general applicability. The laboratory, even with all precautions taken, can hardly be a neutral place in which individuals forget the lives they usually lead (as sometimes some experimentalists seem to hope). Their lives are not those of neoclassical agents. On the contrary, lives are full of ideas, beliefs, knowledge and information. All of these elements differ in different players, as they are linked to and dependent on their personal attitude, experiences, studies, jobs, and so on. And all of these differences enter the laboratory with the players and guide their behaviour.3 Furthermore, if players were the agents of neoclassical economics, experiments would be useless as results would be identical to those reached in the models! As we will see, many different factors influence players’ behaviour. All these factors cause loss of control and increase variability among players. It should be underlined that people can react in different ways to the same treatments because of problems related to social experimentation, and also because of their personal differences; experiments (luckily) cannot cancel this fact. As we will see in the following sections, aggregated data can sometimes be inevitably misleading, as they picture a mean player that is different from each real player. To understand how people behave, it then becomes necessary to study each individual case. The following methodological remarks on the analysis of the firm by Herbert Simon (1992: 20) seem useful in relation to experiments: If you are trying to understand what firms are and how they operate, you will learn a lot from this kind of very detailed study of the processes of decision . . . Of course, we should not stop with five firms. Biologists have described millions of species of plants and animals in the world, and they think they’ve hardly started the job. Now, I’m not suggesting that we should go out and describe decision making in a million firms; but we might at least get on with the task and see if we can describe the first thousand. That doesn’t immediately solve the aggregation problem, but surely, and in spite of the question of sampling, it is better to form an aggregate from detailed empirical knowledge of a thousand firms, or five, than from direct knowledge of none. But the latter is what we have been doing in economics for too many years.

Towards a cognitive experimental economics

87

Learning, individual behaviour and methodological problems in experiments An experiment on individual learning The first example proposed is an experiment on individual learning. Additional notes and results on this experiment and on its theoretical framework are presented in Novarese and Rizzello (2003), on which this section is partly based. Here the attention is focused mainly on its methodological aspects. In particular, two issues are dealt with: • •

significant differences among individuals can arise even within the same treatment group; an individual-based descriptive analysis can be very useful to understand macro results.

The experiment Participants were given the following instructions: A subject has to take a series of exams. Each of these exams should be evaluated as: very good, good, middle, bad, very bad. If the subject scores ‘very good’ or ‘good’, he or she will pass the exam. If he/she scores ‘bad’ or ‘very bad’ he/she will not pass the exam. If he/she scores ‘middle’ he/she will have to repeat the exam. Each exam is evaluated with a new system of scoring; it receives: • • •

a colour (black, blue, white or yellow); a shape (heart, circle, square or rectangle); a dimension (big or small).

Then, for example, an exam could be evaluated as ‘blue–heart–small’. The connection between shapes–colours–dimensions and the final result of the exam is not known. The game worked in this way (through specific software): • • • •

the subject was presented with the first combination of shape–colour– dimension; he/she had to choose one of the possible scores; he/she was informed whether the answer is right or not and which is the right solution; he/she was presented with another combination;

and so on for 231 rounds.

88 Marco Novarese Between the combinations of information and the right answer there is a logical relation, stable for all rounds, described in Table 3.1.4 At the beginning of the session players did not know the relationship between information and results, so they have to find it. As the game is based on a relatively small number of sequences of information (there are 4 4 2  32 different combinations of shapes– colours–dimensions, and the same combination appears many times during the game), subjects might try to remember them. In this case, the results of the experiment would scarcely be interesting – more artificial and conditioned by the fact of being in a laboratory. Therefore subjects were not explicitly told that there were fixed and repeated sequences (they were informed, though, of the existence of a logical relation between information and results). Remembering 32 sequences is, besides, not easy. This experiment presents a few similarities with those belonging to the so-called ‘diagnostic task’ series. Kelley and Friedman (1999), for example, asked the participants in their experiment to forecast the price of a good. The price was determined according to a linear stochastic process with two independent variables. Players know the model, but ignore its parameters. In each of the 480 rounds of the experiment they have to forecast the price, after seeing the changing values of the independent variables. After each round they get information on the real value of the price. Kelley and Friedman’s experiment is built to test the ‘least square’ models of learning (see, for example, Marcet and Sargent 1989), where economics agents are expected to be able to estimate the parameters of known models. According to their goal, then, Kelley and Friedman test the accuracy of their players’ implicit estimates. In Novarese and Rizzello (2003) there is a quite different idea of learning, seen – as in the psychological literature – as the human capacity to modify behaviour in a more or less permanent way, whenever new experience is acquired. Table 3.1 Logical relationships between combinations and results Score

Colour

Shape

Dimension

Very good

Bad

Bright (white and yellow) Dark (black and blue)

Very bad

Dark (black and blue)

Not angular shapes (circle and heart) Not angular shapes (circle and heart) Not angular shapes (circle and heart) Angular shapes (square and rectangle) Angular shapes (square and rectangle) Angular shapes (square and rectangle)

Big

Middle

Bright (white and yellow) Bright (white and yellow) Dark (black and blue)

Good

Small Small or big

Small Big

Towards a cognitive experimental economics

89

Then, as a first point, experimental subjects do not know the model used in the experiment (as usually happens in the real world), but they have to create their own knowledge. Moreover, subjects are not expected to understand the real relationships in Table 3.1, as, in our view, there is learning even if they develop a wrong system of rules. Even a wrong model of the world, in fact, allows them to face a given environment. Data are then analysed in a different manner. There is no model to test and there is no analysis of the convergence on true parameters. The aim is mainly that of collecting empirical evidence and looking for eventual regularities among subjects, at the most guided by some theoretical hypothesis. Results The experiments were realized at the Centre for Cognitive Economics in Alessandria in October 2000. Sixty-four subjects (students in law, with no economics background) participated.5 Three main aspects can be detected among subjects: 1 2 3

the tendency to confirm wrong answers to given combinations of information; the tendency to change right answers, even after many right choices; the tendency to develop (right and also wrong) rules.

These features characterize all subjects, but their specific content differs. In other words, not all subjects confirm the same wrong choice or change the same right one. Table 3.2 helps us to understand the first two points. As said, a single combination of shape–colour–dimension appears many times during the game (but never in two rounds running). By taking into account all the answers a subject gives to a particular combination, we can count the number of wrong answers and the number of times wrong choices are confirmed. For example, the sequence yellow–square–small (bold type in Table 3.2 An example of confirmation of wrong answers and change of right choice Round

Colour

Shape

Dimension

Correct result

Subject’s result

9 16 33 43 50 60 67

Yellow Yellow Yellow Yellow Yellow Yellow Yellow

Rectangle Square Square Rectangle Square Rectangle Square

Big Small Small Big Small Big Small

Middle Middle Middle Middle Middle Middle Middle

Middle Good Good Very good Good Very good Good

90 Marco Novarese Table 3.3 Distribution of the mean ratio of wrong answers confirmed by players Percentage

No. of players

0–25 25–50 50–75 75–100

13 38 13 0 64

Total

Table 3.2) appears four times in the part of the game shown. The participant in the experiment exemplified here confirms the same wrong answer after the three errors in the table. We can then say that he confirms three errors after three wrong choices (100 per cent of confirmation). It is possible to calculate a similar rate for all the combinations faced by a player and to determine a mean individual value. Table 3.3 shows the distribution of this rate among players6 and the relevance of the phenomenon under examination. Table 3.2 also shows the sequence yellow–rectangle–big. At its first appearance, player 1 gives a right answer (round 9) but in the following one he makes a wrong choice. Then we can say that (at round 43) he does not confirm a right solution. Many other players show a similar behaviour in many cases and for different combinations. There are players who give a right answer to the same sequence for four or five (or even more) rounds, and then change it until the end of the game. Table 3.4 shows, for the whole of the players, the number of right answers ‘forgotten’ in at least two following appearances. For example, there are thirteen cases in which some players give the right answer to a given sequence in four subsequent rounds, but later, in the following two (or more) appearances, they make a wrong choice. It is useful, now, to concentrate on the last part of the game, and see if players develop regularities of behaviour and an understanding of the rationale of the game.

Table 3.4 Number of right answers ‘forgotten’ in at least two following appearances related to the number of right answers in following appearances 1 2 3 4 5 6 7 8

332 85 27 13 7 5 0 2

Towards a cognitive experimental economics

91

As a first step it is necessary to give a definition of the rules specific to this experiment. Take into account the third part of the game (the last 77 rounds), where many sequences appear three or four times. For each sequence we can count how many times a subject gives the same answer: • •

‘rule 75’ means that the subject gives the same answer 75 per cent of the times; ‘rule 100’ means that the subject gives the same answer 100 per cent of the times.

The idea is that if a player always, or almost always, gives the same answer when faced with the same sequence, he has probably developed a kind of routine.7 There is wide heterogeneity among players, but most of them develop very routinized behaviour. At a mean level, 70 per cent of the possible ‘rule 100’ are developed. Some players develop all possible rules. Does this mean that most of the subjects have understood the game? No, because players tend also to develop wrong rules (15 or 16 per cent of the total), as shown in Table 3.5. It is interesting to see that the mean time taken to answer to ‘rule 100’ (wrong or right) is lower than the overall mean time. The mean time for a reply to a rule is, in fact, about six seconds, while the overall mean time is about ten seconds. A shorter reaction time is one of the features that denote a routine (Cohen et al. 1996). Understanding the empirical findings Is it possible to understand why players behave in this apparently strange way? Why do they seem to act so irrationally? Are the three tendencies related? To try to give an answer to such questions it is necessary to look at each player. A generalized analysis is, in fact, incomplete, as the development of each experiment is very different from the others. Instead, in many cases, an individual analysis seems to allow a full comprehension of the cognitive processes undertaken by the player.

Table 3.5 Percentage of rules (total and wrong) developed by experimental players, mean values*

Total percentage of rules developed Percentage of wrong rules developed

Rule 75

Rule 100

75 16

70 15

Note *Values are computed only for the sequences that appear at least four times during the whole game.

92 Marco Novarese Take into account the player in Table 3.2. As seen, he confirms several times the wrong answer ‘good’ to the sequence yellow–square–small. He does not confirm the right answer to the sequence yellow–rectangle–big. Table 3.6 helps us to understand the link between these facts. It shows the answers given by this individual to some selected sequences in the first 77 rounds (period I) and in the last 77 rounds (period III). For example, in the first period, the subject answers in all cases (100 per cent) ‘very good’ to the sequence yellow–circle–big. This sequence appears three times in the period (as we can see in the last column, where frequency is equal to three). The shaded cell indicates the right answer. The answer ‘good’ (confirmed even if wrong) to the sequence yellow–square–small is coherent with the routines developed (from the very beginning) for the sequences yellow–circle (or heart)–small. On the contrary, the right answer (not confirmed) ‘middle’ for the sequence yellow– rectangle–big is not coherent with the system of rules emerging in the last part of the game, when ‘square’ and ‘rectangle’ are compared to ‘circle’ and ‘heart’. The system of rules developed (also the part which is not shown here) is then wrong, but has an evident and intelligible internal coherence. Table 3.6 also shows answers to the sequence where the colour black appears. There is an evident analogy with sequences characterized by yellow. The sequences with blue (not shown) are similar to the black ones, and the white ones are analogous to those with yellow.8 This is just an example, but a similar system of rules emerges in many cases, and it is probably related to the order of appearance of the information during the experiment. In fact, at the beginning of the game there are many cases of bright unangular shapes and of dark angular shapes (i.e. the ones that receive a higher number of right answers). In this perspective, then, this result is coherent with the hypothesis of path-dependence in individual learning, stated at a theoretical level by Rizzello (1999) and by Egidi and Narduzzo (1997) in experiments on team organization.9 Apart from those who understand the right logical system, there are players who develop a set of rules that is partially different from the one evident in Table 3.6 (in other words, they build a different knowledge of this experimental environment), though they start from the same configuration of the game and on the basis of similar cognitive devices. Consider in Table 3.7 the final part of the game of another player. There is more variance in the answers, but there are also some clear differences from Table 3.6. Black and blue (the latter not shown in table) are generally associated with the answer ‘bad’ (the last choices shown in the table in bold enforce the idea that this is the path of convergence), while the sequences with a yellow are almost always understood. Both systems of rules in Tables 3.6 and 3.7 are not optimal, as in some cases they lead to wrong choices. Yet both of them have an important feature: they are easy to remember and apply (and this is important, as people have bounded rationality).

III

Yellow

I

Black

Yellow

Black

Colour

Period

Rectangle

Square

Heart

Circle

Rectangle

Square

Heart

Circle

Rectangle

Square

Heart

Circle

Rectangle

Square

Heart

Circle

Shape Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small

Dim.

67%

100%

100%

100%

50% 67% 75%

Very bad

75% 33% 100%

100%

100%

40%

50% 33%

Bad

Table 3.6 A more detailed analysis of the behaviour of the player analysed in Table 3.2

25%

25% 20%

50%

Middle

100%

100%

100%

100%

40%

100%

80%

100%

Good

100%

100%

100%

100%

50%

100% 20%

100%

Very good 3 5 2 5 0 4 4 0 0 0 0 0 4 3 4 5 3 4 1 3 1 4 3 1 1 2 2 1 3 4 3 4

Freq.

Yellow

III

Rectangle

Square

Heart

Circle

Rectangle

Square

Heart

Circle

Shape Big Small Big Small Big Small Big Small Big Small Big Small Big Small Big Small

Dim.

Note *Bold characters show the last answer given to each sequence.

Black

Colour

Period

50%

Very bad

100% 50% 100% 67% 25% 67% 100%

Bad

Table 3.7 The system of rules developed by a player different from those in Table 3.6*

33% 25% 33%

50%

100% 33% 100% 100%

Middle

67%

100% 100%

100%

Good

100%

100%

Very good

3 4 1 3 1 4 3 1 1 2 2 1 3 4 3 4

Freq.

Towards a cognitive experimental economics

95

An experiment on team learning This section describes the results of an experiment on team learning (see Novarese 2003 for more details and results). The experiment under examination is called ‘sum 10’, and aims at analysing if and how teams reach coordination in a situation of partially opposite incentives. ‘Sum 10’ presents elements from games on both organizational learning and cooperation (trying to go beyond the simple traditional prisoner’s dilemma). Two aspects emerge: • •

a strong effect of individual psychological traits (as measured by personality tests) on the behaviour in the experiment; the usefulness of the interview realized at the end of the experiment to understand players’ behaviour, motivations, aims, etc.

Both this points question, again, the ceteris paribus assumption. The experiment Teams of three players are, anonymously and at random, built among participants. The game has 36 rounds. Teams are stable throughout the game. Each of the players has a set of numbers. This set remains unchanged in every round and is composed of the values: 0, 1, 3, 4, 10. In every round each player has to declare one of the numbers in his set. The numbers of the three people playing together are then summed. According to the sum, each player receives a payoff, following this rule: • • •

if S(i)  10, I(i)  40  D(i) if S(i)  10, I(i)  30  D(i) if S(i) 10, I(i)  0  D(i)

where S(i)  sum of the team I, of which player i is a member I(i) player is: individual payoff D(i): number declared by player i Then subjects have to co-operate to reach a sum equal or higher than ten, by using the lowest possible number. In treatment group one, players know the total number of rounds; while in treatment two they do not receive such information. The players’ behaviour can be influenced by many interrelated elements. These considerations have general applications to almost all experiments, but for some of them (as this one) they are probably more important: •

The first and obvious element is their actual comprehension of the experiment. Even if the game is not difficult some players show very

96 Marco Novarese

•

strange behaviour; sometimes, the interview after the game indicated that their comprehension or interpretation of the game was flawed (apparently this problem did not emerge in the example shown in the previous section). Subjects can participate in the experiment with many and different motivations. The introduction of remuneration (as prescribed in the most important handbooks, such as that by Hey (1991)) is supposed to stimulate at least a reasonable degree of care, since subjects should have the same motivations which are presumed in the theory to test. Sometimes this reward can be unnecessary. In fact, important results (for example, Allais 1953) have been reached in experiments with no reward. Sometimes students (the main experimental subjects) attend experiments with motivations different from monetary reward (for example, their interest in the subject or just their curiosity). Besides, during the game they can be involved in the accomplishment of that peculiar task, independently of the reward. But these facts, inevitably, lead to a loss of control, at least in relation to the standard view of the laboratory. Besides, experiments are usually similar to games, and people like games. It is not impossible that they react and behave differently from experimenters’ and theories’ hopes (making more points as possible and getting the highest reward). On the contrary they can play just for fun or even to experiment with the behaviour of their partners: •

•

•

The attitude towards other players can be seen in terms of dichotomies: altruism vs. egoism, freeriding vs. giving, rationality vs. emotionality and the desire of reciprocation.10 The attitude towards other members of the team (and, as a consequence, altruism or egoism) can depend on each subject’s motivation in participating in the experiment. A player interested in ‘making money’ will probably behave differently from one who is enjoying the game. Subjects should be able to co-ordinate their choices by communicating their own eventual strategies and/or understanding those of the others. Using the language of game theory, we could say that the experiment has many possible equilibria (like a game or a super-game). Subjects should co-ordinate to reach one of these.

All such (different) behaviours and motivations emerge from the interview conducted before the end of the experiment. Usually experimental papers seem to neglect this problem, though it has important effects on behaviour during the game. We will now analyse this issue in detail.

Towards a cognitive experimental economics

97

Psychological traits and behaviour in the experiment This experiment was conducted in Alessandria, at the Centre for Cognitive Economics, in May 2000; 36 subjects (students in law) participated. The software used was developed by Swiee (Boero 2001). The effects of personality traits on individual economic behaviour have been studied in relation to egoist/altruist behaviour by Boone et al. (1999), for example. They propose a survey and new empirical evidence leading to the conclusion that the personalities of the players matter. In this and other papers, players’ personality is analysed with appropriate psychological tests and related to the behaviour in the experiment (usually the prisoner’s dilemma or similar games). In many cases a relation between attitude to cooperation and psychological traits emerges. A similar analysis is proposed here, connecting the behaviour in the game ‘sum 10’ with individual extroversion, measured with Eysenck and Wilson’s test (1975). This aspect of human personality, as recalled by Boone et al. (1999), could influence behaviour in experiments on cooperation. The most important result is that introverted individuals get a higher mean score in the game. The variable ‘extroversion/introversion’ accounts for about 25 per cent of the variance in the score among individuals. This individual psychological trait, then, has an important role in determining individual (but also organizational) performance in this experiment. Eysenck and Wilson’s test measures extroversion and introversion along seven dimensions, defined by the following dichotomies: • • • • • • •

dynamism/passivity sociability/unsociability boldness/caution impulsiveness/self-control expressiveness/inhibition practicalness/reflectiveness irresponsibility/responsibility

For each of these dichotomies, the test, based on 210 questions, proposes a score ranging from zero to 30. As specified by the authors, individual scores should be compared with those of a reference population. The absolute level of the variables has, in fact, for itself, low significance. According to the position in each of the seven distributions, it is possible to state whether an individual is extroverted or introverted. Introversion is related to the predominance of high values of the first term of each of the dichotomies. Subjects were first assigned to one side of the distribution (above or below the appropriate median) for each of the seven dichotomies. Then subjects with values above the median in the distribution of at least four of the seven variables were classified as ‘extroverted’.

98 Marco Novarese This procedure obviously has its limits (for example, it does not account for possible differences in the variables that define extroversion), but it is coherent with the ideas of Eysenck and Wilson. Besides, it allows us to build two groups composed of individuals with significantly different psychological traits. The group of extroverted individuals was composed of twelve subjects (six from each of the treatment). Different analyses are possible on the two groups. Here attention is focused on differences in the score (i.e. the main measure of performance) realized in the experiment. Tables 3.8. and 3.9. propose the results of the analysis of the variance aimed at comparing mean scores among extroverted and introverted subjects, controlling for the treatment group. The following equation was then estimated: score (i)  intercept
a* extroversion
b* treatment

(1)

where: • •

extroversion  1 if the individual is classified as ‘extroverted’; 0 elsewhere knowing_rounds  1 if treatment group is 1 (subject knows the number of rounds); 0 elsewhere.

The group composed of extrovert individuals has a mean score significantly lower than the introverted one. On the contrary, there are no Table 3.8 Analysis of the variance for the estimation of equation (1)

Model Error Total

df

Sum of squares

Mean square

F value

P value

2 29 31

1,434,118.8 1,313,530 1,747,649

217,059.4 45,294.14

4.792218

0.01

Table 3.9 Parameters estimation for equation (1)

Intercept Extroversion Knowing_rounds R squared  0.25

Coefficient

Standard error

Stat t

P value

1,079.76 29.13 29.125

60.66 77.71 75.24

17.79 3.07 0.39

0.0001 0.0045 0.7015

Towards a cognitive experimental economics

99

significant differences among treatments related to the information on the number of rounds. At first glance this result might seem surprising. Extroversion, in fact, is supposed to be associable to a tendency to co-operate. In this experiment, the extroverted get worse results. It is necessary to observe that ‘sum 10’ is also a game of coordination. The attitudes of some extroverted individuals (such as activity, impulsivity, irresponsibility . . .) can hinder the attachment of this goal. These individuals, in fact, either try to impose their own focal points and strategies, or react in an adverse way to others’ choices. Introverted individuals are, on the contrary, more adaptive and more likely to understand the behaviour of other players (in certain respects, it is comparable to a rational behaviour). Besides, as in this game there is no direct communication, results from extroverts might be less useful. In other kinds of organizational tasks, personal psychological traits can have different effects on the performance. Interviews on the behaviour in the game After the end of the game ‘sum 10’, players were required to describe, in a written interview, their strategies, their perception of the other members of their group (partners or rivals?) and of their behaviour. While similar interviews are probably quite common among experimentalists, there are no standards in their realization and presentation. Normally, then, the results are not proposed at all in the papers.11 Why? The reasons are probably related to the recalled applications and aims of experimental economics in the traditional approach. If scientists are interested in testing theories and their validity, moving from the paper to reality, interviews become useless. Their interest, in fact, is not focused on understanding players’ behaviour but on testing a theory or hypothesis in a more realistic environment. The problem is that differences among players (even if not under analysis) can make the same test of theories more difficult, introducing variance and ‘errors of measure’. Consider the following examples. Figure 3.1 shows one of the team playing the game, reporting individual choices in every round and the relative sum. The team under examination gets a sum equal to ten only in a few rounds of the game. In other rounds they go above or below this number. Yet in the interview all players express their satisfaction in quite similar words: ‘We were able to reach ten in some of the rounds, without using costly strategies (i.e. strategies requiring to declare ten).’ These players’ expectations, then, were not those of always reaching the maximum goal. This is important information that can be useful to explain their performance in relation to other groups which were able to reach ten for almost all the game. These other players, in fact, show generally very different goals, sometimes declaring regret for the few rounds in which they have not been able to get the best result.12

100

Marco Novarese

12 11 10 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

Rounds Player 1

Player 2

Player 3

Sum

Figure 3.1 The game played by one of the teams in the experiment ‘sum 10’

Many problems and risks are related to the use of this kind of interview. In relation to the example: subjects might have changed their level of expectation during the game, and in the interview they might have expressed only their final opinions (it would be necessary13 to test satisfaction also before and during the game). Besides, some players’ statements result in completely useless information. In other cases it is not easy to link interview and real behaviour (because people did not express their ideas clearly). Then, there could be a possible distortion in using only the part of the interview that seems useful. It seems undeniable, however, that sometimes these statements give useful information and allow a better comprehension of players’ behaviour in the game; therefore it would be absurd to ignore them. Another example is shown in Figure 3.2. The main obstacle to the reaching of a stable equilibrium, for this team, seems the behaviour of player 1. He changes his strategies many times and breaks the equilibria. The interview gives interesting insights. His strange behaviour originates from the willingness to play and reach the goal in different ways and not just in one. He is not so interested in getting top scores, but he likes playing, and he tries to avoid repetitiveness. In other cases (not shown here), the reasons behind the difficulties in getting a stable equilibrium rely on the willingness of some players to reciprocate others. These players cannot accept that one of the other members of the team is forced to play a more costly strategy. These examples show the variety of motivations and the usefulness of

Towards a cognitive experimental economics

101

18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

Rounds Player 1

Player 2

Player 3

Sum

Figure 3.2 The game played by another team in the experiment ‘sum 10’

interviews (as, like psychological tests, they can help to better understand players’ behaviour and also to create appropriate ‘control variables’), and so the necessity to develop scientific criteria and appropriate methodologies to perform them (on this, see also Simon 2000).

Experiments and simulations While there are relatively few papers focused on ‘cognitive experimental economics’, recently there has been a growing diffusion of papers on agent-based simulations (Gilbert and Terna 2000). Sometimes the term ‘experiment’ is used also for simulations with artificial agents, as they show many features usually related to scientific empirical analysis. In fact, these models allow observing emergent unpredictable phenomena, starting from a population of evolving and learning agents. Once the modeller has fixed initial conditions and built the environment, he has just to observe without intervention, as in a ‘culture-dish laboratory experiment’ (Tesfatsion 2002: 2). The interrelation between experimental economics and agent-based

102

Marco Novarese

computational economics has increased because of the frequent use of common sets and games (the classical example are the tests of Axelrod 1984, on the prisoner’s dilemma, widely carried out both with human and artificial agents). As simulations are easier to realize than experiments with human agents, and as they also show some advantages (for example, there is no risk that players might misunderstand the rules of the game or have unpredictable motivations), it is not difficult to understand why they are so diffused. Therefore the following question seems reasonable: why not use simulations instead of experiments? The answer is easy: simulations are not the same as experiments, and artificial agents are not human agents. Tesfatsion (2002) analyses some of the possible risks and problems related to simulations. Two of them should be remembered here. The results obtained in simulations can be strongly affected by the representation of learning processes and by the parameters–values setting used. There are inevitably many risks. For example, as ‘many of these learning representations were originally developed with global optimality objectives in mind’ (Tesfatsion 2002: 3), results could be the same as those of neoclassical models. In fact, if agents (real or artificial) were similar to the traditional homo oeconomicus, experiments would be useless, as the same results of the theory would constantly emerge. Therefore, experiments and simulations are not alternative ways of gathering data. On the contrary, they can be usefully integrated. Tesfatsion (2002) suggests that experiments can help to fix parameters and the decision models to be used in simulations. As we have seen in the previous chapter, experiments always show unexpected results in relation to participants’ behaviour. Some of the related heterogeneity can be considered as a noise and it could be useful to eliminate it. In other cases heterogeneity depends on individuals’ peculiarity, and it is an unavoidable part of the world. Only by means of a really empirically founded research program can these features be recognised. If simulations want to start from reality, and not from a hypothetical world (even if it is different from the neoclassical one), cooperation with experimental economics is necessary. Moreover, the analytical power of simulations can be fully expressed if it is used to account for the variety and complexity of the real world (as shown by experiments), which is impossible to manage through usual models. On the other hand, simulations can also help us understand and read experimental data, comparing hypothetical agents and real ones (Tesfatsion 2002). But there is also another way of integrating experimental economics and simulations, exemplified by an application of ‘sum 10’ performed at the Turin University in collaboration with the Centre for Cognitive Economics and described in Novarese (2003). In this case artificial and human agents play together. Artificial agents are used to create specific situations

Towards a cognitive experimental economics

103

and training environments for humans. For example, in the application recalled, a group of subjects played the first part of the game with two artificial ‘egoist’ agents. Another group interacted with artificial ‘altruist’ agents. In this way it is possible to evaluate possible differences in the performance by later carrying out a game in which all players interact together.

Conclusion Cognitive economics imposes and suggests a series of methodological and instrumental novelties in the realization of experiments and in the analysis of data. Some of these novelties mainly interest the works related to this approach; others have a more general appliance. One of the main aspects that emerged here, in different ways, is a definite heterogeneity among individuals. This heterogeneity depends partly on the fact that experiments are artificial situations carried out in a laboratory and the subjects react to this environment in different ways (in terms of motivations and aims, for example). Other differences are, instead, a reflection of the nature of human beings. The laboratory is not a neutral place. Subjects bring with them their personal and idiosyncratic knowledge, ideas, experience . . . The laboratory environment cannot eliminate these differences. All these factors cause a loss of control, as they question the validity of the ceteris paribus condition. Subjects are, in fact, different, even when acting in identical experimental conditions. If we are interested in testing whether a theory resists when we pass from the books to the real world, we can probably ignore these problems; but we have to remember them if we want to explain why people behave in a particular way. Complex environments, instead of simple games, are also necessary in order to advance in the understanding of human behaviour. Inevitably these experiments are sometimes less controlled and also more difficult to manage. Data gathered in this way are necessarily more difficult to read. Yet such experiments give rise to new opportunities, like that of discovering unplanned events (if researchers are open to look at them). As is well known, penicillin was discovered thanks to an accident! Often the apparent ‘noises’ that emerge in unexpected ways are very interesting. The satisficing behaviour manifested by some experimental players could create problems, but it shows, again, the validity of Simon’s model. Apparently strange choices make it very difficult to understand and read data, but are sometimes the result of innovative tendencies induced by the boredom of repetitive behaviour (see Witt 1993, for a theoretical analysis of this mechanism). We have seen that an individual analysis of data, especially if guided by psychological tests and post-experiment interviews, can help understanding these phenomena. But these new methodologies and procedures need

104

Marco Novarese

to be studied, fully understood and developed. More analyses and specific tests will probably be necessary to make these novelties accepted by the scientific community, but their indications are already so strong that they cannot be neglected.

Notes 1 The experiments discussed in this chapter were realized with the contribution of many people. Special thanks to Riccardo Boero, Salvatore Rizzello, Pietro Terna and Cesare Tibaldeschi. 2 As in the laboratory it is possible to satisfy ceteris paribus conditions, experimental economics can also be in contrast with econometrics, the discipline born to test economic models using data gathered from nature’s uncontrolled experiments. Therefore experimental and field data are sometimes seen as contrasting. 3 A simple example is the well-known tendency of economics students to freeride. They bring the mentality learned in their books into the experiments, behaving much more egoistically than other people (see, for example, Franck et al. 1993). Consider also the following statements by Andreoni (1995), which are very representative of the mainstream experimental approach under many respects: ‘laboratory experiments are designed to be neutral and to minimize social effects like kindness. Hence, regular public-goods experiments may already be eliminating a large amount of subjects’ natural tendency to be cooperative . . .’ (p. 900); ‘social and cultural propensities for kindness and generosity must clearly be very strong, and that such motives cannot easily be removed from experimenters simply by providing neutral environments and pledges of anonymity’ (p. 892). 4 The categories of colour and shape are not explicitly used in the presentation of the game to the players. 5 Similar results were obtained in a pilot experiment realized in March 2000 in Alessandria with a smaller group of subjects remunerated with money. The same experiment, but with a different frame, was replicated in July 2002. All main findings have been confirmed and new evidence has emerged. 6 As shown in Novarese and Rizzello (2003), these results can hardly be explained by subjects’ bad memory. 7 As the game is very long and repetitive, it seems reasonable to think that a subject who has developed a rule can make a mistake and give, in some cases, a different answer. Moreover, the rule could be in evolution and not perfectly defined at the beginning of what we define as the third period. That is the reason why we also use a definition like ‘rule 75’. 8 It is evident, now, that the subject under examination is reasoning and not simply memorizing. 9 As some results could be context-specific, it is very important to find coherent results in different experiments. 10 There is an extensive theoretical and experimental literature on this subject. It is not possible to analyse it here. For a review, see Novarese (2002a, 2002b). 11 Andreoni (1995: 898) talks about this kind of interview, but in a methodological paper. 12 In the interview, player 1 in Figure 3.1 explained the reason for the choice at round 17. After having played three or four, he decided ‘to save some points’. Thus this apparent strange choice is not a signal sent to other members of the team, or a way to communicate new strategies, and it is not an error. 13 It would also be possible, but it would increase the management problems of the experiment.

Towards a cognitive experimental economics

105

References Allais, M. (1953) ‘Le comportement de l’homme rationnel devant le risque: critique des postulats et axiomes de l’école américaine’, Econometrica, 21: 503–546. Andreoni, J. (1995) ‘Cooperation in Public-Goods Experiments: Kindness or Confusion?’, The American Economic Review, 85(4): 891–904. Axelrod, R. (1984) The Evolution of Cooperation, New York: Basic Books. Boero, R. (2001) ‘SWIEE a Swarm Web Interface for Experimental Economics’, in F. Luna and A. Perrone (eds), Agent Based Methods in Economics and Finance: Simulations in Swarm, Boston, Dordrecht and London: Kluver Academic Publishers. Boone, C., De Brabander, B. and van Witteloostuijn, A. (1999) ‘The Impact of Personality on Behavior in Five Prisoners’ Dilemma Games’, Journal of Economic Psychology, 20: 343–377. Cohen, M.D. and Backdayan, P. (1991) Organizational Routines are Stored as Procedural Memory: Evidence from a Laboratory Study, University of Michigan, mimeo. Cohen, M.D., Burkhart, R., Dosi, G., Egidi, M., Marengo, L., Warglien, M. and Winter, S. (1996) ‘Routines and Other Recurring Action Patterns of Organizations: Contemporary Research Issues’, Industrial and Corporate Change, 5(3): 653–698. Egidi, M. and Narduzzo, A., (1997) ‘The Emergence of Path-Dependent Behaviours in Cooperative Contexts’, International Journal of Industrial Organization, 5: 677–709. Eysenck, H.Y. and Wilson, G. (1975) Know Your Own Personality, London: Temple Smith. Franck, R.H., Gilovich, T. and Regan, D.T. (1993) ‘Does Studying Economics Inhibit Cooperation?’, Journal of Economic Perspectives, 7: 159–171. Gilbert, N. and Terna, P. (2000) ‘How to Build and Use Agent-Based Models in Social Sciences’, Mind & Society, 1(1): 57–72. Hey, J.D. (1991) Experiments in Economics, Oxford: Blackwell. Hey, J.D. and Dardanoni, V. (1988) ‘A Large Scale Experimental Investigation into Optimal Consumption Under Uncertainty’, Economics Journal, 98: 105–116. Innocenti, A. (1995) ‘Le origini della Experimental Economics (1948–1959): una valutazione critica sull’evoluzione delle metodologie sperimentali’, Working papers, Università di Siena. Kahneman, D. and Tversky, A. (1979) ‘Prospect Theory: An Analysis of Decision Under Risk’, Econometrica, 47: 263–291. —— (1992) ‘Advances in Prospect Theory: Cumulative Representation of Uncertainty’, Journal of Risk and Uncertainty, 5: 297–323. Kelley, H. and Friedman, D. (1999) ‘Learning to Forecast Price’, Working paper, University of California Santa Cruz, Economics Department. Knight, F.H. (1921) Risk, Uncertainty and Profit, London: The London School of Economics and Political Science. Marcet, A. and Sargent, T. (1989) ‘Convergence of Least Squares Learning Mechanisms in Self Referential Linear Stochastic Models’, Journal of Economic Theory, 48: 337–368. Novarese, M. (2002a) ‘Altruismo ed egoismo in economia e nelle altre scienze

106

Marco Novarese

sociali’, Working paper, Dipartimento di Scienze Giuridiche ed Economiche, Università del Piemonte Orientale. —— (2002b) ‘Lealtà, apprendimendo ed innovazione nella teoria eterodossa dell’impresa’, Working paper, Dipartimento di Scienze Giuridiche ed Economiche. —— (2003) ‘Some Experiments on Team Learning’, Working paper, Dipartimento di Scienze Giuridiche ed Economiche. Novarese, M. and Rizzello, S. (1999) ‘Origin and Recent Development of Experimental Economics’, Storia del Pensiero Economico, no. 37: 201–234. —— (2003) ‘A Cognitive Approach to Individual Learning: Some Experimental Results’, Revue d’Economie Politique, forthcoming. Rizzello, S. (1999) The Economics of the Mind, Aldershot: Edward Elgar. Roth, A.E. (1993) ‘On the Early History of Experimental Economics’, Journal of the History of Economic Thought, 15: 184–209. Simon, H.A. (1992) ‘Colloquium with H.A. Simon’, in M. Egidi and M. Marris (eds), Economics, Bounded Rationality and the Cognitive Revolution, Aldershot: Edward Elgar. —— (2000) ‘Bounded Rationality in Social Science: Today and Tomorrow’, Mind & Society, 1(1): 25–40. Tesfatsion, L. (2002) ‘Agent-Based Computational Economics’, ISU Economics Working Paper No. 1, Department of Economics, Iowa State University. Witt, U. (1993) ‘Emergence and Dissemination of Innovations – Some Problems and Principles of Evolutionary Economics’, in R. Day and P. Chen (eds) Nonlinear Dynamics and Evolutionary Economics, Oxford: Oxford University Press.

4

Accounting for social knowledge in economic analysis The relevance of Adam Smith’s framework1 Pier Luigi Porta and Roberto Scazzieri

Introduction: co-ordination and social knowledge The relationship between freedom of choice and co-ordination is a central theme in economic theory. Freedom of choice is inherently associated with the multiplicity of preferences and the set inclusion criterion. Multiplicity of preferences suggests a situation of social variety, such that different individual rankings may exist side by side. Set inclusion implies that any given agent is at least as free in the largest opportunity set as s(he) is in the smaller set. The combination of multiple preferences and set inclusion leads to the economist’s riddle: heterogeneous agents are conceived as capable of achieving an effective pattern of interaction (competitive equilibrium) by co-ordinating their respective freedoms of choice (see Arrow 1994; see also Arrow 1986). A closer look at the above conception of freedom reveals that the economist’s perspective is somehow distinct from the classical analysis of ‘negative’ or ‘positive’ freedom (see Constant 1815; Berlin 1958; Matteucci 1993). On the one hand, negative freedom (that is, freedom from constraints) is seldom associated with the analysis of co-ordination; on the other hand, theories of positive freedom may have considerable difficulty in dealing with the multiplicity of preferences and the associated riddles of social choice. Recent work in economic theory has called attention to the role of shared information and beliefs in making social co-ordination feasible (Schelling 1960; Aumann and Drèze 1974; Hahn 1978, 1987; Bacharach 1993; Greif 1994; Sugden 1995; Rubinstein 2000; Gilboa and Schmeidler 2001; Schmidt 2001). It has also suggested that economic belief structures are complex blends of formal and informal knowledge, and that the analysis of the way in which beliefs influence co-ordination presupposes the decomposition of existing belief structures into their primitive or elementary elements (Scazzieri 1999a: 299). In this way, co-ordination may be associated with the identification of equivalent elements belonging to different belief structures (ibid.).2

108

Pier Luigi Porta and Roberto Scazzieri

Social knowledge, which may be defined in general terms as a cognitive framework by which human beings make sense of each other’s behaviour, has emerged as a distinctive field of investigation in economic analysis. In particular, social knowledge is considered to be at the root of economic performance and its characteristics are often associated with distinct paths of economic dynamics. This kind of thinking has been developing throughout the literature on political economy. Examples are provided by Douglass North’s analysis on institutions and economic performance (North 1990; see also North 1999); on the same line of thought we understand Paul David’s contributions on the role of institutions and in particular his emphasis on path dependence (David 1994). Significantly the literature we have in mind leans toward a distinctive emphasis on social capital as, for example, in Robert Putnam’s analysis of civic traditions (Putnam 1993), in Paolo Sylos Labini’s emphasis upon the combined influence of culture and institutions in determining the characteristics of long-run dynamics (Sylos Labini 2001), and in Partha Dasgupta’s investigation of the influence of social capital in economic development (Dasgupta 2002). Understandably, all this contributes to identifying an interdisciplinary area of thought in which philosophers, historians, sociologists and political theorists work side by side with economists (see Greif 1994; Bortis 1997; Landes 1998). The realisation that social knowledge (as distinguished from specialised economic knowledge) is a critical feature of economic co-ordination is a relatively recent discovery. It derives from a number of different research lines in theoretical and applied economics. Among others, the analysis of co-ordination games suggests that relatively persistent patterns of coordination (such as those leading to social norms and institutions) often presuppose a process of ‘convergent expectations’ leading to ‘focal points’ (Schelling). These expectations may derive from individuals’ ability to ‘communicate by hint and by suggestive behaviour’, as is outlined in the classic work by Thomas Schelling (1960: esp. p. 85). Economic coordination presupposes a structure of beliefs, symbols and communicative codes; that is, a body (tradition) of mutual adjustments and interdependent decisions by which the outcome of social interaction is constrained. However, convergent expectations do not require a fixed set of codes and beliefs: co-ordination is a dynamic process in which mutual adjustment derives from active participation ‘in the creation of traditions’ (Schelling 1960: 106). An important contribution to the economic theory of social knowledge comes from another line of thinking, directly focused on a critical reconstruction of the notion of rationality and rational choice: the research line associated with Herbert Simon’s analysis of bounded rationality. In this case, limited inferential capability is compensated by effective focusing devices, which enable individuals to select relevant information and to elaborate effective cognitive patterns. As argued by Simon, ‘we are all

Social knowledge in economic analysis 109 very limited in how fully we can calculate our actions and in how rational we can be in a complicated world. But institutions provide a stable environment for us that makes at least a modicum of rationality possible . . . [O]ur institutional environment, like our natural environment, surrounds us with a reliable and perceivable pattern of events’ (Simon 1983: 78). Interesting implications of the bounded rationality approach relate to the individual and collective ability to interpret empirical evidence, to assign causes to events on the basis of identifiable patterns, and to elaborate a conception relative to human manipulative ability. Bounded rationality entails a shift from computation to pattern identification. This means that focal points (as ‘centres of gravity’ of individual expectations) can be essential in screening information and inducing an effective interpretation of evidence. The formation of ‘indirect knowledge’ (knowledge not stemming from primary evidence or feeling) appears to be closely associated with analogical inference. In other words, indirect knowledge may be associated with reasoning based upon the partial similarity of objects.3 Bounded rationality is not simply focusing attention to constraints upon human inferential ability. It is also suggestive of a cognitive paradigm in which a thought process characterised by horizontal associations rather than explicit utilisation of rules of inference takes a primary role. In this case, surprise (rather than Humean regularity) can be an important incentive to the growth of knowledge. Surprise presupposes pattern recognition. An event that is unexpected because it cannot be specified in advance is surprising precisely because it does not fit any pre-existing pattern. And surprise may induce a reconfiguration of the space of events, which may eventually lead to a new cognitive pattern and to the explanation of the unexpected occurrence (see also Scazzieri 2001a, 2001b). Linked with the latter there is a wide field of research on economics and psychology, which directly concerns rationality and the process of decisionmaking and choice. Finally, learning processes have received considerable attention in recent literature on technical progress and structural change. In particular, the immaterial side of technical change has been emphasised and human capital has been considered a critical prerequisite of economic growth. Learning has a twofold character, depending upon whether producers’ or consumers’ learning is considered. In the former case, the growth of (general) knowledge has been discovered to be a prerequisite of an increasing (optimal) division of labour (Becker and Murphy 1992). In the latter case, the evolution of consumers’ preferences has been considered an important factor governing the transformation of productive structure and the diffusion of technical change (Pasinetti 1981, 1993). In both cases, effective learning presupposes a suitable cognitive and institutional framework. This framework is characterised by two essential features. First, general knowledge generates production-specific knowledge by means of

110

Pier Luigi Porta and Roberto Scazzieri

division of labour and the associated partition of the economic system into a certain number of ‘lineages’, which include teachers, students and workers (Becker and Murphy 1992: 1154). Second, social knowledge and institutions generate a suitably differentiated (and flexible) range of preferences which enhance the possibility of consumers’ learning and the likelihood that a new demand structure will be conducive to learning and improvements on the production side (Pasinetti 1993: ch. IV). The epistemic twist of economic theory suggests a new perspective upon the relationship between co-ordination and social codes. The reason is that co-ordination appears to be primarily a cognitive process, by which salient features are identified and communication achieved. This entails that likelihood of co-ordination is higher when individuals (or social groups) come to share a common set of communication beliefs. The consideration of social knowledge has a manifold influence in recasting the problem of economic co-ordination. Recent theoretical work has emphasised a number of connected themes. In particular, contributions in economic theory have considered the role of social knowledge in determining salient features, reducing complexity and achieving ‘focused’ interaction (Bacharach 1993; Mehta et al. 1994; Sugden 1995; Rubinstein and Zhou 1999; Sen 1999). Other contributions have examined the influence of epistemic patterns and cognitive skills in making a ‘constructive’ approach to uncertainty possible, and in allowing for treatment of unexpected occurrences (surprises) when prior probabilities are unknown (Suppes 1984, 1996, 1997; Crisma 1988; Arrow and Hahn 1999). Finally, analytical and historical research has led to the investigation of the connection between formal knowledge, tacit knowledge and effective learning, which is often learning by means of ‘exemplars’ (North 1990, 1999; David 1994; Poni and Scazzieri 1994; Bacharach 1998; Bruni and Sugden 1998). A new general theory is gradually emerging, whereby knowledge substitutes pricing as the critical co-ordinating device. This perspective entails a radical shift in the theory of economic rationality. In particular, the view of rationality as primarily a problem-solving ability gives way to rationality as a problem-setting capacity. The new conception of rationality entails a shift within the theory of economic choice. Its salient feature would be learning about a search space (exploration) rather than maximisation from among a set of known alternatives.

Some building blocks: a retrospective view The epistemic twist of economic theory suggests a new approach to the relationship between co-ordination and social knowledge. In particular, attention for the cognitive basis of co-ordination makes the latter highly sensitive to the congruence between conceptual spaces, rather than to the extent of directly shared information. Congruence suggests the relevance of cognitive practices that can be shared within a community of hetero-

Social knowledge in economic analysis 111 geneous agents. In particular, attention for congruence highlights that a cognitive ability that is most critical for co-ordination is the ability to ‘decompose’ diversity into constituent elements, some of which could be common to a large number of individuals (or social groups). It appears that co-ordination often results from a shared attitude to social reflection; that is, from a shared attitude to reflect upon other agents’ motives and beliefs as if they were our own. This attitude is, to a certain extent, similar to the ‘impartial spectator’ discussed by Adam Smith, even if, in this case, the implications for social co-ordination, not for moral judgement, are considered. There is thus an important reason to examine the cognitive basis of social co-ordination in terms of Smith’s theory of social situations and inductive analogy. The Smithian themes that are most directly relevant in this framework are: (1) the analysis of social communication and contracting codes; (2) the investigation of inferential procedures behind the formation of inductive knowledge; (3) the focus upon human labour and learning as the fundamental prerequisites for the growth of social wealth. The intellectual achievement of Adam Smith suggests a fresh outlook in all three areas and highlights the possibility of reducing important features of social communication, knowledge formation and technical learning to a common set of principles concerning the discovery of connections and the treatment of unexpected occurrences (surprises). The retrospective aspect is an essential component of our argument from a methodological standpoint (see Porta 1992); moreover, it provides the distinctive elements of the classical approach to political economy. Smith’s analysis of social interaction places great emphasis upon the combined influence of imagination and sympathy. Both are rooted in a failure of direct knowledge: ‘[a]s we have no immediate experience of what other men feel, we can form no idea of the manner in which they are affected, but by conceiving what we ourselves should feel in the like situation’ (Smith [1759] 1976a: 9). The ability to consider a counterfactual setup is a necessary condition of social judgement: for direct knowledge never extends ‘beyond our own person’ (ibid.) and ‘it is by the imagination only that we can form any conception of what are [another individual’s] sensations’ (ibid.); and imagination is associated with ability to conceive of a possible world different from the actual one, not only in the sense that we may attempt an assessment of another individual’s sentiments but also in the stronger sense that the spectator could imagine another individual’s situation to be his or her own and express on this basis his or her own independent emotions and judgement. In Smith’s account, sympathy presupposes some notion of social distance, which is a necessary condition for the imagination of possible worlds to take place. Imagination, in turn, brings about sympathy by following a variety of routes, which can be arranged according to a ‘hierarchy’ of communication patterns, ranging from ‘semiotic sympathy’, to ‘subsidiary’ and ‘mutual’

112

Pier Luigi Porta and Roberto Scazzieri

sympathy. First, Smith considers the case in which imagination is the immediate consequence of semiotic congruence (often a merely physical association of feelings): ‘[u]pon some occasions sympathy may seem to arise merely from the view of a certain emotion in another person. The passions, upon some occasions, may seem to be transfused from one man to another, instantaneously, and antecedent to any knowledge of what excited them in the person principally concerned. Grief and joy, for example, strongly expressed in the look and gestures of any one, at once affect the spectator with some degree of a like painful or agreeable emotion’ (Smith [1759] 1976a: 11). Second, Smith investigates the ‘higher order’ sympathy that arises when the spectator, as a result of perceived social distance, attains a view of the situation that makes him capable of a feeling of which the individual under consideration seems to be unable: ‘[w]e blush for the impudence and rudeness of another, though he himself appears to have no sense of the impropriety of his own behaviour; because we cannot help feeling with what confusion we ourselves should be covered, had we behaved in so absurd a manner’ (ibid.: 12). Finally, ‘mutual sympathy’ is considered to be the highest stage of social interaction: ‘nothing pleases us more than to observe in other men a fellowfeeling with all the emotions of our own breast; nor are we ever so much shocked as by the appearance of the contrary’ (ibid.: 13). In this case, imagination takes a roundabout turn: individual A understands (by virtue of imagination) the situation of individual B and expresses a fellow-feeling towards her joy or grief; at the same time, individual B recognises the fellow-feeling of A and finds her grief alleviated or her joy enlivened. Reciprocation of feeling becomes an essential element of social interaction, and is itself at the origin of a new ‘layer’ of sentiments: in general, human beings are pleased when ‘able to sympathize’ with the feelings of another human being, and distressed ‘when unable to do so’ (ibid.: 15). Smith’s analysis of imagination suggests that sympathy takes a different route depending upon the specific task that imagination fulfils: semiotic sympathy requires an almost instinctive use of imaginative power; subsidiary sympathy shows the possibility of an ‘inversion’ of feelings (our consideration of the situation of another human being suggests feelings different from her own); mutual sympathy (fellow-feeling) reveals the possibility of a ‘double reciprocation’ of sentiments (the spectator shares a certain feeling by virtue of imagination exerted upon the observed individual’s situation; this latter finds his joy enlivened or his grief alleviated by virtue of imagination exerted upon the spectator’s view of his own situation).

Social knowledge in economic analysis 113

Smithian themes: language, analogy and connecting principles More particularly, it is Smith’s theory of imagination and sympathy that provides a distinctive approach to the issue of the formation of human knowledge. In this connection, we need to consider the distinction between direct knowledge and indirect knowledge. Direct knowledge is the body of information that can be immediately derived from experience without having recourse to inferential procedures. Indirect knowledge is the set of propositions that can be derived from knowledge of other propositions and a judgement concerning the likelihood of certain associations of characters. The formation of social knowledge reflects, in a Smithian perspective, the practice of certain codes of association by which sympathy can be established (see p. 111). In particular, social knowledge differs from other types of knowledge as a result of a different process by which the association of characters can be detected: sympathy presupposes a capability to ‘generate’ a specific view of other people’s situations, and the latter is made possible by imagination. The practice of imagination leads to the representation of ‘internal worlds’ different from the one that is directly experienced, and makes the association of characters to reflect a perceived commonality of possible situations rather than an actual similarity of features. As a result, social knowledge takes shape by virtue of imagination and likelihood, and reflects a special route by which indirect knowledge can be established. In particular, the judgement of another individual’s situation makes it necessary to consider: (1) this individual’s situation as a world that the spectator’s imagination can grasp; (2) the observed individual’s situation as a possible world that is reasonably likely to come about; (3) the sentiments of the observed individual as a cluster of feelings that the spectator could eventually share. In this way, social knowledge presupposes a sophisticated interplay of imagination, likelihood and proximity: likelihood is required for the spectator’s imagination to grasp another individual’s situation; proximity is necessary for imagination to produce semiotic sympathy, ‘approbation’ and, possibly, ‘fellow-feeling’. Indirect knowledge about other people’s feelings and beliefs stems from the ability to ‘reconstruct’ external worlds as internal worlds, and to understand them in terms of ‘approbation’ and commonality of feeling. As a result, the formation of social knowledge presupposes a particular combination of direct experience and induction. In Smith’s view, the ‘first perceptions, as well as all other experiments upon which any general rules are founded, cannot be the object of reason, but of immediate sense and feeling. It is by finding in a vast variety of instances that one tenor of conduct constantly pleases in a certain manner, and that another as constantly displeases the mind, that we form the

114

Pier Luigi Porta and Roberto Scazzieri

general rules of morality’ (Smith [1759] 1976a: 320). On the other hand, rules of inference come to play an essential role in the evolution of ‘standards of judgement’ and social norms: ‘[t]he general maxims of morality are formed, like all other general maxims, from experience and induction. We observe in a great variety of particular cases what pleases or displeases our moral faculties, what these approve or disapprove of, and, by induction from this experience, we establish those general rules’ (ibid.: 319). In this way, the formation of social knowledge is closely associated with the general process of knowledge formation, even if social knowledge entails a peculiar emphasis upon the role of direct knowledge and imagination is presupposed not only in the process by which general (social) rules are identified, but also in the preliminary phase in which the spectator’s feelings are born. In other words, imagination not only determines the characteristics of the ‘connecting principles’ (see also Bagolini 1975: 153) at the root of any pattern of social congruence, but is also a necessary condition for ‘direct’ moral experience.4 Social knowledge presupposes a double route which imagination should follow: (1) moral sentiments are born from the internal reconstruction of an external world, which leads to ‘direct’ moral experience; (2) social norms and moral codes (Smith’s ‘general maxims of morality’) are determined by the introduction of connections that bind together isolated judgements (moral feelings) and turn them into a general rule. In this way, social knowledge and moral criteria appear to derive from a complex interplay of ‘experience’ (for example, ‘the first perceptions of right and wrong’) and inference (the ‘general rules of morality’). Inductive inference can be considered ‘as one of the operations of reason’ (Smith [1759] 1976a: 319). However, according to Smith, ‘it is altogether absurd and unintelligible to suppose that the first perceptions of right and wrong can be derived from reason, even in those particular cases upon the experience of which the general rules are formed’ (ibid.: 320). As a matter of fact, ‘immediate sense and feeling’ complement inductive inference as means by which standards of judgement are derived: ‘nothing can be agreeable or disagreeable for its own sake, which is not rendered such by immediate sense and feeling’ (ibid.). As a result, social knowledge presupposes both moral feelings and inferential ability: the former makes imagination a source of ‘indirect experience’ (concerning other individuals’ feelings and beliefs); the latter makes imagination the source of connecting principles and turns indirect (moral) experience into a general standard of judgement. From a general perspective, Smith’s theory of social knowledge emphasises the ‘active’ role of the spectator in the generation of the moral sense and in the formation of social norms by the discovery of ‘connecting principles’. At the same time, connecting principles are not purely subjective, as general standards derived from induction and the latter emerge from consideration of ‘a great variety of particular cases’ (Smith [1759] 1976a:

Social knowledge in economic analysis 115 319). In this regard, the dynamics of social knowledge follows the pattern of human knowledge in general, and makes social knowledge to depend upon the sentiments of ‘surprise’, ‘wonder’ and ‘admiration’ (see Smith [1795] 1980: 33–34). These sentiments suggest focal points to the development of social knowledge and make the latter to reflect: (1) the cognitive dissonance associated with the observation of ‘extraordinary and uncommon objects’ (wonder); (2) the ‘violent and sudden change’ that the human mind undergoes when an unexpected occurrence takes place (surprise); (3) the mental disposition to look for ‘resemblances’ and constant association of events, and to be pleased with it (admiration). The formation of social knowledge presupposes a sophisticated interplay of ‘experience’ and ‘induction’: (1) ‘immediate sense and feeling’ are necessary conditions for the emergence of custom and of a certain idea of the social order; (2) partial, or ‘imperfect’, resemblance between different moral and social situations (and the ensuing formation of moral codes and rules) is itself rooted in features of salience deriving from individual or collective history. Social knowledge presupposes moral experience (in Smith’s own sense) and the ability to detect partial similarities within a diverse social universe. However, moral experience is different from the experience of natural events and processes, because ‘immediate sense and feeling’ in the moral sphere is itself based upon the working of imagination. Imagination in the moral sphere reflects the general principles sustaining association of ideas: ‘It is evident that the mind takes pleasure in observing the resemblances that are discoverable betwixt different objects. It is by means of such observations that it endeavours to arrange and methodise all its ideas, and to reduce them into proper classes and assortments. Where it can observe but one single quality, that is common to a great variety of otherwise widely different objects, that single circumstance will be sufficient for it to connect them all together, to reduce them to a common class, and to call them by one general name’ (Smith [1795] 1980: 37–38). The association of ideas takes a peculiar twist when social interaction is considered. The reason is that the formation of categories is closely related to the formation of language. As a result, ordinary objects ‘would all soon be distinguished by particular names, as they would have frequent occasion to make their thoughts about these known to one another’ (Smith [ms. 1762–63] 1983a: 9). Under these conditions, human beings ‘would by mutual consent agree on certain signs whereby this might be accomplished’ (ibid.).5 It may be noted that, in Smith’s account, social interaction gives rise to language, while language itself is moulded by social intercourse. Effective communication presupposes common associations between ‘names’ and ‘things’. This entails that effective communication takes place against a cultural background in which common associations of ideas can be detected. This feature of

116

Pier Luigi Porta and Roberto Scazzieri

language has a remarkable consequence for social co-ordination. For coordination presupposes a core of common concepts (beliefs), and the latter is associated with the formation of a common language (or set of congruent languages). In this connection, Smith calls attention to the process of language transformation in a framework of increasing social diversity. In particular, Smith argues that any given language goes through a stage of increasing complexity (both in grammar and syntax) as long as language transformation takes place in a setting characterised by social homogeneity. But the evolution of language takes a different course as soon as a ‘mixture of different nations’ requires an effective means of communication. In this case, the structure of language gets simpler: ‘[w]hen two nations . . . meet, when “one” was at a loss to express himself in the other language he would be led to supply this defect in some easy manner’ (Smith [ms. 1762–63] 1983a: 13). In this way, ‘compounded languages’ (that is, languages deriving from a blend of different cultural and linguistic traditions) become ‘more simple in . . . rudiments and principles, just in proportion as [they grow] more complex in [their] composition’ (Smith [1761] 1983b: 223). Smith’s theory of simple (or ‘uncompounded’) and compounded languages suggests a possible connection between the evolution of language and social co-ordination. The initial stage of language formation is one in which a set of individuals gives origin to its own particular language without borrowing terms or structures from other languages. In this case, new terms are formed ‘either by composition, or derivation from some other word or words’ in their own language. This means that the evolution of language takes place against a background of common terms and mental associations. Language expansion takes place by the introduction of increasingly complex twists of the original names and verbal expressions. In a subsequent stage, ‘mixtures’ of different nations bring about a different pattern of language transformation. The need to achieve effective communication between two (or more) ‘simple’ languages may lead to a new language, which is born from the ‘composition’ of the original languages. In Smith’s view, ‘[e]ach nation, in order to make itself intelligible to those with whom it was under the necessity of conversing, would be obliged to learn the language of the other. The greater part of individuals too, learning the new language, not by art, or by remounting to its rudiments and first principles, by rote, and by what they commonly heard in conversation, would be extremely perplexed by the intricacy of its declensions and conjugations. They would endeavour, therefore, to supply their ignorance of these, by whatever shift the language could afford them’ (Smith [1761] 1983b: 220). Social differentiation (the ‘mixture of different nations’) leads to the introduction of compounded languages generally ‘more simple’ in terms of ‘rudiments and principles’. For example, declensions may be dropped and conjugations may become simpler. The construal of human experience often reflects established ‘habit and custom’: ‘[i]t is well known that

Social knowledge in economic analysis 117 custom deadens the vivacity of both pain and pleasure, abates the grief we should feel for the one, and weakens the joy we should derive from the other’ (Smith [1795] 1980: 37); on the other hand, ‘transports of rage or grief . . . often . . . distract the fortunate and successful’ (ibid.). Connecting principles are at the origin of human knowledge both in the natural and the moral sphere. However, the ‘reflective’ character of moral feelings makes moral experience itself to be a product of imagination. In this way, connecting principles are both a condition and a product of social knowledge: the ability to detect ‘partial resemblances’ is at the root of general moral principles and social codes, but is in turn the product of received features of salience. The above conceptual framework suggests that any given pattern of economic (and social) co-ordination is primarily a relational setting shaped by convergent expectations. But expectations are more (or less) likely to converge depending on the internal configuration of belief structures. As a result, co-ordination emerges as an epistemic riddle, and its likelihood is associated with the extent to which intersubjective communication of knowledge is feasible under given social conditions.

Conceptual spaces, diversity and ‘imperfect resemblance’6 The previous argument suggests that co-ordination reflects convergent beliefs, and that convergence presupposes a common core of primitive beliefs. In particular, co-ordination in the midst of social diversity presupposes the possibility to ‘decompose’ any given set of beliefs into a certain number of primitive elements. The criterion of ‘imperfect resemblance’ (Smith) suggests that partially overlapping belief systems could be conducive to extensive co-ordination across a large number of different individuals or social groups. The reason is the cognitive attitude thus described by Smith: ‘[i]t is evident that the mind takes pleasure in observing the resemblances that are discoverable betwixt different objects. It is by means of such observations that it endeavours to arrange and methodise all its ideas, and to reduce them into proper classes and assortments’ (Smith [1795] 1980: 37–38). The resemblance criterion works differently depending on which objects are compared. If the distance between objects is very narrow (so that these objects are almost identical), even a ‘superior resemblance’ is not satisfactory (ibid.: 182). On the other hand, if objects are very different from one another, ‘we are often contented with the most imperfect resemblance’ (ibid.). The ‘imperfect resemblance’ criterion calls attention to the epistemic framework of induction by analogy (as described, for instance, in John Maynard Keynes’s Treatise on Probability, Part III). In particular, the ‘satisfaction’ associated with imperfect resemblance suggests the conjecture that social co-ordination may be more likely between distant rather than between close but different individuals (or social groups). This conjecture

118

Pier Luigi Porta and Roberto Scazzieri

may be explored by examining under which conditions a sufficient degree of social diversity may enhance the likelihood of social co-ordination. A preliminary step in this investigation is required. The reason is that likelihood of co-ordination reflects characteristics of the prevailing modes of mental associations across different individuals or social groups. Coordination presupposes a specific structure of conceptual spaces. To the analysis of this structure we now turn. Any given conceptual space may be defined, following Gardenfors (1994), as a set S consisting of a class D1, . . . , Dn of ‘quality dimensions’ (p. 124). This means that any given element belonging to set S may be represented as a n-dimensional vector x°  (d1, . . . , dn), such that each numerical value d1, . . . , dn measures the particular characteristic of element x° that corresponds to that specific dimension. The interpretation of conceptual space S in terms of a ‘social universe’ (that is, as a space of individuals and social groups) suggests that each individual may be considered as a point in a n-dimensional space of social characteristics. For example, individual x° will be identified by a set of social characteristics such as gender, cognitive ability, economic status. It has been argued that ‘[t]he most fundamental reason for separating a cognitive structure into “dimensions” is that this separation expresses the assumption that an individual can be assigned some properties independently of other properties when the properties belong to different “dimensions” ’ (Gardenfors 1994: 124). The above argument entails that, when a social universe is considered, any given individual can be assigned certain characteristics independently of characteristics associated with other dimensions. For example, individual Ai can be assigned a certain identity independently of her/his socio-economic condition, but no individual can be described as ‘rich’ or ‘poor’ at the same time, due to the fact that predicates such as ‘rich’ and ‘poor’ belong to the same dimension of the social universe. The above representation of the social universe as a conceptual space has a number of interesting features, which we shall now discuss in some detail. First, a comparison between two individuals Ai and Aj can be undertaken only if characteristics belonging to the same dimension are considered. On the other hand, Ai and Aj are incommensurable if different dimensions are considered.7 A result of the previous discussion is the discovery that convergence presupposes commensurability of conceptual spaces, and that the latter requires some degree of internal differentiation of these spaces. This means that increasing differentiation within each conceptual space increases the likelihood of a common core of shared beliefs across heterogeneous agents. This proposition has noteworthy implications as to the feasibility of economic and social co-ordination. Co-ordination presupposes some degree of commensurability: the commensurability of different conceptual spaces is a necessary condition for individuals and social groups to be able to take full advantage of diversity.

Social knowledge in economic analysis 119 Commensurability of conceptual spaces is a necessary condition for effective diversity, since radically incommensurable conceptual spaces make it impossible to ‘share’ a common core of indirect social knowledge. There is a seeming paradox here: effective diversity presupposes some degree of commensurability, while commensurability itself is enhanced by diversity (see above). As a matter of fact, some degree of diversity is a necessary condition for effective social co-ordination (as in the case of trade or division of labour). However, ‘extreme diversity’ makes communication impossible: for any given pair of conceptual spaces Si and Sj, a complete breakdown of communication takes place if the intersection of Si and Sj is an empty set. Diversity across conceptual spaces enhances the likelihood of communication as long as it is associated with diversity within each conceptual space. Communication presupposes the inherent variety of any given conceptual space, and internally differentiated conceptual spaces are more likely to be mutually commensurable. The possibility of effective ‘translation’ from one conceptual space to another is associated with the existence of common dimensions. In particular, the possibility to translate cultural beliefs from one conceptual space to another presupposes a common core of shared ‘primitive’ beliefs.8 Spontaneous conformity reflects the ability to recognise salient features that may suggest a disposition to adopt mutually acceptable patterns of behaviour. In this case, co-ordination presupposes the ability to ‘discover’ suitable focal points (see Schelling 1960, 1978; Bacharach 1993; Sugden 1995). A focal point is a centre of convergent expectations that may generate either a change of theory or a change of strategy (see Rubinstein and Zhou 1999). This formal symmetry between the two types of change suggests some degree of congruence between epistemic processses and choice processes. In other words, any given change of theory (or strategy) presupposes a suitable frame of mind (or set of cultural beliefs). This implies that the existence of a ‘congruent set’ of choices and actions presupposes the congruence of corresponding beliefs. Economic (and social) co-ordination appears to be rooted in some congruence of human expectations (belief congruence) and the latter is made possible by commensurability of conceptual spaces. The relationship between commensurability of conceptual spaces and internal diversity of these spaces (see above) could be a reason behind the paradox of co-ordination considered at the beginning of this section. Diversity seems to be at the same time a factor of cohesion and a source of economic (and social) breakdown. To investigate this issue, it is worth exploring the structure of conceptual spaces that is generally presupposed in economic (and social) co-ordination. In this connection, we may assume that economic (and social) interaction primarily derives from reciprocal expectations concerning the processes by which social knowledge is obtained. It follows that the relevant conceptual spaces would be spaces of cognitive beliefs. This means that human decisions as to whether to

120

Pier Luigi Porta and Roberto Scazzieri

‘co-ordinate’ or not, and as to the specific pattern of co-ordination to be followed, are likely to reflect reciprocal expectations as to the course taken by the processing of information and the formation of knowledge. It seems reasonable to assume that individuals’ decisions as to whether to ‘co-ordinate’ or not primarily reflect individuals’ reciprocal expectations about their respective cognitive dispositions and capabilities. In other words, coordination between individual (or social group) Ai and individual (or social group) Aj presupposes not so much a commonality of interest as the reciprocal expectation that Ai and Aj are likely to adopt similar descriptions of the world, and are likely to follow congruent cognitive procedures. The above argument entails that belief spaces relevant to economic and social co-ordination are primarily spaces of (expected) epistemic beliefs. It is worth exploring the implications of the above conceptual set-up as to the relationship between diversity and co-ordination. As previously argued, the commensurability of conceptual spaces is enhanced by the inherent variety of any given space. It follows that the inherent variety of any given collection of epistemic beliefs is likely to enhance the likelihood that different belief spaces would be commensurable. This suggests that co-ordination within a social universe encompassing a variety of different conceptual spaces is more likely when the inherent diversity of each conceptual space is significant. In particular, the likelihood of coordination is enhanced by the diversity of any given space of epistemic beliefs. This means that different individuals (or social groups) are more likely to ‘co-ordinate’ when their respective belief spaces are inherently diverse, and in particular when any given individual (or social group) is able to ‘recognise’ a sufficiently differentiated set of admissible reasons and arguments.9 The above discussion suggests that, given a social universe characterised by heterogeneous belief spaces, economic (and social) co-ordination is more likely if the inherent diversity of each belief space is large enough, and if social agents are able to derive indirect social knowledge by a variety of inferential procedures. In a sense, inherently diverse procedures of inference (which we may call ‘mixed arguments’) enhance the likelihood of social co-ordination. The reason is that a mixed argument is more likely to intersect a significant number of different belief spaces, thus making these spaces commensurable.

Sentiments, fellow-feeling and social co-ordination This chapter has outlined a general conceptual framework aimed at the analysis of economic and social co-ordination as a process associated with a certain ‘harmony’ of feelings and cognitive frameworks (conceptual spaces). A critical feature of our investigation has been the emphasis upon the continuum between direct and indirect social knowledge, and the idea that social co-ordination often presupposes a sophisticated blend of imme-

Social knowledge in economic analysis 121 diate experiences and feelings derived by ‘moral inference’.10 Moral sentiments (in Smith’s own sense) are at the root of social (and economic) coordination (see also Macfie 1967a, 1967b; Sen 1987; Rothschild 2001). The economist’s riddle (combination of freedom and variety) may be solved in terms of a suitable belief structure. Individuals and social groups are able to find a pattern of congruence by means of self-command, self-restraint, and other behavioural criteria by which self-interest becomes historically determinate and socially acceptable. The way in which ‘social harmony’ is achieved by means of belief coordination is discussed in an impressive section of Smith’s Theory of Moral Sentiments. Here Smith starts with the distinction between direct and indirect feelings. The former are immediate experiences, such as one’s own joy, passion or grief. The latter derive from imagination, when this faculty is applied to other persons’ feelings. The possible divergence between direct and indirect feelings is, in Smith’s view, the most critical issue for social co-ordination. And the achievement of a satisfactory equilibrium of feelings is considered by Smith as ‘sufficient for the harmony of society’ (Smith [1759] 1976a: 22). In particular, a correspondence between our own feelings and those of the person ‘whose sentiments we judge of’ (ibid.: 20) is important ‘[w]ith regard to those objects which affect in a particular manner’ either ourselves or the other (ibid.). The reason is that ‘[t]he person principally concerned . . . desires a more complete sympathy. He longs for that relief which nothing can afford him but the entire concord of the affections of the spectators with his own’ (ibid.: 22). In general, however, spectators cannot ‘conceive, for what has befallen another, that degree of passion which naturally animates the person principally concerned’ (ibid.: 21). This moral setting is one of normal divergence between direct and indirect feelings. However, the likelihood of such a divergence is precisely the situation from which a correspondence of feelings may originate. The route by which this is achieved presupposes a relational propensity on both sides of a virtual equilibrium of sentiments. On the one hand, ‘the spectator must, first of all, endeavour, as much as he can, to put himself in the situation of the other, and to bring home to himself every little circumstance of distress which can possibly occur to the sufferer’ (ibid.: 21). On the other hand, the person who is ‘principally concerned’ knows that, to achieve the correspondence of feeling s(he) is longing for, s(he) must learn how to reduce her (his) own sentiments into a shape compatible with the spectators’ perceptive grid. For this to be possible, that person ‘must flatten, if I may be allowed to say so, the sharpness of [the] natural tone [of his passion], in order to reduce it to harmony and concord with the emotions of those who are about him’ (ibid.: 22). This mutual adjustment is essential to social co-ordination and makes it to coincide with a sophisticated interplay of sentiments and beliefs. Smith’s approach suggests that social co-ordination may vary both in

122

Pier Luigi Porta and Roberto Scazzieri

‘density’ and scope. In the former case, the ‘harmony of society’ is associated with the degree to which a ‘more complete sympathy’ may be achieved (ibid.: 22). Social co-ordination is more or less ‘dense’ depending upon the intensity of sentiments that individuals may reasonably be expected to share.11 On the other hand, co-ordination is more or less ‘extensive’ depending upon the number of social spheres in which we may reasonably expect to find some degree of ‘sentimental proximity’. An important implication of Smith’s analytical framework is that different layers of social co-ordination presuppose different degrees of ‘harmony and concord’. This means that individuals expect ‘less sympathy from a common acquaintance than from a friend’, and ‘still less sympathy from an assembly of strangers’ (ibid.: 23). However, the degree of correspondence that is ‘sufficient for the harmony of society’ is different in the various cases. The ‘tranquillity and sedateness’ that comes with successful social interaction (ibid.: 22) presupposes an almost perfect ‘harmony and concord’ in the case of friends (or relatives). The lesser sympathy we may reasonably expect from a ‘common acquaintance’ makes us ‘fix our thoughts upon those general outlines of our situation which he is willing to consider’ (ibid.: 23). This makes social co-ordination less ‘dense’, but also increases its scope (we may be able to co-ordinate with a larger number of people precisely because the degree of correspondence sufficient for the ‘harmony of society’ has decreased). And, in a striking twist of his argument, Smith points out that interaction with common acquaintances could even enhance the satisfaction to be drawn from social intercourse. The reason is that ‘we cannot open (to them) all those little circumstances which we can unfold to (a friend)’ (ibid.: 23). The same argument leads Smith to the view that an ‘assembly of strangers’ is even more likely to enhance the advantages of social interaction: ‘we assume . . . still more tranquillity before them, and always endeavour to bring down our passion to that pitch, which the particular company we are in may be expected to go along with’ (ibid.: 23). Smith’s conclusion is that: ‘[I]f we are masters of ourselves, the presence of a mere acquaintance will really compose us, still more than that of a friend; and that of an assembly of strangers still more than that of an acquaintance’ (ibid.: 23). In short, the effectiveness of social interaction is proportional to its ‘civilizing influence’; that is, to the extent to which social interaction induces individuals (and social groups) to ‘tone down’ their sentiments (and claims) in order to make them acceptable within a larger social universe. Smith’s theory of social beliefs and social intercourse leads him to the remarkable view that the advantages of co-ordination are greatest when skilled social actors are left free to interact within an ‘open space’ of mutual adjustments: ‘[s]ociety and conversation . . . are the most powerful remedies for restoring the mind to his tranquillity . . . Men of retirement

Social knowledge in economic analysis 123 and speculation, who are apt to sit brooding at home over either grief or resentment, though they may often have more humanity, more generosity, and a nicer sense of honour, yet seldom possess that equality of temper which is so common among men of the world’ (ibid.: 23).

An epilogue This chapter has considered a set of connected themes suggested by Smith’s theory of social co-ordination. In particular, we have examined the relationship between direct and indirect knowledge and investigated the way in which this relationship may highlight the formation of social codes of communication (social knowledge). We have argued that this process takes place through the discovery of ‘imperfect resemblances’ across different social situations. The epistemic turn of recent economic theorising (see p. 110) calls attention to the role of socially diffused and often informal knowledge as an important route to economic (and social) co-ordination. In this perspective, the existence of some fundamental congruence between linguistic and economic co-ordination may be suggested (see, for example, Scazzieri 1999a: particularly pp. 290–294; Rubinstein 2000: particularly pp. 3–36). This chapter has explored the relationship between the linguistic construal of human experience and the ability to identify a set of codes allowing social co-ordination.12 In this perspective, co-ordination appears as an outcome of effective signalling and successful induction. Co-ordination (in a complex setting) is seldom achieved by simply adopting a system of congruent codes. The direct knowledge acquired through pattern identification is generally complemented by the indirect knowledge developed through mental associations and reasoning. Adam Smith’s theory of ‘imperfect resemblance’, likelihood and analogy suggests an interesting framework for the analysis of the relationship between social knowledge and social co-ordination. Social knowledge may be identified with a collection of conguent (that is, mutually compatible) communication beliefs. These are widely held conjectures about common standards by which information is transferred within any given social set-up. The above analysis (see, in particular, pp. 115–116) has shown that, in Smith’s view, the linguistic construal of any given situation is closely associated with the ability to detect cases of ‘imperfect resemblance’ within a diverse universe of objects. In the ‘first formation of languages’, ‘new objects had none of them any name of its own, but each of them exactly resembled another object, which had such an appellation’ (Smith [1971] 1983b: 203–204). The criterion of ‘imperfect resemblance’ is also seen at the root of the formation of abstract words such as adjectives: ‘[t]hose . . . who first invented the words green, blue, red, and the other names of colours, must have observed and compared together a great number of objects, must have remarked their resemblances and dissimilitudes in respect of the quality of colour, and must have arranged them, in

124

Pier Luigi Porta and Roberto Scazzieri

their own minds, into different classes and assortments, according to those resemblances and dissimilitudes’ (ibid.: 206). The formation of language shows a blend of direct knowledge (the primitive ‘coding’ of objects) and indirect knowledge (the introduction of general categories by means of analogy and induction). It is worth noting that the interplay between language formation and social coding is, in Smith’s view, an important feature in the formation of some of the simplest verbal utterances: ‘[w]hen [primitive speakers] had occasion . . . to mention, or to point out to each other, any of the new objects, they would naturally utter the name of the correspondent old one, of which the idea could not fail, at that instant, to present itself to their memory in the strongest and liveliest manner’ (ibid.: 204; our italics). The subsequent evolution of languages is also closely associated with the evolution of the social setting in which verbal communication takes place. As shown by Smith’s awareness of the language blends deriving from ‘mixtures of nations’, social knowledge is built from coding needs and opportunities (see p. 116 above). In particular, Smith suggests that ‘mixtures of nations’ could be associated with the formation of a modular lexicon more suitable to effective communication across different languages and cultures (ibid.: 221–223). To sum up, the relationship between language and social coding suggests that social co-ordination presupposes articulation and evolution of language according to a specific sequence of logical (and historical) steps: 1

2

3

Elementary linguistic codes are necessary even at a primitive stage of social co-ordination. At this level, linguistic codes are needed in order to ‘grasp’ human experience in terms of a useful lexicogrammatical system (see also Halliday 1999: 92). The cognitive criterion of ‘imperfect resemblance’ allows human beings to develop their language in terms of abstract words (categories), by which the effectiveness of communication is enhanced. In terms of Smith’s conceptual framework, it appears at this stage that language allows communication beyond closely knit social groups and communities. The human faculty of ‘imagination’ suggests that significant components of human experience may be shared, to an increasing extent, if complex sets of communication codes are decomposed into a collection of elementary expressions. In Smith’s view, the formation of ‘mixtures of nations’ is likely to be associated with the transition from simple to ‘compounded’ languages; that is, to languages characterised by greater utilisation of modular expressions (see pp. 115–116 above).

The above conceptual framework calls attention to the role of ‘imagination’ in filling the gaps of human (direct) experience. This faculty makes it possible for human beings to establish customary connections and to introduce a virtual (but nonetheless effective) pattern of social congruence. In

Social knowledge in economic analysis 125 this way, both language and communication beliefs (as defined above) are associated with the development of a reflective faculty, by which individuals (and social groups) come to look upon themselves in the mirror of others’ expectation and judgement. In Smith’s words: ‘our continual observations upon the conduct of others, insensibly lead us to form to ourselves certain general rules concerning what is fit and proper either to be done or to be avoided’ (Smith [1759] 1976a: 159). Human beliefs about what is to be expected in social interaction are inherently relational. That is to say, individual A’s expectation about B’s behaviour is intertwined with A’s expectation about her (his) own behaviour under similar circumstances. On the other hand, A’s expectation about himself (herself) reflects his (her) expected perception in the eyes of (relevant) spectators. In other words, own perception is primarily perception in the mirror of others. This perspective suggests a close relationship between the formation of moral sense and the structure of communication beliefs. As Joseph Butler had pointed out before Adam Smith, moral sense (conscience) may be associated with a ‘principle of reflection’ (Butler, 1736). This principle entails a ‘dispassionate reflective approval or disapproval of actions and principles (motives)’ (Darwall 1995: 245n). In its original formulation, Butler’s ‘reflex approbation or disapprobation’ (Butler, 1726, as quoted in Darwall 1995: 246) is primarily addressed to an individual’s self-evaluation (see Darwall 1995: 245n). Smith’s conceptual twist makes self-evaluation a consequence of the individual’s beliefs about herself (himself) in the mirror of others. This suggests an important relationship between moral sentiments and social co-ordination. The structure of moral sense (conscience) presupposes, in Smith’s view, ‘a continuing, conversational morality’ (Rothschild 2001: 231). On the one hand, standards of behaviour are born from the propensity of individuals to shift from self-evaluation to evaluation of oneself in the mirror of others. On the other hand, this same propensity tones down human passions in order to enhance the likelihood of a correspondence of feeling (see p. 117). Social interaction induces the formation of the moral sense: ‘we suppose ourselves the spectators of our own behaviour, and endeavour to imagine what effect it would, in this light, produce upon us’ (Smith [1759] 1976a: 112). As a result, society appears to be ‘the only looking-glass by which we can, in some measure, with the eyes of other people, scrutinize the propriety of our own conduct’ (ibid.).13 This point of view, as argued above (see pp. 121–122), is consistent with the consideration that a variety of moral looking-glasses is possible, depending on the distance of the relevant spectator(s) in each particular case. In his investigation of moral sentiments, Smith calls attention to the critical influence of communication and linguistic standards in establishing moral criteria. Indeed, Smith suggests that the moral sense itself presupposes social interaction. However, the evolution of social interaction into sophisticated patterns of ‘social harmony’ requires the existence of

126

Pier Luigi Porta and Roberto Scazzieri

skilled social actors. In particular, it requires social actors capable to decompose a web of social connections into partially overlapping social spheres. Each social sphere would correspond to a particular domain of co-ordinated beliefs, and would be associated with a specific pattern of ‘fellow-feeling’ and ‘social harmony’. To conclude, social knowledge is born from increasingly common codes of communication. At the same time, the existence of such codes enhances the likelihood that social actors be able to detect domains of ‘fellow-feeling’ among otherwise distant individuals or social groups.

Notes 1 This chapter is a further step in the exploration of co-ordination and connecting principles, which we have already considered in Porta and Scazzieri (2001). As in our former contribution, the conceptual framework of our analysis is inspired by Adam Smith’s theory of social induction. In this chapter, we devote special attention to the relationship between connecting principles, ‘fellowfeeling’ and social knowledge. Financial support from the Italian Ministry of Education and Research (MURST) – national project on ‘Heterogeneous Agents, Financial Fragility, Capital Markets and Complex Dynamics’ – is gratefully acknowledged. 2 It is worth recalling, in this connection, Schelling’s description of the cognitive requirements of co-ordination games: ‘Whenever the communication structure does not permit players to divide the task ahead of time according to an explicit plan, it may not be easy to co-ordinate behaviour in the course of the game. Players have to understand each other, to discover patterns of individual behaviour that make each player’s actions predictable to the other; they have to test each other for a shared sense of pattern or regularity and to exploit clichés, conventions and impromptu codes for signalling their intentions and responding to each other’s signals’ (Schelling 1960: 84–85). An interesting analogue of the imperfect congruence of social belief structures is the co-existence of different, but partially overlapping, research lines in certain fields of scholarship, such as economic theory (see, for example, Baranzini and Scazzieri 1986: particularly pp. 1–5 and 77–78). 3 In John Stuart Mill’s formulation, ‘two things resemble each other in one or more respects; a certain proposition is true of the one; therefore it is true of the other’ (see J.S. Mill, ‘On Analogy’, in Nagel and Brandt 1965: 350). 4 In this regard, Bagolini has argued that ‘the prerequisites of sympathy, precisely as a “connecting principle”, are the diversity and contrasts of the human situations upon which sympathetic judgement can be applied’ (Bagolini 1975: 153). The relationship between this cognitive set-up and division of labour suggests a new, interesting field of investigation (see, in particular, Loasby 1996, 1999). 5 Smith’s ‘mutual consent’ should not be associated with a formal agreement of the ‘contractual’ type. It should rather be conceived as a tacit convention stemming from a de facto utilisation of common signs. The latter interpretation would be consistent with Adam Ferguson’s view that ‘[m]en, in general, are sufficiently disposed to occupy themselves in forming projects and schemes: but he who would scheme and project for others, will find an opponent in every person who is disposed to scheme for himself. Like the winds, that come we know not whence, and blow whithersoever they list, the forms of society are derived from an obscure and distant origin; they arise, long before the date of

Social knowledge in economic analysis 127

6 7 8

9

10 11

12

13

philosophy, from the instincts, not from the speculations, of men . . . No constitution is formed by concert, no government is copied from a plan’ (Ferguson, [1767] 1966: 122–123). Cesare Beccaria, in his analysis of the principles of a monetary economy written a few years before Ferguson, had explicitly objected to the contractualist reconstruction of social codes and institutions: ‘(T)he introduction of money was not born from an explicit convention (which has never been the source of a universal establishment), but was born from what ordinary people call chance, that is, from an arrangement of circumstances not designed by human beings’ (Beccaria [1762] 1986: 54). This section takes up, within a Smithian framework, a theme initially explored in Scazzieri (2000). The above condition for commensurability is briefly considered by Gardenfors (1994), who refers to the earlier discussion by W.E. Johnson (1921). It may be argued that translation from belief space Bi to belief space Bj requires a rational reconstruction of existing beliefs, by which: (1) existing beliefs are decomposed into elementary belief units; (2) a mapping is introduced from elementary belief units in Bi to elementary belief units in Bj; (3) translation is completed by identifying a suitable cluster of elementary belief units in Bj (see Scazzieri 1999a: 300). Patrick Suppes has recently argued that, in an uncertain environment, diversity enhances freedom of choice (see Suppes 1997). The above argument suggests that ‘inherent diversity’ may be at the root both of freedom of choice and social co-ordination. The relevance of the above epistemic continuum for social co-ordination is briefly considered in Marzetti and Scazzieri (1999). (See also Scazzieri 2002.) There may be a connection between a society characterised by dense social coordination (as defined on pp. 121–122) and a society characterised by social closure, such as the ‘segmentary community’ described by Ernest Gellner (Gellner 1994). In this type of social setting, the individual is turned ‘into an integral part of the social sub-unit’ (Gellner 1994: 8). It may be conjectured that, under these conditions, the correspondence of sentiments is expected to be almost complete, at least in an ‘exterior’ or ritualised way. The conjecture that any given language is associated with a particular way of ‘ordering’ direct experience has been investigated by Michael Halliday (Halliday 1999). In particular, Halliday considers the relationship between the identification of congruent patterns and the formation of grammar: ‘a grammar – that is, a lexicogrammatical system of any and every natural language – is a theory of human experience. It is a way of imposing order, in the form of mutually defining relationships and categories, on the flux and unboundedness of the perceptual world’ (Halliday 1999: 92). Halliday’s conception of a ‘theory of human experience’ absorbed in the process of language acquisition bears a close resemblance with the view put forward by Thomas Kuhn in the late phase of his investigation on incommensurability and translation in science: ‘the acquisition of conceptual vocabulary requires giving to some laws of nature a definitional role that makes their cognitive status like that of Kant’s synthethic a priori. As other laws are discovered with the aid of those initially posited, they too inherit that cognitive status . . . It is laws that do or could enter the language acquisition process in that way that language thereafter projects back upon the world’ (Kuhn 1999: 36). It may be argued that, depending on established cultural beliefs, society may alternatively enhance or thwart the individual’s thrust to the realisation of one own’s capabilities. The role of society as a distorting mirror was carefully investigated by Pietro Verri (see Verri 1773; see also Porta and Scazzieri 2002: particularly pp. 92–96).

128

Pier Luigi Porta and Roberto Scazzieri

References Arrow, K.J. (1986) ‘Rationality of Self and Others in an Economic System’, in R.M. Hogarth and M.W. Reder (eds), Rational Choice. The Contrast between Economics and Psychology, Chicago: The University of Chicago Press, 201–216. —— (1994) ‘Methodological Individualism and Social Knowledge’, American Economic Review, 84(2): 1–9. Arrow, K.J. and Hahn, F.O. (1999) ‘Notes on Sequence Economies, Transactions Costs, and Uncertainty’, Journal of Economic Theory, 86(2): 203–218. Aumann, R.J. and Drèze, J.H. (1974) ‘Cooperative Games with Coalition Structures’, International Journal of Game Theory, 3: 217–237. Bacharach, M. (1993) ‘Variable Universe Games’, in K. Binmore, A. Kirman and P. Tani (eds), Frontiers of Game Theory, Cambridge, Mass.: The MIT Press, 255–275. —— (1998) ‘The Problem of Trust: A Signal-Theoretic Approach’, Paper presented at the Seminar on Trust and Information, All Souls College, Oxford, Trinity Term. Bagolini, L. (1975) La simpatia nella morale e nel diritto. Aspetti del pensiero di Adam Smith e orientamenti attuali (3rd edn), Torino: Giappichelli. Baranzini, M. and Scazzieri, R. (1986) ‘Knowledge in Economics. A Framework’, in M. Baranzini and R. Scazzieri (eds), Foundations of Economics. Structures of Inquiry and Economic Theory, Oxford and New York: Basil Blackwell, 1–87. Beccaria, C. ([1762] 1986) Del disordine e de’ rimedi delle monete nello Stato di Milano nell’anno 1762, in A. Quadrio Curzio and R. Scazzieri (eds), Sul disordine delle monete a Milano nel Settecento, Milan: Electa for Banca del Monte di Milano, 49–100. Becker, G. and Murphy, K.M. (1992) ‘The Division of Labour, Coordination Costs, and Knowledge’, The Quarterly Journal of Economics, 107(4): 1137–1160. Berlin, I. (1958) Two Concepts of Liberty. An Inaugural Lecture Delivered Before the University of Oxford on 31 October 1958, Oxford: Clarendon Press. Bortis, H. (1997) Institutions, Behaviour and Economic Theory, Cambridge: Cambridge University Press. Bruni, L. and Sugden, R. (1998) Moral Canals. Trust and Social Capital in the Work of Hume, Smith and Genovesi, University of East Anglia, School of Economic and Social Studies, Norwich, mimeo. Butler, J. (1726) A Dissertation of the Nature of Virtue, London. —— (1736) The Analogy of Religion, Natural and Revealed, to the Constitution and Course of Nature, London, printed for John and Paul Knapton. Constant, B. (1815) Principes de politique, applicables à tous les gouvernments représentatifs et particulièrement à la constitution actuelle de la France, Paris. Crisma, L. (1988) ‘Dalla certezza all’incertezza: aspetti dinamici in una impostazione soggettiva’, in Atti del Convegno su incertezza ed economia, University of Trieste, Faculty of Economics, Trieste: Edizioni Lint, 11–46. Darwall, S. (1995) The British Moralists and the Internal ‘Ought’: 1640–1740, Cambridge: Cambridge University Press. Dasgupta, P. (2002) Social Capital: A Multifaceted Approach, Oxford: Oxford University Press. David, P. (1994) ‘Why are Institutions the “Carriers of History”? Path Depen-

Social knowledge in economic analysis 129 dence and the Evolution of Conventions, Organizations and Institutions’, Structural Change and Economic Dynamics, 5(2): 205–220. Ferguson, A. ([1767] 1966) An Essay on the History of Civil Society, Edinburgh: Edinburgh University Press. Gardenfors, P. (1994) ‘Induction, Conceptual Spaces and Artificial Intelligence’, in D. Stalker (ed.), Grue. The New Riddle of Induction, Chicago and La Salle, Ill.: Open Court, 117–134. (Originally published in Philosophy of Science, 1990, 57: 78–95.) Geertz, C. (1978) ‘The Bazaar Economy: Information and Search in Peasant Marketing’, The American Economic Review, 68(2): 28–32. Gellner, E. (1994) Conditions of Liberty. Civil Society and its Rivals, London: Penguin Books. Gilboa, I. and Schmeidler, D. (2001) A Theory of Case-Based Decisions, Cambridge: Cambridge University Press. Greif, A. (1994) ‘Cultural Beliefs and the Organization of Society: A Historical and Theoretical Reflection on Collectivist and Individualist Societies’, Journal of Political Economy, 102: 912–950. Hahn, F.O. (1978) ‘On Non-Walrasian Equilibria’, Review of Economic Studies, 45: 1–17. —— (1987) ‘Conjectural Equilibria’, in The New Palgrave. A Dictionary of Economics (eds J. Eatwell, M. Milgate and P. Newman), London and Basingstoke: Macmillan, 575–579. Halliday, M.A.K. (1999) ‘The Grammatical Construction of Scientific Knowledge: The Framing of the English Clause’, in R. Rossini Favretti, G. Sandri and Roberto Scazzieri (eds), Incommensurability and Translation. Kuhnian Perspectives on Scientific Communication and Theory Change, Cheltenham, UK and Northampton, Mass.: Edward Elgar, 85–116. Johnson, W.E. (1921) Logic, London: Cambridge University Press. Keynes, J.M. (1921) A Treatise on Probability, London: Macmillan. Kuhn, T.S. (1999) ‘Remarks on Incommensurability and Translation’, in R. Rossini Favretti, G. Sandri and Roberto Scazzieri (eds), Incommensurability and Translation. Kuhnian Perspectives on Scientific Communication and Theory Change, Cheltenham, UK and Northampton, Mass.: Edward Elgar, 33–37. Landes, D. (1998) The Wealth and Poverty of Nations. Why Some Are So Rich and Some So Poor, London: Little, Brown & Company. Loasby, B.J. (1996) ‘The Division of Labour’, History of Economic Ideas, IV(1–2): 299–324. —— (1999) Knowledge, Institutions and Evolution in Economics, The Graz Schumpeter Lectures, London: Routledge. Macfie, A.L. (1967a) ‘Adam Smith’s “Moral Sentiments” as Foundation for his “Wealth of Nations” ’, in Macfie (1967b): 59–81. —— (1967b), The Individual in Society. Papers on Adam Smith, London: Allen & Unwin. Marzetti Dall’Aste Brandolini, S. and Scazzieri, R. (1999) ‘Introduzione’, in S. Marzetti Dall’Aste Brandolini and R. Scazzieri (eds), La probabilità in Keynes: premesse e influenze, Bologna: CLUEB, 11–23. Matteucci, N. (1993) ‘Per una definizione teorica della libertà’, Filosofia Politica, no. 2: 277–283. Mehta, J., Starmer, C. and Sugden, R. (1994) ‘The Nature of Salience: An

130

Pier Luigi Porta and Roberto Scazzieri

Experimental Investigation of Pure Coordination Games’, The American Economic Review, 84(3): 658–673. Nagel, E. and Brandt, R.B. (eds) (1965) Meaning and Knowledge, New York: Harcourt, Brace and World. North, D.C. (1990) Institutions, Institutional Change and Economic Performance, Cambridge: Cambridge University Press. —— (1999) Understanding the Process of Economic Change, Twenty-eighth Wincott Memorial Lecture, Occasional paper no. 106, London: Institute of Economic Affairs. Pasinetti, L.L. (1981) Structural Change and Economic Dynamics. A Theoretical Essay on the Dynamics of the Wealth of Nations, Cambridge: Cambridge University Press. —— (1993) Structural Economic Dynamics. A Theory of the Economic Consequences of Human Learning, Cambridge: Cambridge University Press. Poni, C. and Scazzieri, R. (eds) (1994) Production Networks: Market Rules and Social Norms, Paper presented at the 11th International Economic History Congress, Session c.11, Milan, Bocconi University, 12–16 September, Department of Economics, University of Bologna, mimeo. Porta, P.L. (1992) ‘The Present as History in Economic Analysis’, in History of Thought: How and Why? (ed. A. Salanti), Quaderni del Dipartimento di Scienze Economiche, Università di Bergamo, 31–34. Porta, P.L. and Scazzieri, R. (2001) ‘Coordination, Connecting Principles and Social Knowledge: An Introductory Essay’, in P.L. Porta, R. Scazzieri and A.S. Skinner (eds), Knowledge, Social Institutions and the Division of Labour, Cheltenham, UK and Northampton, Mass.: Edward Elgar, 1–32. —— (2002) ‘Pietro Verri’s Political Economy: Commercial Society, Civil Society and the Science of the Legislator’, History of Political Economy, 34(1): 81–108. Putnam, R.D. (1993) Making Democracy Work: Civic Traditions in Modern Italy, Princeton, N.J.: Princeton University Press. —— (1995) ‘Bowling Alone: America’s Declining Social Capital’, Journal of Democracy, 6(1): 65–78. —— (2000) Bowling Alone: the Collapse and Revival of American Community, New York: Simon & Schuster. Rossini Favretti, R., Sandri G. and Scazzieri, R. (1999) ‘Translating Languages: An Introductory Essay’, in R. Rossini Favretti, G. Sandri and R. Scazzieri (eds), Incommensurability and Translation. Kuhnian Perspectives on Scientific Communication and Theory Change, Cheltenham, UK and Northampton, Mass.: Edward Elgar, 1–29. Rothschild, E. (2001) Economic Sentiments. Adam Smith, Condorcet and the Enlightenment, Cambridge, Mass. and London, UK: Harvard University Press. Rubinstein, A. (2000) Economics and Language. Five Essays, Cambridge: Cambridge University Press. Rubinstein, A. and Zhou, L. (1999) ‘Choice Problems with a “Reference” Point’, Mathematical Social Sciences, 37(3), 205–209. Scazzieri, R. (1999a) ‘Economic Beliefs, Economic Theory and Rational Reconstruction’, in R. Rossini Favretti, G. Sandri and R. Scazzieri (eds), Incommensurability and Translation. Kuhnian Perspectives on Scientific Communication and Theory Change, Cheltenham, UK and Northampton, Mass.: Edward Elgar, 289–306.

Social knowledge in economic analysis 131 —— (1999b), ‘Analogia, fiducia e società civile: la struttura morale della probabilità nel contributo di John Maynard Keynes’, in S. Marzetti Dall’Aste Brandolini and R. Scazzieri (eds), La probabilità in Keynes: premesse e influenze, Bologna: CLUEB, 189–216. —— (2000) ‘Conceptual Spaces and the Varieties of Inference. A Perspective from Economic Analysis’, Paper presented at the inaugural workshop of the Department of Mathematics applied to Economic and Social Sciences, University of Bologna. —— (2001a) ‘Analogy, Causal Patterns and Economic Choice’, in M.C. Galavotti, P. Suppes and D. Costantini (eds), Stochastic Causality, Stanford: CSLI Publications, 123–139. —— (2001b) ‘Patterns of Rationality and the Varieties of Inference’, Journal of Economic Methodology, VIII: 105–110. —— (2002) ‘Practical Reason and Probabilistic Coordination’, Paper presented at the Colloquium ‘Rationality and Pragmatism; Workshop in Honour of Patrick Suppes’, University of Bologna, 22–23 March. Schelling, T.C. (1960) The Strategy of Conflict, Cambridge, Mass. and London, UK: Harvard University Press. —— (1978) Micromotives and Macrobehavior, New York and London: W.W. Norton & Company. Schmidt, C. (2001) ‘From the “Standards of Behaviour” to the “Theory of Social Situations”: A Contribution of Game Theory to the Understanding of Institutions’, in P.L. Porta, R. Scazzieri and A. Skinner (eds), Knowledge, Social Institutions and the Division of Labour, Cheltenham, UK and Northampton, Mass.: Edward Elgar, 153–167. Sen, A.K. (1987) On Ethics and Economics, Oxford: Basil Blackwell. —— (1999) Reason Before Identity. The Romanes Lectures for 1998, Oxford: Oxford University Press. Simon, H. (1983) Reason in Human Affairs, Oxford: Basil Blackwell. Skinner, A.S. (1996) A System of Social Science. Papers Relating to Adam Smith (2nd edition), Oxford: Clarendon Press. Smith, A. ([1759] 1976a) The Theory of Moral Sentiments (eds D.D. Raphael and A.L. Macfie), Oxford: Clarendon Press. —— ([1776] 1976b) An Inquiry into the Nature and Causes of the Wealth of Nations (general editors R.H. Campbell and A.S. Skinner; textual editor W.B. Todd), Oxford: Clarendon Press. —— ([1795] 1980), Essays on Philosophical Subjects, edited by W.P.D. Wightman and J.C. Bryce . . . (edited by I.S. Ross; general editors D.D. Raphael and A.S. Skinner), Oxford: Clarendon Press. —— ([ms 1762–1763] 1983a) Lectures on Rhetoric and Belles Lettres (edited by J.C. Bryce; general editor A.S. Skinner), Oxford: Clarendon Press. —— ([1761] 1983b) ‘Considerations Concerning the First Formation of Languages, and the Different Genius of Original and Compounded Languages’, in Lectures on Rhetoric and Belles Lettres (edited by J.C. Bryce; general editor A.S. Skinner), Oxford: Clarendon Press, 223–226. Sugden, R. (1995) ‘A Theory of Focal Points’, The Economic Journal, 105(May): 533–550. —— (1996) ‘Rational Coordination’, in F. Farina, F. Hahn and S. Vannucci (eds), Ethics, Rationality and Economic Behaviour, Oxford: Clarendon Press.

132

Pier Luigi Porta and Roberto Scazzieri

Suppes, P. (1984) Probabilistic Metaphysics, Oxford and New York: Basil Blackwell. —— (1996) ‘The Nature and Measurement of Freedom’, Social Choice and Welfare, 13: 183–200. —— (1997) ‘Freedom and Uncertainty’, in H.G. Natke and Y. Ben-Haim (eds), Uncertainty: Models and Measures, Mathematical Research, Berlin: Academie Verlag, 69–83. Sylos Labini, P. (2001) Underdevelopment: A Strategy for Reform (Federico Caffè Lectures), Cambridge: Cambridge University Press. Verri, P. (1773) Idee sull’indole del piacere: Discorso, Livorno.

5

The making of a behavioural economist Herbert A. Simon and the early evolution of bounded rationality* Mie Augier

Introduction In the outpouring of appreciation that followed Herbert Simon’s death on 9 February 2001, there has been almost complete agreement that he was a towering figure in many twentieth-century developments in science, and his work crossed disciplinary boundaries.1 Thus judgement seems to be based on the fact that he significantly shaped the disciplines in which he worked; pioneered developments such as artificial intelligence and behavioural economics; and brought ideas such as satisficing and bounded rationality to the limelight. People want to give credit where credit is due. Unfortunately, the effect of viewing the man through the lens of subsequent development of the fields is a picture of him that is incomplete and distorted. For instance, modern research in behavioural economics sometimes credits Simon for introducing ‘limited rationality’, while ignoring most of his other insights, such as satisficing, altruism, and organizational identification (Rabin 1998, 2002; Mullainathan and Thaler 2000). Similarly, while much of today’s work in organizational economics employ terms such as incomplete contracts and bounded rationality, it really is only the work of Oliver Williamson, Richard Nelson and Sidney Winter which comes close to the heart of the ideas Simon aspired to.2 Finally, while game theorists and macro theorists seemingly have taken up the challenge of building bounded rationality into their theories, their work is far from always in the spirit of what Simon wanted.3 To do justice to Herbert Simon and his thought, we need to approach him on his own terms and in the context of his own time, rather than on our own terms and in retrospect. Understanding Herbert Simon we come also to understand something more about the development of the fields he did so much to foster, since we can view these developments through his eyes and from his perspective. We will, furthermore, come to see that he didn’t care much about differences between the disciplines; preferring instead to emphasize their commonalities. He was unusually firm in his resistance for disciplinary loyalty: ‘If you see any one of these disciplines dominating you,’ he said in conversation, ‘you join the opposition and you

134

Mie Augier

fight it for a while.’4 As a result, Simon could appear to be always leaving and never finding home; always embracing a new discipline with passion and intensity, but at the same time always appearing to be moving away. In keeping with this perspective, Simon never really joined an established disciplinary community, preferring instead to establish his own domains (such as behavioural science, cognitive psychology, artificial intelligence). Domains which were close enough to mainstream disciplines to allow recognition, yet different enough to not expect Simon to be their follower – for how can you follow when your instinct is to lead? It was this instinct to be independent and to be different, in addition to an enormous intellectual curiosity, which allowed Simon to be so broad in the scope of his research, while at the same time emphasizing the same problem in all his work. Simon was especially concerned with reforming disciplines that had an unrealistic conception of human decision-making, and particularly with representation of rationality, a concern central to many of the disciplines in which he was particularly interested. Consequently, one key to Simon’s work is his effort to work out his own analysis of decision-making at several disciplinary as well as analytical levels. To an unusual degree, Simon’s mind was comfortable with both the very concrete and the highly abstract, and he leapt readily from one to the other and back again, without spending much time in the middle where most of us live. His mind was a mind at once preoccupied with the historical instance and the big picture. As a man and a scientist, his mind was always seeking to derive models of human behaviour – be they verbal, mathematical or analogue (Simon 1954). But he wanted models which could provide concrete expressions about human behaviour. As a result, the dynamic character of Simon’s mind and personality can at least in part be traced to the dialectic relationship between the two sides of his interest. On the one hand, he was a man of science who dreamed of a better world with scientific models and he first entered the domain of social science with the ambition of spreading the use of mathematics, thinking that the fields of social science needed a little ‘stiffening up’. On the other hand, those models should correspond to the empirical realities of the real world. For instance, to Simon, mathematics was a language that could add considerably to the social sciences if it was empirically sound, but if it was not empirically sound, it didn’t matter that it was good mathematics for it was not enough that it was logically consistent. Another man might have handled the possible tension between admiration for models and the need for ‘the empirical’ by compartmentalization, but Simon found bounded rationality and managed to channel the tension into an unusual productive program of research. In effect, Simon’s interest in the empirical tied him to the mast, like Ulysses, and enabled him to attend the siren song of economics without losing the critical distance so vital for interdisciplinary and empirical inquiry. In all these ways (and many others), Simon was a man who did not fit in

H.A. Simon and bounded rationality

135

the standard boxes provided by the post-war social science professions. Neither an empiricist nor a theorist, as those terms came to be used, Simon was both in some measure, just as he was both a political scientist and an economist and a psychologist without fitting into either category very neatly. He testified to Congress, to the National Academy of Science, served as adviser for the President’s Science Advisory Committee (PSAC) and the Committee on Science and Public Policy (COSPUP), and occasionally wrote for popular magazines, but he wasn’t a policy-maker either.5 He worked outside the mainstream of economics, partly by necessity and partly by choice, and yet his name and work became widely known and influential in mainstream economics circles such as the Cowles Commission and the RAND Corporation, and in 1978 he won the Nobel Prize in Economics. This chapter discusses how Simon’s vision for behavioural economics (and social science generally) was found in the context of his early work in public administration and political science and was strengthened as Simon proceeded to make contributions to economics, and, finally, found a home with the establishment of the behavioural science in the 1950s. The next section examines the first part of this development, which was the formulation of the idea of limited rationality and the initial attempt to understand decision-making problems in the context of political/administrative organizations. The second step for Simon in realizing his behavioural vision (see pp. 139–144) was his attempt to translate his ideas into the language of economics, and his increasing awareness that much more than economics was needed to bring bounded rationality to fame. The third (and, for the purpose of this chapter, final) part in the making of behavioural economics, the beginning of the program in behavioural economics at the graduate school of industrial administration (GSIA) at Carnegie Mellon University, is covered on pp. 145–150.6 This formulation of behavioural economics has its roots in the Ford Foundation’s support of the Carnegie research in general, and Simon in particular, and their attempt to establish an interdisciplinary ‘behavioral science’. The final section concludes with some remarks on how the field of behavioural economics now, curiously, is being published in top economics journals, awarded ‘genius’ awards, and being taught at major universities. Very much contrary to Simon’s vision, the ambition of most modern behavioural economists is to contribute to the heart of mainstream economics, not to replace it with a more behavioural program. Indeed, for scholars familiar with Simon’s own work in behavioural economics, it can come as a surprise that modern behavioural economists seem so unable or unwilling to follow Simon fully into the promised land.

136

Mie Augier

Discovering the limits of rationality Born in 1916, Simon spent his early years with his parents and his older brother on the west side of Milwaukee in a middle-class neighbourhood. As a boy, he was rather shy, preferring to spend most time reading by himself; and he often walked to a park as early as 5 a.m. to climb into a tree and read before breakfast. Attending public schools, Simon at first intended to study biology. However, after he went on a strawberry hunting trip and discovered that he was colour-blind (unable to distinguish the strawberries from the plants), he changed his mind, thinking that colourblindness would be too big a handicap in biology. He then thought briefly about studying physics, but he gave up that idea after discovering that there weren’t really any major advances left to be made in physics: ‘They have all these great laws’, he said in conversation. ‘Newton had done it, no use messing around with it.’ As a result, upon finishing high school in 1933, Simon enrolled instead at the University of Chicago with an interest in making social science more mathematical, and an intention to major in economics. In keeping with his strong wish to be independent, Simon preferred reading on his own instead of taking classes; and he particularly refused to take the class in accounting, which was required to graduate in economics.7 As a result, he majored instead in political science. Political science wasn’t physics, of course, with all its ‘great laws’. However, as a science, it could encompass both theory and practice; and, being an empirical science, it had to take the data seriously. Furthermore, Simon found an appeal to interdisciplinary thinking (in particularly psychology) in understanding political behaviour, which attracted him. The details of Simon’s mature work differ, but the underlying ideas, interdisciplinary thinking and the necessity of bringing together theory and reality, remain. Also present from the start was the essential idea of limited rationality which would stay with Simon as he proceeded to translate his insights in political science and public administration into his work in economics, organization theory, psychology, and artificial intelligence. In 1935, Simon wrote a paper for a class entitled ‘Administration of Public Recreational Facilities in Milwaukee’.8 The paper was organized around the study of problems in public administration caused by the growth of municipal recreation facilities. In particular, the outgrowth of park and school activities invited studying their administration in the light of possible problems of relations between the school boards and the local city governments (p. 2). This paper, Simon often said in retrospect, provided the first insight into the idea of bounded rationality; for it was here that Simon discovered that neoclassical ideas on utility maximization didn’t fit the way in which the budget allocation process under study (in the Milwaukee Parks Department) really worked. The paper discusses the politics of the budget allocation process, and describes the governmental structure of the recreational administrative functions (such as planning

H.A. Simon and bounded rationality

137

activities, construction of new facilities, maintenance of facilities) and the legal considerations. At first, Simon wanted to use basic ideas of utility maximization in analysing the budget allocation process, having rational agents choosing the alternatives that produced the highest utility for a given allocation of resources. This rational process would assume that a ranking existed for the different possibilities and a set of preferences so that the utility for each alternative could be derived (comparing the utility of things such as ‘parks and parkways’, ‘playgrounds’, ‘museums’ and ‘municipal athletics’). What is wrong with this perspective? Perfect knowledge about the alternatives is assumed and administrative decision-makers knew what the current outcomes were. Furthermore, the decision-makers would also be able correctly to predict future ones in a way so that rationality wasn’t just ‘local’ and restricted to maximizing local alternatives but ‘global’ and maximizing all alternatives. Translated into the case of public administration, this would mean that the local administrator would evaluate the utility gained by spending money for a particular purpose and allocate the money in a way that maximized global utility by dividing up budgets between competing programs in a rational manner and to see maximization as its main goal. But this wasn’t how the budget process went on in real life, Simon observed. People couldn’t agree on the relative values of playground and parks. Rather, it appeared that decisions were determined, at least in part, by the local environment in which he worked and the local aims of that environment. For instance, one department supervisor reported that ‘the primary function of the department is to provide guidance which will make possible the more useful and enjoyable employment of leisure time by the people of the city’ (pp. 15–16).9 Thus, Simon wrote: ‘My training in economics, evoked in the context of a budget situation, disclosed a contradiction between what theory taught me ought to be happening and what my eyes and ears showed me was actually happening’ (1991: 371). He never repudiated this early work. In fact, during his career, he made the insights more elaborated and made it the centre of his research. As Simon recalled: ‘Now I had a new research problem: How do human beings reason when the conditions for rationality postulated by the model of neoclassical economics are not met?’ (Simon 1989: 376).10 As a result of this early work, Simon was invited by Clarence Ridley to participate as a research assistant in a project for the International City Manager’s Association (Simon 1991: 64). Together with Ridley, Simon published the results of this project in several articles as well as a book, Measuring Municipal Activities (Ridley and Simon 1938). This brought an invitation to join the University of California’s Bureau of Public Administration to study local government. While working in Berkeley on directing a study of the administration of state relief programs, intended to demonstrate how quantitative empirical research could contribute to understanding and improving municipal government problems (1991: 82), Simon was

138

Mie Augier

also working on an early manuscript of his thesis (what became Administrative Behavior), intended to reforming administrative theory (Simon 1947). The first working title of Administrative Behavior was ‘The Logical Structure of an Administrative Science’ (Simon 1977: xiii). Simon had intended the book to have a heavy philosophical component; in particular due to being influenced by Rudolph Carnap.11 The connection to the Milwaukee study is clear in Simon’s dissatisfaction with neoclassical rationality. ‘The implication might be drawn from this discussion that any rational choice between alternatives involves a complete description of the possibility consequential on each alternative and a comparison of these possibilities’, he wrote. ‘We would have to know in every single respect how the world would be changed by our behaving one way instead of another, and we would have to follow the consequences of behavior through unlimited stretches of time, unlimited reaches of space, and unlimited sets of descriptive variables. Under such conditions even an approach to rationality in real behavior would be inconceivable’ (Simon 1943: 38). Furthermore, Simon introduced the importance of organizations for individual decision-making, a theme later elaborated in March and Simon (1958) especially. ‘Human rationality’, he wrote, ‘gets its higher goals and integrations from the institutional settings in which it operates and by which it is molded . . . [Therefore] . . . [t]he rational individual is, and must be, an organized and institutionalized individual’ (Simon 1947: 101–102). Simon argued that organizations make it possible to make decisions by virtue of the fact that they constrain the set of alternatives to be considered and the considerations that are to be treated as relevant. Organizations can be improved by improving the ways in which those limits are defined and imposed. Finally, Administrative Behavior criticized existing administrative theory for being based on ‘proverbs’ (often contradictory common-sense principles), a perspective he wanted to replace with a more empirically oriented perspective investigating the nature of the decision processes in administrative organizations. In Simon’s view, Administrative Behavior was the first place in which he systematically examined the importance of limits to human rationality. ‘The dissertation contains both the foundation and much of the superstructure of the theory of bounded rationality that has been my lodestar for nearly fifty years’, he wrote (Simon 1991: 86). The core chapters of this book intended to develop a theory of human decision-making which was broad and realistic enough to accommodate both ‘those rational aspects of choice that have been the principal concern of the economist, and those properties and limitations of the human decision making mechanisms that have attracted the attention of psychologists and practical decision makers’ (1947: xi). Bringing together insights from economics and psychology, Simon laid the foundation for the later establishment of behavioural economics and for organization theory. In Simon’s view, the significance of his early work was in substituting ‘economic man’ with ‘administrative

H.A. Simon and bounded rationality

139

man’ by bringing insights from psychology to bear in studying decisionmaking processes (Simon 1947: xxv). Decision-making, as it is portrayed in Administrative Behavior, is purposeful, yet not rational, because rational decision-making would involve a complete specification of all possible outcomes conditional on possible actions in order to choose the single alternative action that was best. This is the important view of a man whose key ideas would remain intact for the next fifty-plus years and whose intellectual project was driven by the need to address this question, no matter what the disciplines.12 The research question that Simon had found on rationality and decisionmaking was relevant to, and had implications for, administrative theory; and since administrative theory could be scientific, it should be scientific – scientific and empirical, rather than only empirical, and rather than only scientific. The language that Simon used to develop such a theory, most notably in Administrative Behavior, was mostly English. However, youthful exposure to other languages, in particular mathematics, proved good preparation for Simon’s work in economics, and it also explains why it was relatively easy for Simon to translate some of his ideas in public administration into economics after he returned to Chicago in 1942 when he accepted a position at the Illinois Institute of Technology. For the burning question in Simon’s mind remained the same, and it concerned not what kind of discipline to use to address the overall question, but rather how to use different disciplines to form the answer, and it is through that lens that we must understand his intellectual trajectory and, hence, also Simon’s writings in economics.

Life among the Econ13 In an environment in Chicago in the early 1940s where most of his fellow researchers were believers in rational decision-making, Simon remained a strong advocate of the idea of limited rationality. He began to discuss his ideas with prominent economists, in particular those connected to the Cowles Commission which was a group of mathematical economists doing pioneering research in econometrics, linear and dynamic programming, and decision theory, among other things (Christ 1994; Mirowski 2002). The economists connected to the Cowles Commission included such wellknown names as Kenneth Arrow, Jacob Marshak, Tjalling Koopmans, Roy Radner and Gerard Debreu, and they held regular seminars to discuss their research. During the last years of Simon’s stay in Chicago he began attending the Cowles Commission seminars, and this became very important to him both because, as he noted in his autobiography, his interaction with Cowles almost made him ‘a full time economist’ (Simon 1991: 140), and because several members of the Cowles Commission would become good friends. Furthermore, Simon seemed to have realized that, for his life among the economists, the possibilities for exploring the limits

140

Mie Augier

of rationality were themselves limited, for only later did he proceed to construct the broad behavioural program upon which the foundation of the psychology of decision-making in behavioural science could rest. ‘In none of [the] early papers’, wrote Simon, ‘did I challenge the foundations of economic theory strongly’ (1991: 270). The Cowles Commission itself had a large impact on Simon, in particular with regard to his relationship to economics (Simon 1991; Augier and March 2002). Although he had already had a ‘half-baked elementary training’ in economics when enrolling at the University of Chicago in 1933, and had taken several advance courses in economics, without the participation in the Cowles seminars, Simon noted, ‘I would surely not have had a full command of the tools of economic analysis’ (p. 3). Simon used these tools to continue his interest in human decision-making, as witnessed for example in his work on causality (Simon 1952, 1953). But perhaps ‘the greatest impact of the Cowles exposure on me’, Simon says, ‘was to encourage me to try to mathematize my previous research in organization theory and decision making, especially the theory developed in Administrative Behavior’.14 Furthermore, it made Simon’s work visible to economists; in particular, it ‘made Administrative Behavior visible not merely to scholars in the discipline of public administration but to others, as well, who could sense how crucial decision processes are to explaining human rationality. Thus Administrative Behavior did not languish in its provincial homeland, but was noted by economists and decision theorists’ (1991: 116). In particular, Simon mentions three papers that were significant with respect to his influence in economics and his relations to the Cowles Commission. Those were ‘A Behavioral Model of Rational Choice’ (Simon 1955), ‘A Formal Theory of the Employment Relation’ (Simon 1951), and ‘A Comparison of Organization Theories’ (Simon 1952–1953).15 The first paper, ‘A Behavioral Model of Rational Choice’, was written in part at the RAND Corporation where Simon was working in the summers on projects related to organization theory. The paper begins with Simon’s critique of perfect information and unlimited computational capacity in which he wants to replace the assumption of global rationality with an assumption that was more in correspondence with how humans (and other choosing organisms) made decisions, their computational limitations and how they accessed information in their current environments (1955: 99). In Simon’s illustration of the problem, the influence of his early ideas outlined in Administrative Behavior is ever-present, echoing the view that decisions are reasoned and intendedly rational, yet limited (Augier 2001). He first suggests a simple and very general model of behavioural choice which analyses choosing organisms (such as humans) in terms of basic properties to understand what is meant by rational behaviour. He introduces the simplifying assumptions (such as the choice alternatives, the pay-off function, possible future states and the sub-set of choice alternatives which is considered, as well as the information about the probability

H.A. Simon and bounded rationality

141

that a particular outcome will lead to a particular choice) (Simon 1955: 102). But immediately afterwards he turns to the simplifications of this model, stressing that upon careful examination, ‘we see immediately what severe demands they make upon the choosing organism’ (1955: 103). Whereas in models of rational choice, the organism must be able to ‘attach definite payoffs (or at least a definite range of payoffs) to each possible outcome’ (ibid.), Simon suggests that ‘there is a complete lack of evidence that, in actual human choice situations of any complexity, these computations can be, or are in fact, performed’ (ibid.). As a consequence of the lack of computational power, decision-makers have to simplify the structure of their decisions, one of the most important lessons of bounded rationality.16 The other important idea introduced in this paper is that of satisficing. Introducing his behavioural model of rational choice, Simon assumes a simple pay-off function where decision-makers interpret outcomes as either satisfactory or unsatisfactory, and in which an aspiration level constitutes the boundary between satisfactory and unsatisfactory. Whereas decision-makers in rational choice theory would list all possible outcomes evaluated in terms of their expected utilities, and then choose the one that is rational and maximizes utility, decision-makers in Simon’s model face only two possible outcomes, and look for a satisfying solution, continuing to search only until they have found a solution which is good enough. The second paper singled out by Simon as important from his years at Cowles is ‘A Formal Theory of the Employment Relation’ (Simon 1951). This paper was written at Cowles and also reprinted as a Cowles working paper. Furthermore, the paper was in at least one case used as a basis for Cowles selecting their graduate students.17 The paper began by emphasizing the traditional Simon view that models ought to correspond to the empirical realities that are neglected in most economic models of the employment contract (Simon 1951: 293). He then turns to a concept that was so central to him in Administrative Behavior – namely, the concept of authority. Central to the employment relation, Simon says, is the fact that the employer accepts a certain amount of authority of the employee for which he pays a wage and the employee accepts this authority within certain ‘areas of acceptance’ (p. 294). His model applies the idea of satisfaction functions to the employment problem; yet it is still ripe for extensions because it still is ‘highly abstract and oversimplified, and leaves out of account numerous important aspects of the real situation’ (p. 302). The model appears to be considerably more realistic in the way it conceptualizes the nature of the employment relationship; yet it is still about ‘hypothetically rational behavior in an area where institutional history and other nonrational elements are notoriously important’ (p. 302). The model suggests a way to reconcile administrative theory and economics through the economic nature of the employment relation; yet it is still limited by the ‘assumptions of rational utility-maximization behavior incorporated in it’

142

Mie Augier

(p. 305). Thus, Simon used the framework of economics (however limited it might be) to discuss an issue he had been interested in since his thesis, and he concluded his analysis by pointing out the limitations of a constrained model and the necessity of also accounting for non-rational elements. The third paper that Simon mentions as growing out of his Cowles experience was ‘A Comparison of Organization Theories’ (Simon 1952–1953). He compared the economic theory of the firm with organization theory. According to Simon, traditional theories of the firm (‘Ftheory’) are marked by the emphasis on production functions and the maximization of profits, whereas organization theory (‘O-theory’) is recognized by its concern with the conditions for organizational survival (Simon 1952–1953). Faced with issues of organizational behaviour and with empirical facts of the process of decision-making in organization, Simon appeals to those institutionalist economist and organization theorists ‘who have perhaps been in closer contact than traditional economic theorists with the empirical phenomena’. In addition to those three papers, Simon wrote a fourth paper during those years which deserves mention. This is ‘Rational Choice and the Structure of the Environment’ (1956) in which Simon introduced the idea that the environment influences decision-making as much as informationprocessing abilities do. He examines the influence of the structural environment on the problem of ‘behaving approximately rationally, or adaptively’ in particular environments (1956: 130). Simon would later elaborate these ideas in his book The Sciences of the Artificial, using the famous ‘ant on the beach’ metaphor to illustrate his idea (Simon 1969: 51–53). The ant makes his way from one point to another, using a complex path, the complexity consisting of the patterns of the grains of sand along the way, rather than the internal constraints. Just so with human beings, Simon argues, and states his thesis as: ‘Human beings, viewed as behaving systems, are quite simple. The apparent complexity of our behavior over time is largely a reflection of the complexity of the environment in which we find ourselves’ (p. 53). When all is said and done, these papers (with the possible exception of the 1956 paper) did not depart dramatically from (what was then) mainstream economics. However, Simon’s life among the Econ made him realize the importance of pursuing the idea of limited rationality further than was possible with pure economics tools. Thus, to treat the case of limited rationality in a fundamental way, Simon eventually became more and more sceptical towards neoclassical economics and wanted to establish his own research program in behavioural economics. In addition to an unhappiness with the treatment of rationality outlined above, Simon’s dissatisfaction with orthodoxy departed from several generalized disagreements with the way economics was developing at that time. For instance, he did not like the tendency for economics to argue for

H.A. Simon and bounded rationality

143

the use of unrealistic assumptions. The need for a correspondence between the assumptions of a theory and reality was absent. What mattered instead were the predictions the theory could generate. This position, particularly associated with Milton Friedman, was dismissed by Simon as ‘nonsense’ (1963: 230). In particular, in a discussion published in the American Economic Review, Friedman was arguing that no matter what the real nature of decision-making was, it was all right to proceed as if they were maximizing utility because that was consistent with economic data. ‘No one has,’ Simon replied, ‘in fact, observed whether the actual positions of business firms are the profit-maximizing ones; nor has anyone proposed a method of testing this proposition by direct observation’ (1963: 230). He suggested replacing Friedman’s ‘principle of un-reality’ with what he calls the ‘principle of continuity of approximation’ which asserts that given that the conditions of reality approximate sufficiently well with the assumptions of a given ideal type, the derivations we draw from such assumptions will be approximately correct (p. 230). ‘Unreality of premises’, he concluded, ‘is not a virtue in scientific theory; it is a necessary evil – a concession to the finite computing capacity of the scientist that is made tolerable by the principle of continuity of approximation’ (p. 231). Furthermore, through the Cowles Commission seminars, Simon became aware of early work done by John von Neumann and Oskar Morgenstern on game theory. Although at first Simon was enthusiastic about this new theory of games, hoping that it could deliver some new tools to deal with problems in administrative theory, he found game theory in no shape to deal with ideas on limited rationality, because it relied too heavily on ideas of economic man (1947: xxviii).18 In fact, of the many important ideas in game theory, Simon finds that the only ‘common conceptual roof’ under which both economic and administrative man can exist is the idea of representing decision alternatives in a tree of possible future behaviours (p. xxix). Whereas Simon was searching for a theory of how psychological elements can determine decision-making, game theory was searching for techniques to emphasize the predictable behaviour of humans. Similarly, whereas Simon’s starting-point was that organizational and cognitive limitations determined rationality, game theory assumed that there always existed a ‘best’ strategy (and thus decision-makers could be objectively rational). As a result, Simon found that game theory needed to take the decision-maker’s perceptions into account when evaluating rationality (1956). For, Simon says, ‘[t]o predict how economic man will behave we need to know not only that he is rational, but also how he perceives the world – what alternatives he sees, and what consequences he attaches to them’. Thus, as Simon was developing his early ideas on bounded rationality in economics, he also observed that much of the economics that surrounded him wasn’t immediately attracted to his ideas. ‘It was during this period’,

144

Mie Augier

Simon said in looking back, ‘that I began to understand the intensity of economists’ reaction to bounded rationality . . . they were not dealing with it on empirical grounds, but simply rejecting it as irrelevant . . . as they became more strident, so did I.’19 He would soon translate his early ideas on bounded rationality into establishing a new research program which could accommodate the richness of bounded rationality and deal with it on empirical grounds too. While Simon opposed some major developments in economics, he found value in the emerging field of operations research. Although Simon’s marriage with operations research was neither entirely happy, nor permanent, the fact that operations research was well suited to cross disciplinary boundaries immediately appealed to him – in addition to its appeal to the use of computers for heuristic programming.20 Although most of the techniques used in operations research are techniques of constrained maximization, Simon found that they ‘formed a natural continuity with my administrative measurement research’ (1991: 108); and he found Artificial Intelligence to be the logical next step in operations research, something which would eventually bring Simon’s insight on behavioural economics and organization theory to bear in management science, using empirical studies in decision-making in organizations, constructing a mathematical model of the process under study, and then simulating it on a computer (Simon 1965). Simon’s interest in operations research is also evident in his work on the design of optimal production schedules, something which ultimately led to the book, Planning Production, Inventories, and Work Force. Although initiated at the Cowles Commission, this worked was carried out at Carnegie Mellon University, which provided the context for most of Simon’s academic life. It was also at Carnegie that it became clear that Simon was not ‘just’ another economist. Highly respected among most (if not all) distinguished economists of his time,21 Simon himself was much more than an economist. For instance, at Carnegie he quickly re-tooled himself as an organization theorist in order to carry out, with James G. March, a major Ford Foundation study on theories of decision-making in organizations. Most important, at Carnegie Simon found both colleagues and an environment which could accommodate and appreciate his broad interests and honour his vision to cross disciplinary boundaries in pursuing his vision. With the emergence of a behavioural science emphasis at Carnegie came many contributions of cross-disciplinary and interdisciplinary nature. The disciplinary boundary crossing that had been, if not difficult, then different from the mainstream before, became possible and wider spread with the behavioural research focus that Simon helped establish at Carnegie. Having found during his years at Chicago the limits of standard economic theory for dealing with limits to rationality, he turned his attention toward founding a research program in behavioural economics to accommodate his vision.

H.A. Simon and bounded rationality

145

Simon and the early evolution of behavioural economics at Carnegie Arriving at Carnegie in 1949, Simon worked (with William Cooper and George Lee Bach) to build up a promising new business school, namely the School of Industrial Administration (which later became known as the Graduate School of Industrial Administration, GSIA). Business education at that time wasn’t much oriented towards research, but Simon and colleagues wanted to be different. They wanted to do research. They wanted their research to be relevant for business leaders, while at the same time emphasizing the tools of good science (Cooper 2002). Early core courses in the program included ‘quantitative control and business’ (basically consisting of accounting and statistics) taught by Bill Cooper, a sequence of micro and macroeconomics, taught by Lee Bach, and organization theory taught by Simon. As a result of their early efforts to build up a research program at Carnegie Mellon, GSIA was picked by the Ford Foundation as one of the foremost places where the new science of behavioural economics could be developed. GISA became pioneering for the establishment of business education in the United States and must be seen as part of the Simon legacy, perhaps as important as his direct intellectual contributions (Kreps forthcoming). The Ford Foundation had at that time formulated a program for ‘the study of man’ (which became known as ‘the behavioral science research area’), the specific objective of which was stated as follows: ‘The Ford Foundation will support scientific activities designed to increase knowledge of factors which influence or determine human conduct, and to extend such knowledge for the maximum benefit of individuals and of society.’22 Research had to be scientific; embodied in the Ford Foundation’s understanding of the behavioural science concept was ‘its emphasis upon the scientific approach to problem solution’ (p. 4). And it had to be practical, to some extent at least, given the foundation’s interest not in knowledge per se, but in ‘knowledge which promises at some point to serve human needs’. Furthermore, it explicitly encouraged interdisciplinary research. ‘The program is interdisciplinary and inter-field. Its goal is to acquire and apply knowledge of human behavior, and segments of all fields and disciplines will make contributions in varying degrees.’ It is clear that this resonates well with Simon’s vision, and, not surprisingly, he became an adviser to the Ford Foundation research area on the behavioural sciences (Simon 1991: 170–171). Along with other distinguished scientists such as Paul Lazarsfeld, George Miller and Allen Wallis, Simon helped the behavioural science division of the Ford Foundation, answering such questions as ‘what kind of mathematical knowledge is desirable in the training of the behavioral scientist?’; ‘at what level [graduate or undergraduate] and in which areas is mathematical training most desirable?’; and ‘at which institutions should such programs be developed?’.23 And for

146

Mie Augier

years he was very active in advising the Ford Foundation towards developing a mathematically sophisticated but broad social science, emphasizing such things as the development of sophisticated statistical tools as well as a general mathematical knowledge.24 And despite occasional disagreements,25 Simon remained close to the Ford Foundation for many years. In the early 1950s, Simon worked for the Foundation on several projects. For instance, he was part of an ‘Advisory Group on Economics and the Behavioral Sciences’ which was initiated to consider the question whether and how the Ford Foundation should attempt to establish more fruitful relations between economics and the behavioural science. After an initial meeting in New York, smaller one-day conferences were held in San Francisco, Chicago and New York, and participants included such distinguished names as Armen Alchian, Kenneth Arrow, Kenneth Boulding, George Katona, Jacob Marschak, Charles Lindblom, Franco Modigliani, Oskar Morgenstern and, of course, Simon.26 From the outset there was ‘unanimous agreement’ that cooperation between economics and the behavioural sciences was ‘highly desirable’.27 The potential benefits of such cooperation were mutual: ‘such cooperation would be valuable both in the development of new theories about economic behavior and in “mopping up” operations designed to document and refine earlier theories. On the other side, it would be valuable in providing new data and new techniques for the behavioral science.’ Regretting the present situation with the historical separation between the disciplines, the ‘pecking order’ of the social sciences, with economics on the top, was a potential block for interdisciplinary projects and needed to be de-emphasized. Further, collaboration between the disciplines should be seen as more than merely applying the ‘other sciences’ to economic problems; it should also include economic phenomena as one area out of which a general theory of behaviour could be developed (p. 2). Bringing economics to genuine collaboration with the other disciplines, the field of behavioural economics was born. In the early 1950s it was time for Simon to translate the visions from the Ford Foundation into concrete projects, and the place to do this was Carnegie. It was the right time to do this – not too early or too late, for Carnegie Mellon University to let young Simon play a major role and lead major projects – and with the Ford Foundation support there was no shortness of funding. But more than administrative and monetary support would be needed to carry out the vision in Simon’s mind, for he needed a larger group of colleagues with similar interests. In addition to seeking innovative economists such as Franco Modigliani, Richard Cyert, Charles Holt and Merton Miller, Simon and his colleagues sought to build a group at Carnegie that would combine mathematics and quantitative techniques with organizational sociology and social psychology. Thus he hired such scholars as James G. March, Harold Guetzkow and Allan Newell. The group at Carnegie soon consisted of many talented young scholars who

H.A. Simon and bounded rationality

147

were all eager to contribute to this newly formed vision of behavioural science. The spirit at Carnegie was that everybody interacted with everybody else, discussed each other’s research and discussed science, so collaborative teams worked together as well as across each other’s projects. Consisting of different people with different interests, these teams always worked together in a friendly way, despite different disciplines and despite varying degrees of admiration for the idea of rationality.28 The faculty at Carnegie worked together around a number of research projects. Simon was involved in many of them, such as a study done for the Controllership Foundation on the comparative merits of decentralization and centralization in budgeting, a study for the Office of Naval Research on uncertainty and decision-making, and an ‘inventory’ of organization theory funded by the Ford Foundation which led to the book Organizations (1958), co-authored with March with the assistance of Harold Guetzkow (Simon 1991). Each of these projects were different in content, but they were all carried out in the context of Carnegie Mellon, and they all bear the fingerprints of Simon’s vision for understanding decisionmaking (Augier and March 2002). Particularly important for the purpose of this chapter is Simon’s involvement in the Ford Foundation project leading to the book Organizations (March and Simon 1958), for it was here (and in follow-up projects) that the behavioural vision was realized and the seeds to the field of organization theory were planted. In addition to filling a need in the establishment of the behavioural sciences, research on organizations became the emergent discipline of business school education, bringing together different disciplines in the study of decision-making and behaviour in organizations. Furthermore, it significantly influenced major developments in (organizational) economics – in particular Oliver Williamson’s transaction cost theory. Having been a student at Carnegie in those early days, Williamson thus reflected: ‘The selective joinder of organization theory with economics, as these two bear on the theory of the firm, is what motivated my dissertation . . . and describes much of what I have been up to since’ (Williamson, forthcoming: 3). This wasn’t just pure luck; it was part of Simon’s overall vision and long-term strategy to outline a long-range plan for developing a research program on behaviour in organizations at GSIA and to integrate it into the problems he had detected in economics, particularly with regard to the theory of business firms and decision-making in organizations. The administrative strategy was detailed in 1953.29 Finding that the GSIA provided an ‘especially appropriate environment for [organizational] research’, Simon was ready to identify research problems in this area that needed intensive work and also to conduct empirical studies of business organizations in order to give the methodological discussions a more concrete and tangible focus – and in addition to outline a long-term research program into organizational behaviour.

148

Mie Augier

It comes as no surprise that the initial approach of the research program came from the theoretical framework developed in Administrative Behavior. Following the decision-making framework from the dissertation, then, Simon wanted the central concern of the organizational research program to be ‘how people in organizations, particularly at supervisory and executive levels, make choices among alternative courses of action, and how organizational changes affect their choices’, the general hypothesis of the research program being that organizations influence individual behaviour by providing an important part of the environment in which they have to make decisions (pp. 2–3). And in order to understand such decisions, one must examine and understand not only the characteristics of the decision-maker and his relations to the groups, organizations and institutions with which he has daily contact but also the environment constituted by interactions within the whole business organization. Thus, the ‘research will focus on the specific set of influences that we would describe as “the organization” rather than on either the broader social influences or the narrower small-group and individual influences’ (p. 3). Given this focus, three aspects of the organizational decision-making processes in particular deserved attention. First, it was important to understand the roles of formal and informal organizational aspects, and the interaction between them, both with regard to the communication channels through which decision premises are transmitted, and with regard to the development of attributes of the decision premises such as legitimacy, loyalty, goals, identification, etc. The main idea here was that formal organizational influences were significant in establishing initial conditions in organizations, which in turn may determine possible equilibria of the system of organizational interactions. Second, relations between decisionmaking processes and learning had to be explored. Because of bounded rationality, decision-makers see only part of the world and they simplify their decision situation to simpler situations which then serve as their frames of reference. Finding an intimate connection between organizational structure and the learning of frames of references and roles by organizational members, the relationship was seen to be going both ways. Frames of reference would influence what system of formal and information interactions would emerge in the organization. At the same time, if a particular pattern of interaction would be maintained over a period of time, it was possible to predict with some certainty the kinds of roles and frames that the decision-makers would come to adopt. Third, it was important to understand what would happen when an individual’s definition of a particular situation necessitates information for making the decision which he doesn’t have. This is, of course, the problem of uncertainty, greatly simplified in rational models which only consider how a ‘rational’ man would behave facing uncertainty. In contrast to this, the project would examine how individuals and organizations do in fact behave in the face of uncertainty.30

H.A. Simon and bounded rationality

149

This line of thinking explains why Organizations was an inventory of organization theory which used insights from organization theory and social psychology, while also praising mathematical and statistical tools. Intended to generalize insights on organizations as well as supply empirical evidence on those, this project expanded and elaborated Simon’s earlier ideas on behavioural decision-making, search, and aspiration levels. It also elaborated Simon’s ideas on the significance of organizations as social institutions in society and the role of organizations in constraining decision-making. ‘The basic features of organization structure and function’, March and Simon wrote, ‘derive from the characteristics of rational human choice. Because of the limits of human intellective capacities in comparison with the complexities of the problems that individuals and organizations face, rational behavior calls for simplified models that capture the main features of a problem without capturing all its complexities’ (March and Simon 1958: 151). The Ford Foundation subsequently supported a larger project on behavioural theories of organizations which was carried out by Richard Cyert and James March (along with their students, including Julian Feldman, Edward Feigenbaum, William Starbuck and Oliver Williamson). The project originated in the works of Cyert and March to developed improved models of oligopoly pricing by using organization theory (March and Cyert 1955; Cyert and March 1956) and was formally designated as a project area, with Cyert and March as the senior personnel, in 1957. The research aimed at investigating how the characteristics of business firms as organizations affect important business decisions. Integrating theories of organizations with existing (mostly economic) theories of the firm, they developed an empirical theory rather than a normative one, and focused on classical problems in economics (such as pricing, resource allocation, and capital investment) to deal with the processes for making decisions in organizations. This project led to a series of papers, and resulted in another book, A Behavioral Theory of the Firm (Cyert and March 1963). This work extended the earlier ideas, particularly by elaborating concepts of organizational slack, adaptive aspirations, organizational learning, and the role of rules and routines. The book furthermore addressed a major dilemma of organization theory: the choice between a realistic, but unmanageable theoretical model of organization, and a simple, manageable one. Using computer simulations, a relatively realistic description of actual processes was sought, without losing the predictive power so essential to empirical testing (Cyert et al. 1959). This was behavioural economics at Carnegie in the 1950s and early 1960s; and Simon was right in the centre of it, making the research program his intellectual home and a platform from which he would proceed to articulate ideas in areas such as cognitive psychology and artificial intelligence.31 For Simon, behavioural economics meant doing science

150

Mie Augier

in an interdisciplinary way, linking economics to organization theory, cognitive science, sociology and psychology, and centring around concepts such as uncertainty, ambiguity, norms, routines, learning, and satisficing. Emphasizing the concern with the empirical validity of assumptions, Simon thus found that ‘behavioral economics is best characterized not as a single specific theory but as a commitment to empirical testing of the neoclassical assumptions of human behavior and to modifying economic theory on the basis of what is found in the testing process’ (1987: 278). He included in the behavioural economics different approaches such as New Institutional Economics, Evolutionary Economics, and the literature on heuristics coming from Kahneman and Tversky.

Closing This chapter has described the early evolution of Simon’s ideas on bounded rationality and his development of behavioural economics. Along the way, Simon’s early contributions to, and relationship with, economics, have been described, in addition to some of his contributions to political science. It was those experiences which made Simon realize the significance of developing a behavioural program, which he did (along with others) at Carnegie Mellon University in the 1950s and early 1960s. Through the next decades of his life, Simon witnessed many fruits of his work. He was proud of all the intellectual children and grandchildren that his research program had fostered – ranging from ‘Post Simonian Evolutionary Economics’, to influencing transaction cost economics, to cognitive science and artificial intelligence (Dosi 2002; Williamson 2002a, 2002b; Newell 1989; Feigenbaum 1989); he saw all these developments as broadly consistent with many ideas he had developed. ‘Behavioral economics,’ he wrote, ‘with its emphasis on the factual complexities of our world, has not always appeared as attractive as the axiomatized certainties of [neoclassical economics] . . . I am encouraged, at the present time, by what I perceive as a strong ground swell of interest in behavioral economics’ (Simon 1986: 275–276). As the world of the disciplines had evolved, elements also emerged that Simon did not expect – most significant, perhaps, rational expectations theory in economics and connectionism in artificial intelligence – but in time he came to see them mostly as healthy rivals during the transition to his behavioural ideal. As a young man, so strong was the ideal in Simon’s mind that he thought its arrival was immanent, if not already implicit in the real world and that science would have to adapt quickly. As he aged, he became more and more aware of the persistence of deviations from this ideal, and his forecast date for the arrival of the full behavioural ideal retreated even further into the future. Simon died recognizing that the victory of behavioural science was yet quite a distance away, but believing it to be still worth while to work toward its arrival.

H.A. Simon and bounded rationality

151

This chapter has outlined the broad paths of Simon’s early vision and his establishment of an interdisciplinary behavioural economics. From the standpoint of the history of economic thought, then, it is indeed curious that the term ‘behavioural economics’ has now, several decades after Simon’s first efforts, made its way into mainstream journals and is being taught at major universities. Does this reflect the fact that Simon’s vision did manage to transform the economics profession within a few decades in a much more ‘behavioral’ direction? Not quite. Instead of mainstream economics becoming more ‘behavioral’, modern behavioural economics is becoming more mainstream. In fact, some among the new wave of behavioural economists don’t even recognize the Carnegie program as part of their background. ‘I call what I do “second-wave behavioral economics” ’, one prominent behavioural economist said. ‘The pioneers were psychologists who came along 15 years ago and convinced economists that some of their assumptions were wrong. Now, people like me are working to carefully adjust the models and formalize the effect.’32 Much unlike Simon, this new wave of behavioural economics wants to employ mainstream economic tools. ‘This research program’, Rabin declares, ‘is not only built on the premise that mainstream economic methods are great, but so too are most mainstream economic assumptions’ (Rabin 2002). Given the use of concepts in the new wave such as bounded rationality, altruism, and behavioural economics (all Simon ideas), it is somewhat surprising that they rarely address the Simon connection, either positively or negatively. Perhaps they are put off by the strong sense of interdisciplinarity that is a legacy of Simon’s earlier advocacy of behavioural economics, or by youthful exposure to (too) much neoclassical economics, or by career opportunities; as a matter of current institutional fact, it may be a better bet to stay closer to the mainstream, but in a better world to come (Simon’s ideal), it might not. Whatever the reason, though they do not explicitly take exception with Simon’s ideas, the new wave is often extremely forthcoming about their closeness to the mainstream.33 The future relevance of Simon’s work will depend on whether further developments in the disciplines (behavioural economics among them) will bring us closer to Simon’s vision or take us further away. This in turn will depend on the ability of the disciplines to work together and for the sciences to accommodate interdisciplinary vision. For in Simon’s mind, this vision was the first step toward realizing the behavioural ideal. Simon himself enthusiastically embraced several (early) developments in the new wave of behavioural economics (in particular the work of Kahneman and Tversky), although he was always eager to warn that a close relationship to orthodoxy wasn’t anything to wish for; and he probably would have taken up the battle on many fronts of argument had he still been with us. But that reflects who Simon was. He liked to fight more than he liked to win. And he was, by heart, a pluralist. ‘Even though Simon often disagreed with those of us who, to his taste, were “too neoclassical” ’, Williamson

152

Mie Augier

recalled, ‘there was always an outside possibility that, in some degree, we might get parts of it right’ (forthcoming: 12). Thus, Simon concluded, ‘[w]ith enough patience and diligence in exploring the empirical realities of economic processes, and enough ingenuity in modeling these processes, we will be able, in time, to return economics to the real world’ (1986: 276).

Notes * This is part of a book-length study forthcoming with The MIT Press, and the present chapter summarizes, very briefly, some arguments made in the first half of this book. I am very grateful to Kenneth Arrow, Edward Feigenbaum, James G. March, Paul Samuelson, Herbert Simon, and Oliver Williamson for taking time answering many questions and commenting on my ideas; and to Philip Mirowski and Salvatore Rizzello for discussions and improvements. Also thanks to the RAND corporation and to Carnegie Mellon University Library for access to archives and for hospitality when doing research for this book. My greatest thanks, finally, to Herbert Simon for so many enjoyable moments, and for his endless inspiration and energy; and to his family, in particular Kathie Simon Frank and Dorothea Simon, for continuous encouragement of this project. 1 He received major awards from many scientific communities, including the A.M. Turing Award (in 1975), the National Medal of Science (in 1986), and the Nobel Prize in Economics (in 1978). He received the James Madison Award, the Dwight Waldo Award and the John Gaus Lecture Award from the American Political Science Association, and the Gold Medal Award for Life Achievement in Psychological Science from the American Psychological Association. He was a Fellow of the Econometric Society and was elected to the American Philosophical Society, the National Academy of Science, and the American Academy of Arts and Sciences, as well as many foreign honorary societies. He was awarded honorary doctorates by many universities in the United States and abroad, including Harvard University, Columbia University, Yale University, and the University of Chicago. 2 For instance, Foss (in this volume) argues that most research in organizational economics uses bounded rationality in a rhetorical manner with little attention to substance. Williamson has described his debt to Simon and his program on several occasions (1996, 2002, forthcoming). In addition to transaction cost economics, much of evolutionary economics builds on Simonian ideas and concepts (Dosi, forthcoming). 3 ‘Not always’ is probably an understatement as Simon had little patience with the way in which game theorists, for instance, use bounded rationality. ‘I have been watching with fascination game theorists and mathematical economists to introduce various kinds of boundedness into the rationality of their actors’, Simon wrote in a letter to Ariel Rubinstein; ‘that the boundedness is all contained in formal models, with little evidence of its relevance to reality, and almost no effort to demonstrate that relevance, tempers my enthusiasm about this trend’ (letter to Ariel Rubinstein, 2 December 1996. Herbert A. Simon papers, Carnegie Mellon University Library). 4 Interview with Herbert A. Simon, 19 December 2001. 5 Simon’s work in the policy institutions indicates that his significance goes much beyond his written work and pure research, and he used his recognition in such circles to push his vision for social and behavioural science further. For instance, while Simon was the first social scientist to be elected to the National

H.A. Simon and bounded rationality

6

7 8 9 10

11

12

13 14

15 16

153

Academy of Science, he, once in, immediately worked on getting some close colleagues elected (Kenneth Arrow, Tjalling Koopmans, Paul Samuelson, among others). Similarly, while Simon was serving on PSAC during the Johnson and Nixon years as an expert in AI, he worked for getting more social scientists on board in areas (such as defence policy) where the natural sciences had been dominating. While Simon’s career proceeded to other areas later, my story will not develop this aspect here, for the purpose of this chapter is to describe Simon’s early vision and the early evolution of bounded rationality, and how he increasingly saw the need to bring more cognitive elements to bear in the understanding of human decision-making. A good introduction to Simon’s early work in psychology and AI is McCorduck (1979). Among the classes he did take were mathematical economics (with Henry Schulz); logic (with Rudolph Carnap) and mathematical biophysics (with Nicolas Rashevsky). Herbert A. Simon (1935): ‘Administration of Public Recreational Facilities in Milwaukee’, Unpublished manuscript (Herbert A. Simon papers, Carnegie Mellon University Library). This point anticipates his later insights on organizational identification (1997). Also see Simon (1988: 286): ‘The big problem for me, from the very beginning of my work with Clarence Ridley, was to reconcile the way that decisions were actually made in organizations with the way that the economists pretended that they were made.’ The first draft of the thesis, written as a paper for Carnap’s class, discussed both Aristotle, the concept of ‘economic man’ and the construction of administrative science upon the concept of efficiency – so his early interests were very clear. See H.A. Simon, ‘The Logical Structure of an Administrative Science’ (Herbert A. Simon papers, Carnegie Mellon University Library). Many ideas central to Simon’s later work can be found in Administrative Behavior (Simon 1988, 1991; Augier and March 2002). For instance, he has later taken up the theme of organizational identification, as discussed in Administrative Behavior, and contrasted it to the dominating idea of the presence of opportunism and conflict of interest in organizations (Simon 1997). Moreover, his ideas about the hierarchical organization of problem-solving that first were hinted at in Administrative Behavior became central to his later work. The introduction of the idea of a hierarchy of decisions, where one stage in the hierarchy influences the next step, and so on, found use in Simon’s theory of ‘The Sciences of the Artificial’ as well as his work in computer science, seeing the underlying structure of computational memory as hierarchical (Simon 1977). His work on problem-solving also viewed problems as generally decomposable into hierarchical structures (Simon 1989). Thanks to Axel Leijonhufvud for letting me borrow his (1981) phrase. ‘Inquiry into the Cowles Commission’ (Manuscript, Herbert Simon papers, Carnegie Mellon University archives: 4). Simon continues, emphasizing the importance of his early work on his career: ‘I think this project was on the agenda anyway, but the Cowles contact certainly egged me on it and gave it higher priority.’ The following sections build on Augier and March (2002). As Simon explains: ‘For the first consequence of the principle of bounded rationality is that the intended rationality of an actor requires him to construct a simplified model of the real situation in order to deal with it. He behaves rationally with respect to this model, and such behavior is not even approximately optimal with respect to the real world. To predict his behavior, we must understand the way in which this simplified model is constructed, and its con-

154

17

18

19 20

21 22 23

24 25

26 27

Mie Augier

struction will certainly be related to his psychological properties as a perceiving, thinking, and learning animal’ (Simon 1957: 199). As the economist Roy Radner reports, ‘[m]y first contact with Herbert Simon’s work occurred when I was a graduate student (in statistics) at the University of Chicago. Simon had written a paper for the Cowles Commission [Simon 1951] . . . and it was the Cowles Commission’s practice to have its papers “refereed” before they were circulated externally. For this ostensible purpose, I was asked to read Simon’s paper. As I remember it, I was being considered for a research assistantship at the Commission, and I suspect that the paper was given to me more as a test of my own abilities than anything else’ (Radner, forthcoming). See Simon’s friendly review of von Neumann and Morgenstern in Simon (1945). In this review he toned down the differences between his own views and those of game theory, because he wanted his audience to learn from the book (the review appeared in a major sociology journal). Said Simon: ‘I thought that, because of the importance of the book and the issues it raised, my goal in the review should be to get the attention of sociologists so that they, at least the younger ones, would become aware of this topic and its implications for rationality in social situations’ (personal email message from Simon, 2 February 1999). Personal email message from Simon, 2 February 1999. Thus, Simon and Newell wrote in 1958: ‘Even while operations research is solving well-structured problems, fundamental research is dissolving the mystery of how humans solve ill-structured problems. Moreover, we have begun to learn how to use computers to solve these problems, where we do not have systematic and efficient computational algorithms. And we now know, at least in a limited area, not only how to program computers to perform such problem-solving activities successfully; we know also how to program computers to learn to do these things’ (p. 6). See, for instance, Paul Samuelson (forthcoming): ‘Simon’s many published works were part of my inventory of knowledge, as they were for any avant garde theoretical economist of the post-1933 years.’ ‘The Ford Foundation Behavioral Science Program: Proposed Plan for the Development of the Behavioral Sciences Program’ (1951, Herbert A. Simon papers, Carnegie Mellon University Library). ‘Memorandum to Participants of Meeting on the Mathematical Training of Behavioral Scientists’, The Ford Foundation, Behavioral Science Division, 11 September 1952 (Herbert A. Simon papers, Carnegie Mellon University Library). Herbert A. Simon to Bernard Berelson, ‘Memorandum: Mathematical Training of Behavioral Scientists’, 13 September 1952, p. 1 (Herbert A. Simon papers, Carnegie Mellon University Library). Simon’s archive material suggests that although he was a long-time adviser to the Ford Foundation, he did not always agree with their agendas. For instance, Simon was strongly opposed to the establishment of what became the Center for the Advanced Study in Behavioral Sciences, finding that the Foundation should instead spend money on different, smaller programs that would integrate many different people. See Simon’s correspondence with Bernard Berelson of the Ford Foundation, 1951 (Herbert A. Simon papers, Carnegie Mellon University Library). ‘Report of Three One-day Conferences on “Economics and the Behavioral Sciences” ’ (Herbert A. Simon papers, Carnegie Mellon University Library). ‘The Ford Foundation, Behavioral Science Division: Advisory Group on Economics and the Behavioral Sciences’, July 1952 (Herbert A. Simon papers, Carnegie Mellon University Library).

H.A. Simon and bounded rationality

155

28 Accounts of the atmosphere and intellectual excitement at Carnegie at the time can be found in Williamson (1996, 2002, forthcoming), Modigliani (2001) and Cooper (2002). 29 Memorandum: ‘Research into Behavior in Organizations – Proposed Program’, 28 February 1952 (Herbert A. Simon papers, Carnegie Mellon University Library). 30 Simon summed up the general plan as follows: ‘to examine organization behavior as a decision making process; to formulate, refine, and test propositions about the organizational determinants of decisions; and to explore the implications of the hypothesis that organizational changes can be regarded largely as changes in “set” of organization members. As two special areas of concentration, we propose to study especially the relationship between formal and informal organization in these terms, and to restate organization theory so as to bring it into closer relationship with learning theory; and in both of these areas to test empirically some of the hypotheses that we derive’ (1947: 4). 31 I shall not elaborate these developments here; but note that they were all natural extensions of Simon’s interest in bounded rationality, and he saw all his subsequent work as making the same point using different tools and different languages (Simon 2001; also see Feigenbaum 1989: Newell 1989; Augier and Feigenbaum 2002; Augier and March 2002). 32 This quote from Mat Rabin can be found at: http://www.berkeley.edu/news/ media/releases/2000/06/14_macarthur.html 33 Also note that the new program in behavioural economics are not much interested in organization theory, whereas in the early contributions by Simon and others, the joinder of economics with organization theory is significant (Williamson, forthcoming).

References Augier, M. (2001) ‘Models of Herbert A. Simon’, Perspectives on Science, 8(4): 407–436. Augier, M. and Feigenbaum, E. (2002) ‘Herbert A. Simon: A Biographical Memoir’, forthcoming in Proceedings from the American Philosophical Association. Augier, M. and March, J.G. (2002) ‘A Model Scholar’, Journal of Economic Behavior and Organization, 49: 1–17. Christ, C. (1994) ‘The Cowles Commission’s Contributions to Econometrics at Chicago, 1939–55’, Journal of Economic Literature, 32: 30–59. Cooper, W.W. (2002) ‘Auditing and Accounting: Impacts and Aftermaths of R.M. Cyert’s Research in Statistical Sampling’, in M.-S. Augier and J.G. March (eds), The Economics of Choice, Change and Organization: Essays in Honor of Richard M. Cyert, Cheltenham, UK: Edward Elgar. Cyert, R.M. and March, J.G. (1956) ‘Organizational Factors in the Theory of Oligopoly’, Quarterly Journal of Economics, 70: 44–64. —— (1963) The Behavioral Theory of the Firm, Englewood Cliffs, N.J.: PrenticeHall. Cyert, R.M., Feigenbaum, E. and March, J.G. (1959) ‘Models in a Behavioral Theory of the Firm’, Behavioral Science, 4: 81–95. Dosi, G. (forthcoming) ‘A Very Reasonable Objective Still Beyond Our Reach: Economics as an Empirically Disciplined Social Science’, in M. Augier and J.G. March (eds), Models of a Man: Essays in Memory of Herbert A. Simon, Cambridge, Mass.: The MIT Press.

156

Mie Augier

Feigenbaum, E. (1989) ‘What Hath Simon Wrought?’, in D. Klahr and K. Kotovsky (eds), Complex Information Processing: The Impact of Herbert A. Simon, Hillsdale, N.J.: Lawrence Erlbaum Associates Publishers, 165–182. Kreps, D. (forthcoming) ‘Beliefs and Tastes: Confessions of an Economist’, in M. Augier and J.G. March (eds), Models of a Man: Essays in Memory of Herbert A. Simon, Cambridge, Mass.: The MIT Press. Leijonhufvud, A. (1981) ‘My Life Among the Econ’, in A. Leijonhufvud (ed.), Information and Coordination: Essays in Macroeconomic Theory, Oxford: Oxford University Press, 347–359. McCorduck, P. (1979) Machines Who Think, San Francisco: Freeman. March, J.G. and Cyert, R.M. (1955) ‘Organizational Structure and Pricing Behavior in an Oligopolistic Market’, The American Economic Review, 45(1): 129–139. March, J.G. and Simon, H.A. (1958) Organizations, New York: Wiley. Mirowski, P. (2002) Machine Dreams, Cambridge: Cambridge University Press. Modigliani, F. (2001) Adventures of an Economist, New York: Texere Publishers. Mullainathan, S. and Thaler, R. (2000) Behavioral Economics, Working paper No. 00-27, MIT Department of Economics. Newell, A. (1989) ‘Putting It All Together’, in D. Klahr and K. Kotovsky (eds), Complex Information Processing: The Impact of Herbert A. Simon, Hillsdale, N.J.: Lawrence Erlbaum Associates Publishers, 399–440. Rabin, M. (1998) ‘Psychology and Economics’, Journal of Economic Literature, XXXVI (March): 11–46. —— (2002) ‘Psychology and Economics’, Working paper No. 97–251, Berkeley Department of Economics, European Economic Review, 46: 657–685. Radner, R. (forthcoming) ‘The Best is the Enemy of the Good’, in M. Augier and J.G. March (eds), Models of a Man: Essays in Honor of Herbert A. Simon, Cambridge, Mass.: The MIT Press. Ridley, C.E. and Simon, H.A. (1938) Measuring Municipal Activities, Chicago: International City Manager’s Association. Samuelson, P. (forthcoming) ‘The Hawkins–Simon theorem revisited’, in M. Augier and J.G. March (eds), Models of a Man: Essays in Memory of Herbert A. Simon, Cambridge, Mass.: The MIT Press. Simon, H.A. (1943) ‘Administrative Behavior’ (Dissertation version), University of Chicago, Unpublished manuscript. —— (1945) ‘Review of the book Theory of Games and Economic Behavior by J. von Neumann and O. Morgenstern’, American Journal of Sociology, 50: 558–560. —— (1947) Administrative Behavior, New York: Free Press. —— (1951) ‘A Formal Theory of the Employment Relation’, Econometrica, 19: 293–305. —— (1952) ‘On the Definition of the Causal Relation’, The Journal of Philosophy, 49: 517–528. —— (1952–1953) ‘A Comparison of Organization Theories’, Review of Economic Studies, 20(1): 40–48. —— (1953) ‘Causal Ordering and Identifiability’, in W.C. Hood and J.C. Koopmans (eds), Studies in Econometric Method, New York: Wiley, 49–74. —— (1954) ‘Some Strategic Considerations in the Construction of Social Science Models’, in P. Lazarsfeld (ed.), Mathematical Thinking in Social Sciences, Glencoe, Ill.: Free Press.

H.A. Simon and bounded rationality

157

—— (1955) ‘A Behavioral Model of Rational Choice’, Quarterly Journal of Economics, (69): 99–118. —— (1956) ‘Rational Choice and the Structure of the Environment’, Psychological Review, (63): 129–138. —— (1957) Models of Man, New York: Wiley. —— (1963) ‘Discussion: Problems of Methodology’, American Economic Review, 53: 229–231. —— (1965) The Shape of Automation (for Men and Management), New York: Harper & Row. —— (1969) The Sciences of the Artificial, Cambridge, Mass.: The MIT Press. —— (1977) Models of Discovery, Dordrecht: Reidel. —— (1986) ‘Preface to “Handbook of Behavioral Economics” ’, Reprinted in Models of Bounded Rationality, vol. 3, Cambridge, Mass.: The MIT Press. —— (1987) ‘Behavioral Economics’, Reprinted in Models of Bounded Rationality, vol. 3, Cambridge, Mass.: The MIT Press. —— (1988) ‘Nobel Laureate Simon “looks back”: A Low-frequency Mode’, Public Administration Quarterly, 12: 275–300. —— (1989) ‘The Scientist as Problem Solver’, in D. Klahr and K. Kotovsky (eds), Complex Information Processing: The Impact of Herbert A. Simon, Hillsdale, N.J.: Lawrence Erlbaum Associates Publishers, 375–398. —— (1991) Models of My Life, Cambridge, Mass.: The MIT Press. —— (1993) ‘Altruism and Economics’, American Economic Review, 83(2): 156–161. —— (1997) An Empirically Based Microeconomics, Cambridge: Cambridge University Press. —— (2001) ‘On Simulating Simon: His Monomania, and Its Sources in Bounded Rationality’, Studies in the History and Philosophy of Science, 32(3): 501–505. Simon, H.A. and Newell, A. (1958) ‘Heuristic Problem Solving: The Next Advance in Operations Research’, Operations Research, 6: 1–10. Williamson, O.E. (1996) ‘Transaction Cost Economics and the Carnegie Connection’, Journal of Economic Behavior and Organization, 31: 149–155. —— (2002) ‘Empirical Microeconomics: Another Perspective’, in M.S. Augier and J.G. March (eds), The Economics of Choice, Change and Organization: Essays in Honor of Richard M. Cyert, Cheltenham, UK: Edward Elgar. —— (forthcoming) ‘Herbert Simon and Organization Theory: Lessons for the Theory of the Firm’, in M. Augier and J.G. March (eds), Models of a Man: Essays in Honor of Herbert Simon, Cambridge, Mass.: The MIT Press.

6

The rhetorical dimensions of bounded rationality Herbert A. Simon and organizational economics Nicolai J. Foss

Introduction1 In this chapter I make a couple of connected arguments about the status of bounded rationality (henceforth, “BR”) in modern economics and the role of Herbert Simon with respect to understanding this status. Following McCloskey (1983), emphasis is placed on the rhetorical aspects, the attempts to persuade, of scientific development in economics. Following Sent’s (1997) fine study of Thomas Sargent’s (rhetorical) appeal to BR, economists’ actual use of BR is examined. The economists whose use of BR I consider are economists of organization. This choice is far from arbitrary. In his key papers directed at an economics audience (Simon 1978, 1979), Simon made several explicit references to the emerging economics of organization and he himself contributed fundamentally to the neighboring field of organization theory. His examples of BR and its implications usually involved the business firm. Indeed, he sometimes took the notion of “administrative man” to be synonymous with a boundedly rational agent. Perhaps not surprisingly, then, the economics of organization was probably the first sub-field of economics where BR was systematically invoked, and it is perhaps still today the sub-field being invoked with the highest frequency.2 More specifically, it will, first, be argued that the use of BR arguments in the economics of organization is primarily rhetorical in the somewhat pre-McCloskeyan (McCloskey 1983) sense of dressing up a theory with arguments that are essentially empty in an explanatory sense, but are nevertheless made because they help to persuade. Specifically, BR is invoked in the rhetorical practice of organizational economists, because it represents a way of conveying the intuition of another, much more central point; however, it is not invoked because it itself is in any way central. Specifically, BR is used in order to explain in a loose, background way the notion of contractual incompleteness. Thus, if people do not have the wits to imagine and make contractual provision for a number of contingencies, they will leave the contract incomplete, giving rise to the kind of externality problems that drive the modern economics of organization. However,

H.A. Simon and organizational economics

159

asymmetric information (a well-defined concept) can do, in these models, what BR (a concept with a multitude of different connotations, but arguably no clear definition) supposedly does, and can do so more “cleanly”. This raises the question why BR continues to be invoked, leading to the second argument. Second, I argue, admittedly in a more speculative vein, that there are strong reasons to suspect that the status, as well as the rhetorical practice, of Herbert Simon plays a decisive role for understanding the way in which BR is presently used. The fact that a Nobel Prize was bestowed upon Simon for his work on BR, and that it was the brainchild of one of history’s more impressive polymaths, should make it hard for almost anyone to simply dismiss it.3 Arguably, however, the rhetorical practice of Simon made it hard for most economists to not in the end dismiss it, if perhaps only discreetly; namely, by tacitly refusing to incorporate it in their theoretical work. Here, rhetorics is understood in a more authentic manner as the “art of discovering good reasons, finding what really warrants assent, because any reasonable person ought to be persuaded” (Booth 1974: xiv; quoted by McCloskey 1983: 482). Thus, although Simon certainly wrote as if “any reasonable person ought to be persuaded” by his arguments, he did not, I argue, give the “good reasons,” those that would “really warrant assent,” for BR. He failed to persuade. Most importantly, in his most “rhetorical” papers, directly aimed at persuading economists, Simon never provided any precise definition of what BR really is. Moreover, he never really gave good heuristic advice on how to incorporate BR into economic models, and simply noted that how exactly notions of BR developed in the theory of computational complexity – his favorite example of successful modeling of BR – would be incorporated into economic models “remain[s] to be seen” (1978: 12). These failures form part of the explanation of why most economists, including economists of organization, so far have not fundamentally taken BR seriously. Some implications and wider implications, notably with respect to the future of BR in the economics of organization, are finally briefly discussed. A conclusion is that there is reason to be more optimistic on behalf of the use of BR in economic reasoning, since there are now better critical and constructive foundations for BR than when Simon tried to make economists take BR seriously, as partly signaled by the publication of the present book.

160

Nicolai J. Foss

Economists’ use of bounded rationality: the case of organizational economics Bounded rationality in organizational economics In his autobiography, Simon (1991a: 270–271) recounts increasingly violent disagreements with mainstream economists, leading him to abandon, for a period, economics in favor of psychology and computation science. “By the time I returned to a concern with economics in the 1970s,” he observed, “the war was open and declared.” Indeed, many of Simon’s writings from that period (Simon 1976, 1978, 1979) are so sharply formulated that it seems quite likely that at least Simon himself felt that he was part of a war.4 (It is actually harder to find specific and concerted critique, at least in print, of Simon and BR, such as one would expect of a genuine war.) However, at exactly the time when the supposed war was going on, an important part of economics, namely the theory of the firm, seemed to be increasingly influenced by considerations of BR. New, serious approaches to various aspects of the theory of the firm that all appeared to be solidly based on bounded rationality were mushrooming. Thus, team theory (Marschak and Radner 1972), transaction cost economics (Williamson 1971), and the evolutionary theory of the firm (Nelson and Winter 1973) all appeared in the beginning of the 1970s (although their roots go further back). These, still flourishing, approaches all seemed to start from bounded rationality,5 exactly as Simon would like them to. And the explicit motivation for such a starting point was that neoclassical theory of the firm and its behavioral starting point in substantive rationality excluded concern with such vital phenomena as incomplete contracts, the role of organizational structure and organizational routines. Today, many – if not all – economists of organization would likely agree that BR is important to the study of economic organization (Milgrom and Roberts 1988, 1992). Indeed, some argue that it is indispensable; that is, a necessary assumption in the theory of economic organization (Williamson 1996; McLeod 2000). References to the need to draw more on psychological research for understanding the workings of organization are quite common now, even among the economics profession’s foremost symbol manipulators (e.g., Holmström and Tirole 1989; Lazear 1991). Thus, a newcomer to the field may, by glancing at contemporary organizational economics, easily get the impression that Simon’s lessons have been absorbed, and that organizational economists have acknowledged the need to place BR centerstage in their theorizing. This is not the case. If anything, the use (or invocation) of BR may have declined.6 To some extent this is because the mainstream economics of organization has developed into a highly formal and axiomatic enterprise, and BR has a bad reputation of only being given to formalization if that formalization is

H.A. Simon and organizational economics

161

fundamentally ad hoc and the axiomatic basis is unclear or non-existent. That reputation may not be entirely justified (Rubinstein 1998), but many economists of organization (particularly contract theorists) certainly act as if it is. Oliver Hart arguably sums up the attitudes of many formal economists when he argues that I do not think that bounded rationality is necessary for a theory of organizations. This is fortunate because developing a theory of bounded rationality in a bilateral or multilateral setting seems even more complicated than developing such a theory at the individual level; and the latter task has already proved more than enough for economists to handle. (Hart 1990: 700–701) In fact, some parts of the economics of organization, particularly contract theory, bear little substantial imprint of BR.7 This is not surprising: contract theory is based entirely on information economics and game theory, which at least in their standard, “toolbox versions” have no room for BR at all. Still, even contract theorists occasionally invoke BR, usually to explain in a loose way why some contingencies may be left out of a contract. Bounded rationality and transaction cost economics It is sometimes argued that transaction cost economics provides considerably more room for BR than contract theory (e.g., Brousseau and Fares 2000). There is something to this claim; for example, Williamson (1975) does invoke BR in connection with, for example, explaining the M-form, and other aspects of organizational structure. He puts much emphasis on the need for adaptation that arises in a world of uncertainty and bounded rationality. Governance mechanisms in Williamson’s work are more than efficient (i.e., second-best) ex ante allocations of property rights (as in Grossman and Hart 1986; Hart and Moore 1990); they are mechanisms for ex post adaptation and conflict resolution. Williamson’s works are replete with references to Simon. Still, however, Williamson refrains from being explicit about how to model BR on the level of the individual agent. He is quite explicit here, noting that “[e]conomizing on bounded rationality takes two forms. One concerns decision processes and the other involves governance structures. The use of heuristic problem-solving . . . is a decision process response” (Williamson 1985: 46). However, in transaction cost economics, “heuristic problem solving” is not central. Instead, transaction cost economics “is principally concerned . . . with the economizing consequences of assigning transactions to governance structures in a discriminating way.”8 In other words, Williamson is interested in making use of bounded rationality for

162

Nicolai J. Foss

the purpose of developing a theory of discriminating alignment rather than for the purposes of explaining administrative behavior, as in Simon (1947). He is not interested in BR as a “decision process response.” For the purpose of explaining why contracts are incomplete, Williamson apparently thinks that it is not necessary to model BR itself; it may be asserted as a “background assumption” that while vital – indeed, necessary – does not need to be explicated itself. Milgrom and Roberts (1992: 128), as well as most other mainstream economists of organization who invoke BR, adopt the same procedure. Thus, BR enters organizational economics reasoning in a loose background sort of way, in which it lends credence to exogenously imposing constraints on the feasible contracting space, but is not modeled itself. It supplies the rhetorical function of lending intuitive support to the notion of incomplete contracts. A Simonian information-processing argument is sometimes invoked in order to be more concrete about how BR produces incomplete contracting (Hart 1990: 698; Schwartz 1992: 80): if agents do not have the mental capacity to think through the whole decision tree (for example, in complicated bilateral trading relations), it seems reasonable to assume that some of the branches of the tree (such as those relating to some future uses of assets) cannot be represented in a contract; the contract is left incomplete. This is indeed a BR argument. However, agents are supposed to deal with this manifestation of BR in a substantively rational manner, as numerous critics have pointed out since Dow (1987). As he observed, this approach provokes a lurking suspicion of a basic inconsistency, for whereas BR is loosely invoked as a background assumption (yet still a necessary one), there is no hesitation to appeal to substantive rationality when the choice between governance structures must be explained. I discuss this next. The irrelevance of bounded rationality: the incomplete contract controversy A recent theoretical debate on the coherence and foundations of incomplete contract theory – called the “incomplete contract controversy” (Tirole 1999) – is pertinent to the issues under consideration here. The debate concerns whether satisfactory foundations for incomplete contracts are offered in the works of Hart and associates (e.g., Grossman and Hart 1986; Hart and Moore 1990). The main critics are Eric Maskin and Jean Tirole (Maskin and Tirole 1999; Tirole 1999). At the core of this debate is the explanatory tension between invoking transaction costs – which may be understood as a consequence of BR – on the one hand and postulating farsighted and substantively rational contracting on the other (i.e., the parties to a contract can foresee the utilities from the relation). Whereas Dow (1987) interpreted this as an inconsistency in transaction cost economics, Maskin and Tirole show that there is no formal inconsistency

H.A. Simon and organizational economics

163

here, and that on this point the incomplete contracts literature has got it right. However, the main thrust of their argument is that the use of transaction costs (i.e., BR) in models of incomplete contracting does not provide additional explanatory insight relative to models that make no use of these (i.e., complete contracting models). The obvious implication is that there is no reason for making use of BR, not even in the “homeopathic” (Dosi 2002) way in which it has hitherto been used. Organizational issues have largely motivated the upsurge in incomplete contract modeling during the last decade. In fact, the founding incomplete contract paper, namely Grossman and Hart (1986), was explicitly motivated by an attempt to model the emphasis in transaction cost economics on asset specificity as a key determinant of the scope of the firm, using modeling conventions and insights already developed in (complete contracting) agency theory and its basis in mechanism design theory. However, whereas Williamson (1996) puts much emphasis on inefficient ex post bargaining, the incomplete contracting approach assumes that ex post bargaining is efficient. Thus, what drives these models are misaligned ex ante incentives, particularly with respect to investment in vertical buyer–supplier relationships. The problem is to motivate what may cause such misalignment. The point of contention in the incomplete contracts controversy is whether transaction costs arising from the inability to perfectly anticipate or describe all relevant contingencies or enforce contract terms – all of which may derive from BR9 – constrain the set of feasible contracting outcomes relative to the complete contracting benchmark. If this is not the case, transaction costs (BR) do not suffice to establish the possibility of inefficient investment patterns. Therefore, they do not suffice to establish a role for ownership, and in turn for a theory of the boundaries of the firm.10 The Maskin and Tirole argument builds on the key assumption in the incomplete contract approach that although valuations may not be verifiable, they may still be observable by the parties (Hart and Moore 1990). This implies that trade can be conditioned on message games between the parties. These games are designed ex ante in such a way that they can effectively describe ex post (where bargaining is efficient) all the trades that were not described ex ante. A further crucial step in the argument is the typical contract theory assumption that parties allocate property rights and choose investments so that their expected utilities are maximized, knowing (at least probabilistically) how payoffs relate to allocations of property rights and levels of investment (i.e., they can perform “dynamic programming”). Given this, Maskin and Tirole (1999) provide sufficient conditions under which the undescribability of contingencies does not restrict the payoffs that can be achieved. In other words, there are no differences in the allocations that can be achieved under incomplete contracting and complete contracting; no real economic content is achieved by adding considerations of BR/undescribability of contingencies/transaction

164

Nicolai J. Foss

costs. This is their “irrelevance of transaction costs” theorem. We might as well call it the “irrelevance of bounded rationality theorem.” Bounded rationality and mainstream modeling BR, it has been argued, is very much a background assumption that is introduced in order to help explaining in an “intuitive” way incomplete contracting, a key ingredient in understanding the efficient boundaries of the firm, and efficient alignment more generally. BR is never explicitly modeled on the level of the individual agent. It is never fundamentally taken seriously in the way economists take arguments seriously, namely by modeling them. After Maskin and Tirole (1999), there would indeed seem to be little reason to take BR seriously at all; its use can at best be “rhetorical.”11 However, one may argue that it is exactly the very “thin” way in which BR is treated in organizational economics that allows for the Maskin and Tirole argument, which purportedly demonstrates the complete irrelevance of BR (Kreps 1996; Foss 2001). After all, a main notion in incomplete contract theory and transaction cost economics is that a very thin Simon may join hands with a rather corpulent Savage, as it were, and it is not really surprising if the very thin Simon turns out to matter very little indeed for the explanatory weight of the whole construct. Still, the puzzle remains why organizational economists have chosen to work with models in which BR occupies at best a small corner – of a mainly rhetorical nature – of the model, the rest of the space being taken up by common priors, dynamic programming, etc. Not only was Simon critical of, for example, dynamic programming (Simon 1978), his work was taken up with rather concrete manifestations of BR often explored in detailed computational models (e.g., Newell and Simon 1972). Simon himself published prolifically on firms and other organizations (e.g., Simon 1947, 1951, 1991a, 1997; March and Simon 1958). Given all this, how can it be that organizational economists have been reluctant to be serious about BR, in the sense of actually modeling BR? The presumably most obvious reason is because of the well-known difficulties of aligning BR with the basic machinery of neoclassical microeconomics and game theory (Conlisk 1996; Rabin 1998; Camerer 1998). Thus, fundamental notions and modeling principles, such as subjective expected utility, common priors, rational expectations/dynamic programming, backward induction, etc., are not too easily aligned with fundamental findings of cognitive psychology (such as gain–loss asymmetries, role-biased expectations, etc.). Moreover, from the mainstream economist point of view, there is a huge price to be paid in terms of analytical tractability and clarity to the extent that one wishes to factor findings from cognitive psychology into economic models. This would seem to be consistent with BR becoming increasingly squeezed out of the economics of organization, as

H.A. Simon and organizational economics

165

the latter has become increasingly formalist, the Maskin and Tirole argument being the culmination of the squeeze-out operation. However, this may not be the entire story. In the following, I argue that Simon himself is partly responsible for the way in which economists have used (rather, not used) BR.

Simon lecturing economists on rationality In a series of papers, Simon (1976, 1978, 1979) made a sustained attempt to convince economists to take BR seriously. Two of these are particularly noteworthy in the present context, namely his Richard T. Ely lecture, “Rationality as Process and as a Product of Thought” (Simon 1978) and the Nobel lecture, “Rational Decision Making in Business Organizations” (Simon 1979). Needless to say, an Ely lecture, held at the annual meeting of the American Economic Association, and a Nobel even more, represent excellent platforms for persuasion exercises. It is therefore worth looking a bit into these papers. The Ely and the Nobel lectures The lectures differ primarily in the dimensions of depth and broadness, the Nobel being more survey-oriented and less analytically engaging. The similarities are, however, much more striking than the differences. First, they cover much the same themes. Second, both lectures are very strongly rhetorical in a number of senses. Thus, they employ a host of familiar rhetorical devices, such as metaphorical reasoning (particularly the Nobel), analogies, reference to authority, and, yes, quotations from Alice in Wonderland. Moreover, the lectures are rhetorical in the sense that they are very much taken up with “probing what men believe they ought to believe, rather than prove what is true according to abstract methods” (Booth 1974: xiii; cited in McCloskey 1983: 482). An aspect of this “probing” is that the lectures are highly polemical, with much strong critique of economists such as Fritz Machlup, Milton Friedman, Edward Mason and other defenders of the neoclassical theory of the firm. Also, there is indeed no attempt to “prove what is true according to abstract methods”; quite the contrary, “abstract methods” (axiomatic, non-empirical, etc.) are criticized and a rather sturdy elementary empiricism is promoted, one that involves “straightforward ‘anthropological’ field study” (1978: 12). General equilibrium theory is condemned as mere intellectual puzzle-solving, it being noted that “[p]erhaps some of these intellectual mountains have been climbed simply because they were there” (1979: 493–494). All these aspects are signaled at the beginning of both lectures. Thus, they begin with the traditional opening gambit for talks of this kind – namely, invoking authorities. The relevant authorities are Richard Ely and Alfred Marshall. Citing and quoting them is motivated by their taking a

166

Nicolai J. Foss

different, and more “realistic,” view of economics from the abstract one associated with Lionel Robbins of how to best allocate scarce means among competing ends, an understanding that is, of course, closely related to the notion of maximizing. The richer and more realistic view of behavior in Marshall and Ely leads directly into the rationality theme. Particularly in the Ely lecture, Simon handles this in a strikingly rhetorical manner. He begins by arguing that indeed most of human behavior has a rational element; however, maximizing rationality may not adequately describe this rational element. He then shows that economists in fact do use weaker forms of rationality, particularly in connection with institutional issues, where economists are as methodologically functionalist as sociologists and anthropologists (transaction cost economics is mentioned here). In general, Simon continues, it will not do to separate the various human and social sciences on whether they ascribe to rationality to human beings or not; they all do (Freudian psychology and sociological social exchange theory are provocatively mentioned as examples), and economics differs only by having adopted a particular, narrow conception of rationality. In particular, economics is not concerned with the process of choice, such as what are effective procedures for searching for solutions, only with the results. However, if attention is not in unlimited supply, it is necessary to account for the allocation of attention in search processes. A theory of this is a theory of procedural rationality. However, Simon admits, he is not aware “that there has been any systematic development of a theory of information and communication that treats attention rather than information as the scarce resource” (1978: 13). The Nobel lecture is slightly more specific about concrete manifestations of BR than the Ely lecture. Here, as in many other places in the two lectures, Simon employs the mode of argumentation of beginning by criticizing a mainstream position and then arguing that the behavioral alternative is superior. Thus, he begins by strongly criticizing the methodological notion (endorsed in economics by Machlup and Friedman) that theories can only be tested with respect to their predictions of aggregate phenomena. This is a methodologically unsound idea; instead, attention should indeed be directed toward the soundness of basic assumptions, not the least with respect to behavior. Moreover, it is insufficient to help discriminating between mainstream and behavioral perspectives, as some of the central mainstream predictions, notably negatively sloping demand curves and first-degree homogeneity of production functions, might be as well explained by a behavioral theory. This leads into a lengthy discussion of “normative decision theory,” much of which, quite appropriately for a Nobel lecture, surveys Simon’s own work. Later in the lecture there are other references to “advances in the behavioral theory,” notably the works of Tversky and Kahneman, Simon’s own work on the psychology of problem-solving, theories of organizational decision-making, and various theories of firm organization, such as the work of Nelson and Winter,

H.A. Simon and organizational economics

167

Cyert and March, Degroot, Radner, Leibenstein, Kornai, Williamson, and, rather surprising, Baumol’s managerialist theory of the firm (which is entirely based on maximization). Perspectives on the lectures In both lectures Simon makes critical and constructive arguments and observations. His main critical targets are the von Neumann/Morgenstern/Savage model, game theory, information economics, oligopoly theory, rational expectations theory on the level of theory and instrumentalism on the level of methodology. These are treated rather harshly. Thus, of subjective expected utility theory we are told “it is hard to take SEU seriously as a theory of actual human behavior in the face of uncertainty” (Simon 1978: 9), and of game theory, we are told that it “is embarrassing in the wealth of alternative solutions it offers” (1978: 10). Not only do these involve excesses of rationality, they also may not lead to determinate solutions (oligopoly theory is mentioned a number of times as an illustration; see also Simon 1976). Simon also offers a number of observations on how bounded rationality has a bearing on economic organization, as well as brief surveys of specific work on boundedly rational behavior. With the substantial benefit of hindsight we can discern a number of reasons why the rhetorics of Simon’s two lectures failed to convince contemporary (and succeeding) economists. Most obviously, Simon’s oeuvre suffered from unusually bad timing (hardly his fault, of course). Thus, the end of the 1970s, the time when Simon gave his two key lectures, is the beginning of the information economics and game theory revolution. Path-breaking work by Arrow, Mirrlees, Stiglitz, Ross, Myersbon, Wilson and others was between five and ten years old. The first statements of contract theory were about five years old; the revelation principle was almost contemporaneous. The first really convincing applications of non-cooperative game theory – incidentally to the part of economics, oligopoly theory, that Simon characterized as “the permanent and ineracidable scandal of economic theory” (1976: 140) – by Spence, Dixit and others were being worked out at about the same time. Rational expectations were moving from being strictly associated with Lucas and Sargent (and with specific policy positions), to becoming a generally acceptable modeling tool, etc. Thus, Simon was fighting a battle that even he, in that particular historical context, was bound to lose. However, even if the historical context had been more favorable to Simon’s arguments, there are still a number of fundamental problems with how Simon tried to convince his economist audiences of the soundness of his ideas. These have to do with the lack of definitions of BR, and the lack of modeling heuristics in Simon (1978, 1979). Consider these in turn. A fundamental problem which many discussions of BR have pointed to

168

Nicolai J. Foss

is that the concept is defined negatively rather than positively: BR tends to be seen as all those aspects of decision-making that substantive rationality is not. The problems with this are, first, that BR only assumes a real existence when viewed against its substantively rational counterpart, and, second, that the set of candidates for boundedly rational behaviors is without bounds. Of course, the problem is inherent in the name of the concept itself, and Simon may have committed a fundamental labeling blunder here. This may explain why he, from about the mid-1970s, used the notion of “procedural rationality” rather than BR, and why he, in fact, uses the concept very little in Simon (1978, 1979). In a discussion of his earlier work, he mentions that In Administrative Behavior, bounded rationality is largely characterized as a residual category – rationality is bounded when it falls short of omniscience . . . There was needed a more positive and formal characterization of the mechanisms of choice under conditions of bounded rationality. (Simon 1979: 502) The theory of satisficing search is, of course, one such characterization, and it, as well as other instances of behavioral decision theory, is discussed in Simon (1978, 1979). It has often been argued that a basic problem with satisficing search is that there is virtually nothing in the theory itself about the merits of alternative search procedures, and certainly not in economics. Simon explicitly argues that in order to understand the relative advantages of different procedures, it is necessary to step outside of economics and consider, for example, work on integer programming. However, his comments on the subject are extremely vague, and he chooses to “leave the topics of computational complexity and heuristic search with these sketchy remarks. What implications these developments in the theory of procedural rationality will have for economics . . . remain to be seen” (1978: 12). Thus, Simon essentially admits that the theoretically developed basis for theorizing on satisficing is virtually non-existent, and, at any rate, will likely emerge outside of economics. The conclusion for a listener to these 1978 or 1979 lectures can only be that work based on satisficing search must make use of rules for search and postulate aspiration levels that are essentially arbitrary, and at best justified by loose empirical considerations of a dubious sociological nature. In other words, Simon’s alternative program would not seem to present any non-arbitrary modeling heuristics. It is not surprising, then, that Simon’s economist audience, being first severely criticized for their methodical practice and listening to a wholesale condemnation of what they likely saw as hot new ideas, then being instructed to take seriously a fundamentally undefined notion, and finally being exposed to examples and applications of procedural rationality that

H.A. Simon and organizational economics

169

came close to what they would consider sociological reasoning, was not persuaded, and that the use of BR in economics remains, at best, rhetorical in a pre-McCloskeyan sense of the word. Could Simon have done it differently, and perhaps more successfully? A fundamental problem is that his message was in many ways so radically counter to most economists’ ingrained habits of thought, and they no doubt must have seen him as someone who refused to play by the rules of the game. Arguably, there was little Simon could have done, and would have wanted to do, about this. However, he could have done something, such as cutting down on the polemical elements, which arguably take up a disproportionate amount of space in the lectures. More importantly, he could have done more to present economists with precise behavioral models and the computational models associated with these (Newell and Simon 1972) in order to more convincingly present the case of a genuine behavioral alternative to mainstream modeling of behavior. It is indeed striking that when Simon talks to economists about BR, he is much less specific than when he addresses audiences in artificial intelligence and psychology about the same subject. Perhaps he felt that what was necessary was conveying the big idea and not go into formal detail (which might not have been appropriate in an Ely or Nobel lecture anyway). That may have been a mistake, for most modern economists like to be told about big ideas in a formal manner (the rational expectations revolution and the new growth theory comes to mind; Lucas 1972 and Romer 1986 are quite formal pieces, indeed).

Discussion Simon’s influence on organizational economists There can be little doubt that Simon has had a strong influence on many of those economists who have directed their analytical efforts toward firms and organizations. Williamson, Winter and Radner immediately come to mind.12 When various economics approaches to firms and organizations began to emerge in the beginning of the 1970s, Simon’s fundamental work on the subject was already two to three decades old (Simon 1947, 1951; March and Simon 1958). For some of the pioneers it was natural to look to Simon’s work for inspiration, perhaps particularly those who stressed the routinized nature of firm behavior (i.e., Nelson and Winter 1973). Bounded rationality seemed to link up directly with the notion of routines, since these may be interpreted as firm-level equivalents to individual behavioral rules that are adopted to reduce complexity under bounded rationality. To others the contribution of Simon’s thought lies elsewhere and is, in a sense, less direct. Thus, what Williamson appears to have gained from Simon is, first of all, a rationale for incomplete contracting, and, second, the notion that institutional choice is discrete (see Simon

170

Nicolai J. Foss

1978). Although he is, of course, familiar with Simon’s work in economics from the 1950s which is taken up with actually modeling boundedly rational behavior (e.g., Simon 1955), it is the Simon of the 1978 lecture – the interdisciplinary Simon who goes into institutional choice as a discrete one at considerable length13 – that seems to loom largest in Williamson’s thinking. As mentioned earlier he explicitly dissociates himself from understanding economizing with BR in terms of, for example, heuristic problem-solving. It is notable that those economists that have been able to utilize aspects of Simon’s thought in their work on economic organization were graduate students in the 1950s and early 1960s when Simon’s influence was perhaps more readily felt, and when he was around for interaction. In contrast, the works of those theorists of economic organization who began to publish from the mid-1970s are arguably much less, and usually not at all, influenced by Simon. One reason why this is so is simply that they were busy assimilating, applying and extending the new information economics and game theory tools when Simon presented his fundamental ideas to economists, and that Simon’s vision was simply too far from what they were up to. A part of that explanation, however, is that the Simon papers they were likely to know would be his two end-of-the-1970s papers. And, as has been argued, there is very little in these papers that may instruct economists wanting to build a BR research program in the economics of organization about the exact nature of BR and how to go about doing it. His later papers are equally barren in this respect. Thus, one of his last papers on organizational issues (Simon 1991b) surprisingly does not go into BR at all, but mostly takes issue with various themes in organizational economics, notably the assumption of opportunism. This has arguably contributed to the absence of a distinct BR research program in the economics of organization. The future of bounded rationality in the theory of economic organization So, what will happen to Simon’s Grand Theme of BR in the theory of economic organization? Will it gradually disappear, as contract theory takes over the whole field, everybody realizing that what some theorists try to say using BR may be said more elegantly with notions of asymmetric information, drawing on standard methodology? Although this may have been a reasonable prediction, say, ten years ago, there are reasons to think that the situation is different now. First of all, the evidence from psychology and experimental economics about the relative failures of expected utility theory is now so large that it cannot be ignored. Although various findings that are contradictory to EU theory were certainly well known at the time Simon gave this lectures, the amount of findings today is many times larger (Camerer 1998), and the

H.A. Simon and organizational economics

171

scientific quality of the relevant experimental methods is superior. There is, in other words, a much better critical foundation for BR. Second, there is now a much-improved positive foundation for theories of BR. Economists do not have to look at (at least to them) esoteric branches of computational theory or AI.14 In the fields of psychology and decision theory, perhaps closer neighbors to economics than AI, theorists and experimentalists have been at work since the end of the 1970s trying to align some aspects of BR with EU theory or develop distinct alternatives to EU theory (see Camerer 1998 for a fine survey and discussion). While economists may still be uncertain or ignorant about alternatives to EU theory – one wonders how many economists are familiar with cumulative prospect theory with rank-dependent weights which Camerer (1998: 166) singles out as the best alternative to EU theory in the light of the evidence – at least they may be increasingly alert to the new alternatives. Third, within mainstream contract theory, the Maskin and Tirole paper (Maskin and Tirole 1999) is not taken to be the last word about BR (and Maskin and Tirole do not appear to think of it in this way, either). Rather, it may be taken as a contribution that demonstrates the inherent limitations of a certain class of models (Kreps 1996), and points to the need to overcome these limitations, possibly by means of sophisticated treatments of BR (see Segal 1999). However, it is far from clear how to incorporate BR in contract theory. A main problem is that BR threatens to complicate drastically the link between current actions (e.g., investments) and anticipations of future payoffs, because BR is hard to square with the rational expectations of these payoffs. There is no BR theory of expectations (as far as I know), and although notions of satisficing may perhaps be invoked we are up against the usual problem that virtually anything can be postulated to be a reasonable aspiration level. Still, there are other uses for BR within contract theory, such as understanding the limitations of hold-up (Carmichael and McLeod 1999) and providing more refined understanding of why contracts are incomplete than merely postulating this by fiat (Mokerjee 1998; Segal 1999; McLeod 2000). Fourth, writers associated with transaction cost economics have increasingly begun to look into the BR component of the approach. Thus, Williamson (1998) himself has put forward a possible modeling strategy for how to incorporate richer notions of BR in transaction cost economics. He argues that the many ramifications of bounded rationality should be explored with a view to first identify those regularities in decision-making that differ from the classical model of von Neumann–Morgenstern– Savage, then work out the implications of these regularities for efficient organization, and finally fold these into the organizational design. The implication is that the efficiency questions of the economics of organization may usefully be reformulated, relying on more elaborate models of BR, so that “organization can and should be regarded as an instrument for utilizing varying cognitive and behavioral propensities to best advantage”

172

Nicolai J. Foss

(1998: 12). A limitation of Williamson’s (1998) paper (if not of the program he sketches) is that he seems mostly intent on demonstrating that findings of cognitive psychology are entirely consistent with “[t]he transaction cost economics triple for describing human actors – bounded rationality, farsighted contracting, and opportunism.” Therefore, he is not very specific about what exactly to do with these findings. However, Foss (2001) sketches various ways in which findings from the bias and heuristics literature may be utilized in transaction cost economics. The main idea is to interpret these findings as potential sources of transaction cost problems (e.g., contribution biases may increase bargaining costs) and argue that such problems influence the choice of governance structure. These ideas have been criticized by Loasby (2002), who argues that they continue an unfortunate tendency in most of organizational economics to only look at organization as something that exists in order to avoid the negative aspects and consequences of human behavior and not to stimulate the positive ones. There is certainly something to this critique. Transaction cost economics needs to address how governance choice influences alternative methods of search and learning. A recent, very interesting attempt at building a theory of this is Nickerson and Zenger (2002). Relying on complexity theory, they argue that Simonian heuristic search, which is usually necessary for problems that involve many, highly interdependent knowledge sets, is likely to require substantial knowledge sharing and ongoing interaction of knowledge sets. However, this exposes those who control the relevant knowledge sets to various knowledge exchange hazards (e.g., the Arrowian paradox of information). Because firm organization makes such hazards less severe and also better enables the building of a specialized language in terms of which the relevant communication may take place, complex problem-solving requiring heuristic search will be organized inside firms. Simpler problems, which may be decomposed into sub-problems, and correspondingly simpler search, may be well organized by market governance. Thus, Nickerson and Zenger elegantly combine key ideas from Simon (notably Simon 1962) with key transaction cost economics ideas in a theory of how governance structures both avoid the negative and promote the positive.

Conclusions This chapter has treated BR as a mainly rhetorical part of the practice of organizational economists. Thus, BR is invoked in a loose, intuitive manner to explain more central concepts, notably incomplete contracting, but it is not itself treated in much detail; it is not defined and modeled in any precise manner. I also argued that this practice may be related to the fact that Simon never really explained, at least to his economist audiences, what BR is, how it may be modeled so that it may be of use to economists, and how exactly BR impacts on those issues that interest economists of

H.A. Simon and organizational economics

173

organization (i.e., existence, boundaries and internal organization of firms; contract design). Thus, although the study of economists’ rhetorical practice may be helpful for illuminating aspects of scientific development, it has also been suggested that there are limits to the role of rhetoric in scientific development. At least, rhetoric in the sense of mere eloquence is not sufficient, because the successes of persuasion attempts are highly context-dependent. Thus, although Simon in his two major attempts to win economists over to the behavioral side (Simon 1978, 1979) exhibited rather considerable eloquence he failed to persuade, arguably because the context was so unfavorable to his arguments. However, that context has changed. Had Simon given his Ely and Nobel lectures in, say, 2001 and 2002 he would have found considerably more receptive audiences than he did in 1978 and 1979.

Notes 1 The comments of Mie Augier are gratefully acknowledged. 2 Game theorists may refer to BR more frequently, but game theory is hardly a distinct sub-field in economics; rather, it is a workshop that supplies machine tools to the various fields. 3 Giovanni Dosi (2002) reports that a few years ago, he (and a couple of coauthors) had a paper on industrial dynamics rejected from Econometrica, based on a review report, the main conclusion of which was that the paper would bring “back the discussion of industrial change to the Dark Ages of Herbert Simon.” Such bluntness is, however, rare. 4 Simon also published in economics in the 1970s on issues that are not directly related to BR – for example, Ijiri and Simon (1974) and Seskin and Simon (1973). 5 Albeit rather different conceptions of BR. Thus, BR in team theory is in actuality maximizing with costly communication, BR in transaction cost theory is the factor that explains why contracts are incomplete, and BR in evolutionary theory is a matter of search. 6 An impressionistic example: in Williamson’s work, bounded rationality looms larger in Markets and Hierarchies (1975) than in The Mechanisms of Governance (1996). 7 For example, rational expectations are central in most contract theory models. Applying the notion (any notion, really) of BR to expectations formation would imply that much contracting would produce unintended consequences, producing a need for ex post governance; however, this is not explored in contract theory. 8 Thus, Simon may be justified in his critique that “the new institutional economics has not drawn heavily from the empirical work in organizations and decision-making for its auxiliary assumptions” (Simon 1991b: 27). See also Simon (1997). 9 In the last case (non-verifiability), it is the enforcing party, such as a judge, that is boundedly rational. 10 That is, within the particular set-ups adopted in contract theory. 11 In contract theory courses at American universities, BR is often introduced (and finished) in the following manner: “We may use BR as a shorthand for anything that makes a contract incomplete.”

174

Nicolai J. Foss

12 For a brief essay on the Simon–Williamson relation, see Augier and March (2001). 13 Incidentally, Simon (1978) presents this theme as somehow intimately connected with BR, and mentions that discrete institutional choice is at variance with marginalism. While discreteness in choice indeed may require other tools (notions of complementarity and the underlying mathematical lattice theory) than conventional marginalist ones, it is entirely consistent with mainstream economics. 14 However, it should be mentioned that various computational approaches have been very usefully applied in the context of the theory of the firm by a number of evolutionary economists. See, for example, Marengo et al. (2000). Also, game theorists have made much use of AI notions of BR (see discussion and references in Lipman 1995).

References Augier, M. and March, J.G. (2001) “Conflict of Interest in Theories of Organization: Herbert A. Simon and Oliver Williamson,” Journal of Management and Governance, 5: 223–230. Brousseau, E. and Fares, M.H. (2000) “Incomplete Contract Theory and New Institutional Economics Approaches to Contracts: Substitutes or Complements?,” in C. Ménard, Institutions, Contracts, and Organizations: Perspectives From New Institutional Economics, Aldershot: Edward Elgar. Camerer, C. (1998) “Bounded Rationality in Individual Decision Making,” Experimental Economics, 1: 163–183. Carmichael, L. and MacLeod, W.B. (1999) “Caring About Sunk Costs: A Behavioral Solution to Hold-Up Problems with Small Stakes,” Olin working paper 99–19, University of Southern California Law School. Conlisk, J. (1996) “Why Bounded Rationality?,” Journal of Economic Literature, 34: 669–700. Dosi, G. (2002) “A Very Reasonable Objective Still Beyond Our Reach: Economics as an Empirically Disciplined Social Science,” Working paper. Dow, G. (1987) “The Notion of Authority in Transaction Cost Economics,” Journal of Economic Behavior and Organization, 8: 13–31. Foss, N.J. (2001) “Bounded Rationality in the Economics of Organization: Present Use and (Some) Future Possibilities,” Journal of Management and Governance, 5: 401–425. Grossman, S. and Hart, O. (1986) “The Costs and Benefits of Ownership: A Theory of Vertical Integration,” Journal of Political Economy, 94: 691–719. Hart, O. (1990) “Is ‘Bounded Rationality’ an Important Element of a Theory of Institutions?,” Journal of Institutional and Theoretical Economics, 146: 696–702. Hart, O. and Moore, J. (1990) “Property Rights and the Nature of the Firm,” Journal of Political Economy, 98: 1119–1158. Holmström, B. and Tirole, J. (1989) “The Theory of the Firm,” in R. Schmalensee and R.D. Willig (eds), The Handbook of Industrial Organization, Amsterdam: North-Holland. Ijiri, Y. and Simon, H.A. (1974) “Interpretations of Departures from the Pareto Curve Firm-size Distributions,” Journal of Political Economy, 82: 315–332. Kreps, D.M. (1996) “Markets and Hierarchies and (Mathematical) Economic Theory,” Industrial and Corporate Change, 5: 561–595.

H.A. Simon and organizational economics

175

Lazear, E.P. (1991) “Labor Economics and the Psychology of Organizations,” Journal of Economic Perspectives, 5: 89–110. Lipman, B.L. (1995) “Information Processing and Bounded Rationality: A Survey,” Canadian Journal of Economics, 28: 42–67. Loasby, B.J. (2002) “Economics After Simon,” Working paper, Department of Economics, Stirling University. Lucas, R.E. (1972) “Expectations and the Neutrality of Money,” Journal of Economic Theory, 4: 103–124. McCloskey, D. (1983) “The Rhetoric of Economics,” Journal of Economic Literature, 21: 481–517. McLeod, W.B. (2000) “Complexity and Contract,” Revue d’Économie Industrielle, 92: 149–178. March, J.G. and Simon, H.A. (1958) Organizations, New York: Wiley. Marengo, L., Dosi, G., Legrenzi, P. and Pasquali, C. (2000) “The Structure of Problem-Solving Knowledge and the Structure of Organizations,” Industrial and Corporate Change, 9: 757–788. Marschak, J. and Radner, R. (1972) The Theory of Teams, New Haven, Conn.: Yale University Press. Maskin, E. and Tirole, J. (1999) “Unforeseen Contingencies and Incomplete Contracts,” Review of Economic Studies, 66: 83–114. Milgrom, P. and Roberts, J. (1988) “Economic Theories of the Firm: Past, Present and Future,” in N.J. Foss (2000) The Theory of the Firm, Vol. 1, London: Routledge. —— (1992) Economics, Organization, and Management, Upper Saddle River: Prentice-Hall. Mokerjee, S. (1998) “Ambiguity Aversion and Incompleteness of Contractual Form,” American Economic Review, 88: 1207–1231. Nelson, R.R. and Winter S.G. (1973) “Towards an Evolutionary Theory of Economic Capabilities,” American Economic Review, 63: 440–449. Newell, A. and Simon, H.A. (1972) Human Problem Solving, Englewood Cliffs, N.J.: Prentice-Hall. Nickerson, J. and Zenger, T. (2002) “A Knowledge-based Theory of Governance Choice – A Problem-solving Approach,” Unpublished paper. Rabin, M. (1998) “Psychology and Economics,” Journal of Economic Literature, 36: 11–46. Romer, P. (1986) “Increasing Returns and Long-Run Growth,” Journal of Political Economy, 94: 1002–1037. Rubinstein, A. (1998) Modeling Bounded Rationality, Cambridge, Mass.: The MIT Press. Schwartz, A. (1992) “Legal Contract Theories and Incomplete Contracts,” in L. Werin and H. Wijkander, Contract Economics, Oxford: Blackwell. Segal, I. (1999) “Complexity and Renegotiation: A Foundation for Incomplete Contracts,” Review of Economic Studies, 66: 57–82. Sent, E.M. (1997) “Sargent versus Simon: Bounded Rationality Unbound,” Cambridge Journal of Economics, 21: 323–338. Seskin, E.P. and Simon, H.A. (1973) “Appendix to the article ‘Residential Choice and Air Pollution: A General Equilibrium Model,’ ” American Economic Review, 63: 966–967. Simon, H.A. (1947) Administrative Behavior, New York: Macmillan.

176

Nicolai J. Foss

Simon, H.A. (1951) “A Formal Theory of the Employment Contract,” Econometrica, 19: 293–305. —— (1955) “A Behavioral Model of Rational Choice,” Quarterly Journal of Economics, 69: 99–118. —— (1962) “The Architecture of Complexity,” in S. Masten and O.E. Williamson (eds), Transaction Cost Economics, Aldershot: Edward Elgar. —— (1976) “From ‘Bounded’ to ‘Procedural’ Rationality,” in S. Latsis (ed.), Method and Appraisal in Economics, Cambridge: Cambridge University Press. —— (1978) “Rationality as Process and as a Product of Thought,” American Economic Review, 68: 1–14. —— (1979) “Rational Decision Making in Business Organizations,” American Economic Review, 69: 493–513. —— (1991a) Models of My Life, New York: Basic Books. —— (1991b) “Organizations and Markets,” Journal of Economic Perspectives, 5: 25–44. —— (1997) An Empirically Based Microeconomics, Cambridge: Cambridge University Press. Tirole, J. (1999) “Incomplete Contracts: Where Do We Stand?,” Econometrica, 67: 741–781. Williamson, O.E. (1971) “The Vertical Integration of Production: Market Failure Considerations,” American Economic Review, 61: 112–123. —— (1975) Markets and Hierarchies, New York: Free Press. —— (1985) The Economic Institutions of Capitalism, New York: Free Press. —— (1996) The Mechanisms of Governance, Oxford: Oxford University Press. —— (1998) “Human Actors and Economic Organization,” Paper for the 1998 Paris ISNIE Conference.

7

Cognition, entrepreneurial conceptions and the theory of the firm Ulrich Witt

Introduction With its quest for explaining the ‘nature’ of the firm, the transaction costs oriented theory of the firm (Coase 1988; Williamson 1979) has successfully shaped the agenda of the research program of New Institutional Economics and, thus, a substantial part of economic theorizing about institutions. As is well known, the ‘nature’ of the firm is explained by putting it in perspective with market exchange and by showing that it is a distinct way of inducing and coordinating the division of labour. However, like much of the theorizing in New Institutional Economics, the Coasian and postCoasian reflections on the firm apply a version of economic analysis which is cast entirely in static and comparative static terms. This makes sense when the purpose is to explain the existence and the limits of an institution by demonstrating that it serves a certain function or functions. On the other hand, the genesis and (endogenous) change of an institution cannot easily be dealt with on the basis of a static mode of reasoning. To borrow a term from Menger ([1883] 1963), who may be considered one of the forerunners of New Institutional Economics (see Vanberg 1998), the processes of genesis and change require a ‘causal-genetic’ explanation. The quest for extending New Institutional Economics to the explanation of the genesis and change of institutions seems more than justified in view of the fact that institutions clearly change over their lifetime. The firm is no exception (see Chandler 1990, 1992): there are small, recently founded enterprises which are as different from large, established companies as a firm organization itself is from some modern forms of markets. Unfortunately, a causal-genetic approach is much less well developed than its static, functional counterpart and cannot be derived from the latter by simply adding a time dimension. The reason is that the genesis and endogenous change of institutions cannot be conceptualized within the framework of invariably given information as it must be presumed in a static mode of reasoning. The evolution of institutions, whether this is intended deliberately or emerges interactively and unintentionally, hinges to a large extent on how the perceptions and understanding of the

178

Ulrich Witt

involved agents change as a consequence of learning and of changing cognitive interpretation (Denzau and North 1994; Rizzello 1999: part III). Therefore, for a causal-genetic approach to materialize, what needs to be expanded in the first place is the underlying theory of economic behaviour so that it can account for cognitive contingencies and the change over time which they imply. In the present chapter an attempt is made in that direction by investigating both the cognitive influences on behaviour adaptation and their systematic effects on the genesis and endogenous change of an institution. The exemplary case is – as in the bulk of New Institutional Economics – the firm. It is here that the attempt to understand the genesis of, and systematic change in, created or planned institutions can build on some important contributions in the past which deal with organizational development (Penrose 1959; Langlois 1992; Langlois and Robertson 1995; see also Foss 2001 for a survey). A somewhat broader approach is necessary to come to grips with the cognitive influences on the firm. This approach will be outlined in the first section. Drawing on earlier work (Witt 1998, 2000), the notion of an entrepreneurial business conception is derived on such a basis in the next section. That notion allows us to understand much better the way in which cognitive frames affect the coordination of intra-firm problem-solving and division of labour. At the same time, the notion sheds some light on the role of the entrepreneur within the firm organization, a much neglected issue in the Coasian theory of the firm. The section to follow starts the discussion on the genesis and change of firm organizations. In that section, focus will be on entrepreneurial business conceptions as conjectures about business opportunities which compete – in the labour market – for becoming material in the form of firm organizations. This competitive process is an important part of the genesis of firms, but it implies a nucleus of change as well: since these conjectures do not necessarily live up to the expectations, the outcome of competition in the labour market may be unstable. If so, certain forms of organizational change may follow. Another section is devoted to a different source of systematic changes in the firm which may be labelled ‘developmental’ changes because they deal with the effect which the growth of the organization over a firm’s life cycle has on its mode of operation or, as it is called here, on the prevailing entrepreneurial regime. The final section presents some conclusions.

Theoretical extension: cognition and social learning Some reflections on the operation of the human mind are necessary to provide a basis for discussing the cognitive contingencies on which a firm organization rests at any point of time (see Anderson 1990: ch. 3; cf. also the excellent discussions in Rizzello 1999). Powerful as man’s cognitive apparatus and memory are, they are subject to constraints as has long

Cognition and the theory of the firm

179

been acknowledged in economics in the debate on ‘bounded rationality’. These constraints induce our minds to process sensory information selectively on the basis of discriminative attention processes. These, in turn, operate by means of cognitive cues. Incoming information is screened on an associative basis to check whether it fits patterns that already exist in the memory. If familiar configurations cannot be easily identified in the incoming information, or if existing patterns cannot be easily extended to encompass it, the information is ignored. The cognitive cues that help memorize patterns and discriminate among the incoming information are themselves organized into larger and more complex systems called ‘cognitive frames’ (see Anderson 1990: ch. 5). Cognitive frames support classificatory activities and thus allow knowledge to be represented in a meaningful way. By their very nature, these cognitive frames are idiosyncratic features of the individual mind. On the basis of a limited number of probably genetically coded cues, a lifelong process takes place in which increasingly extended associative chains with increasingly more complex sets of frames are formed and stored in long-term memory. Against the current state of the cognitive cues and frames, the human mind selectively utilizes information and identifies meaning, intuitively as well as at the level of deliberate reasoning. These cognitive frames can lead to fairly rigid mental conceptions and interpretation patterns. In any case, though different cognitive tasks may be pursued on the basis of different cognitive frames, at any time no more than just one particular cognitive frame is in operation. This also implies that, while in use, such a frame cannot itself be at the same time made an object of cognitive reflection. The fact that mental activity is constrained and selective also affects our capacity to imagine and reflect on alternatives for action. Some particular courses of action are more or less clearly conceived and carefully thought through while others that could, in principle, be imagined are not even recognized. Given the extensive debate on bounded rationality in economics, all this does not appear entirely new. A less-well-known implication of the social-cognitive theory advocated here is the fact that the individual’s cognitive development just outlined does not take place in an entirely autonomous or isolated way, but is moulded by social interactions (Bandura 1986: ch. 2). Communication, particularly of the non-formal, and often one-sided, kind that occurs in socializing with others, enables people to observe, and make comparisons with, the behaviour of other agents and their rewarding or non-rewarding experiences. The more frequent and intense the latter observations are, the more likely that there will be some similarities in the subjective representation of knowledge. The exchange with, and selective imitation of models of behaviour provided by, the social environment tends to induce tacit cognitive commonalities, the more so the more intense communication is. Socially shared interpretation patterns and frames emerge, as do common tacit knowledge of facts,

180

Ulrich Witt

hypotheses, practices, and skills, including socially shared knowledge of rewarding or adverse experiences. Intense communication is usually subject to a spontaneous agendasetting effect. The effect modifies, in a self-reinforcing way which is similar for all the participants in the social interaction, the intensity with which some particular information, rather than other available information, is exchanged and attracts attention. Notwithstanding the idiosyncratic nature of individual cognitive frames and the corresponding diversity that results from each individual’s unique cognitive history, tacit cognitive commonalities are thus a common concomitant of intense communication within social groups. These commonalities extend to both the content of actionrelated knowledge and the way in which it is framed. This means that intense communication tends to induce similarities in the alternatives of action which individuals within an intensely communicating group selectively perceive as feasible and those that they disregard. Yet socially shared constraints in the selective representation of reality do not imply that people are prevented from choosing the most preferred alternative among those they do recognize. (People may, however, deviate from what the rationality postulate would prescribe if information processing were perfect; that is, non-selective and therefore unbiased. Unless the particular cognitive constraints of the communicating population are specified, the concept of rationality is therefore insufficient to actually determine what behaviour should be expected.) One kind of knowledge which is subject to the social learning processes outlined, and is particularly relevant in the context of the firm, is knowledge of how to behave in social interactions. Certain patterns of behaviour which are regularly shown by some person(s), and which can be observed in social interaction, may gain the status of commonly shared models of behaviour (Bandura 1986: ch. 2). The significance of such models lies in their vicarious nature; that is, the fact that their rewarding or non-rewarding consequences can be grasped by observation rather than by own experimentation. Within intensely communicating groups, learning by observation then means that the group members tend to focus on much the same models. Consequently, their individual learning processes produce correlated results which, via the prevailing social models, can intuitively be grasped by newcomers. (Often such models develop from spontaneously established, multilaterally respected, rules of how to behave into normative rules of what conduct should be.) Individual behaviour can, of course, deviate from established social models for many reasons. If so, this is more likely to arouse attention among group members, the more that behaviour challenges existing norms and models. The observed vicarious success or failure of deviant behaviour provides the other agents with information that enables them to assess their current behaviour in the light of the alternative not previously considered (perhaps even an entirely newly discovered action). The vicarious reward or loss of the one who

Cognition and the theory of the firm

181

deviates from established norms and models thus tends to induce or inhibit adjustments by imitation (Bandura 1986: ch. 7). Accordingly, social learning strengthens or weakens the tacit constraints on what is individually perceived as the choice set. If successful, a deviant way of behaving is likely to jeopardize the tacit cognitive commonalities. The corresponding reframing of action knowledge may cause a switch in social interactions from one social model of behaviour to another one.

Entrepreneurial business conceptions and the nature of the firm The multi-person firm is a way of organizing the division of labour. As when the division of labour is organized via markets, this kind of firm has to rely on knowledge dispersed among several agents. Moreover, those agents must be motivated to undertake the physical and mental efforts necessary to make use of their individual knowledge in the interest of the firm, and their efforts must be co-ordinated. How is all this achieved? It will be claimed here that for a true understanding of this problem the cognitive underpinnings of any business venture must be recognized and theoretically accounted for. Any division of labour, be it market based or firm based, has its origin in a venture undertaken by some person or persons. Such a venture rests on an idea of how to (re-)organize work, an idea which initially may not be more than just a speculative imagining. To develop the respective ideas, and to undertake the efforts to see them through, is a crucial entrepreneurial input for organizing the division of labour. While many of these ideas may be achieved exclusively by means of market transactions, there are also more complex ideas which an entrepreneur is unable to realize by her/himself exclusively on the basis of standard market contracts because of cognitive and motivational reasons to be explained in a moment. These ideas are called business conceptions here, and represent the entrepreneur’s image of what business to do and how to do it with the staff hired. A business conception is general, and rather unspecific, in nature. This is an important feature, because the cognitive limitations just discussed prevent the entrepreneur, as much as everyone else, from anticipating all possible business moves that unfold into the future. The general, unspecific nature of the business conception leaves room to cover up-coming information on non-anticipated events and the consequences of own actions which would otherwise be difficult to classify and assess with respect to their implications for the business. Thus a business conception has the features of a cognitive frame. It helps to interpret the current events in the perspective of the firm’s overall orientation and associates appropriate actions with the state of affairs conceived. Moreover, the fact that, at any point in time, the human mind operates on the basis of the one, current cognitive frame turns out now to have decisive meaning for understanding

182

Ulrich Witt

the cognitive basis of the firm organization. An entrepreneur who creates a firm organization to accomplish her/his business conception by drawing on individual knowledge dispersed among the employees and on their work efforts achieves her/his goal the better, the better the decisions which the firm members make within their respective area of discretion fit the entrepreneur’s business conception. Since each firm member operates on the basis of an individual cognitive frame, dispersed knowledge and individual endeavour would be concerted most effectively if all firm members were to share the entrepreneurial business conception as their own cognitive frame. Sharing the entrepreneurial business conception as a cognitive frame also matters from the motivational point of view, since it makes a great difference whether or not people adopt the attitude of contributing to a common goal. Their task perception tends to be framed in a way that makes their attention more devoted to solving problems in the interest of the firm’s goals than to pursuing private short-run inclinations and separate interests. Conversely, it may be concluded that individual knowledge and efforts are not well concerted if there are rivalling business conceptions pursued within the firm, or if all firm members only follow an opportunistic conception of extracting some form of short-run rents for themselves. From a cognitive point of view, the possibility of concerting individual motivation and dispersed knowledge on the basis of a socially shared cognitive frame is therefore a crucial part of what makes the firm organization an attractive alternative to the market mechanism. Indeed, this provides an independent argument for why an entrepreneur creates a firm organization for expanding business instead of realizing her/his ideas on the division of labour exclusively via ordinary market transactions. As a business grows, the number of transactions necessary for buying and selling all services and materials on the market at the right time quickly increase. To conceive of all those transactions, to fix, and to supervise them becomes impossible for a single entrepreneur because of the limited individual information processing capacity. If there were no firm organization, many of these activities would have to be delegated to subcontractors via ordinary market transactions. However, the services of a subcontractor which would suit the entrepreneur’s business conception as the venture unfolds into the future are not yet fully known at the time of contracting. How should it be possible then to contract yet unknown services? Only a contract of rather unspecified nature, it appears, could solve the problem. Such a contractual form is indeed provided in the market, though not by subcontractors but by the employment contract in the labour market. The reason is that there is a precondition for a contract as unspecified as this one to allow the various specialized activities to be co-ordinated and provided in a coherent way: the adoption of one business conception as an interpretative frame. This is not feasible in the case of subcontracting with independent business men who follow their own busi-

Cognition and the theory of the firm

183

ness conception. The predominance of one and the same business conception can only be achieved, if at all, by setting up an own firm organization and hiring employees. From the preceding considerations it should be clear that an entrepreneurial conception can only give rise to an efficient division of labour within the firm, i.e. organizational coherence, and a successful firm performance, if it is uncontested within the firm organization. To achieve this, the entrepreneur’s conception must be communicated to, and adopted by, the firm members in the first place. This is not so easily done, because employees cannot be induced to adopt by order the entrepreneur’s business conceptions as a cognitive frame. Similarly, models of behaviour which would be supportive for the entrepreneur’s goals cannot be implemented by instruction. Rather, the formation of individual cognitive frames and the emergence of group-specific social models follow fairly complex processes in which social learning plays a decisive role. The more intense and the more lasting communication and observational learning among interacting agents, the more likely these agents are to tend to develop collectively shared interpretation patterns as well as common tacit knowledge of facts, hypotheses, practices, and skills. These cognitive commonalities are partly due to the fact that in intense interactions the involved agents’ selective information processing is occupied with much the same topics which, in a sense, are processed in parallel, while there is little or no attention left for other topics. Partly collectively shared interpretative frames emerge in an intuitive way from mutual observational learning. In any case this is something different from organizing a formal communication process and giving orders within that process. What counts much more are the informal communication processes within the firm. However, while, by the nature of the employment contract, the entrepreneur can define and control the agenda of formal communication within the firm, there is much less institutional support for being able to shape the informal communication processes within the firm organization. On that level, entrepreneurial conceptions and social models may well be contested by rival cognitive frames and social models. Failure to prevent these from tacitly taking the lead in the firm’s informal communication can have far-reaching consequences for organizational coherence and for the firm’s performance. Therefore, a particular capacity is required for the entrepreneur to succeed in shaping informal communications in a way that is advantageous to the propagation of her/his business conception. (The same holds with respect to defending the predominance of her/his business conception against potential contestants.) That capacity may be called ‘cognitive leadership’ (see Witt 1998). On the one hand, such a capacity calls for social skills like communicativeness, persuasiveness, and persistence, and for fairness, credibility, appreciativeness. Obviously, the agents assuming the role of an entrepreneur may differ wildly in the extent to which they command these skills. On the other hand, the intrinsic features

184

Ulrich Witt

of business conceptions surely also play an important role for whether or not leadership can successfully be exerted on their basis. If a conception is too complex and sophisticated, if it lacks soundness and appeal – not the least in terms of career options, remuneration, qualification enhancement, and working conditions for the employees – or if it is obviously unsuited for the imagined business, it is difficult to make employees adopt it. Similarly, the nature of the social models matters for whether and to what extent it will be adopted within the firm organization, with or without the entrepreneur’s help.

Entrepreneurial conceptions and the competitive genesis of the firm Without some, at least implicit, form of a business conception no multiperson firm can be run, and the genesis of firm organizations is significantly affected by the kind of business conception used. However, business conceptions and entrepreneurial imaginings can be developed in large numbers and by many people. No doubt there have always been many more of them in the minds of potential and real business people than those actually pursued in an enterprise. Hence, there must be some kind of competitive process that discriminates between imaginings, which are indeed turned into ventures, and the vast number of those that have no consequences. This fact is significant for understanding the cognitive element in the genesis of the firm and its emergence from a competitive process which expresses itself in – and this may come as a surprise – the labour market. This market sets the stage because thinking up a business conception only has real consequences in the form of a firm organization if an attempt is made to hire employees and to organize their work on that basis. In principle, everyone who has thought up a business conception could make an attempt in that direction. Yet hiring is competitive and, moreover, it is mediated through markets which act as allocation devices. Therefore, prospective entrepreneurs have to develop some idea of the wages their business would allow them to offer to potential employees. Since the wage offers must be competitive, the question arises as to how much income the entrepreneur would be able to retain for her/himself after paying out the wages offered. On the other hand, people contemplating their prospects as entrepreneurs can themselves obtain a competitive offer to work as an employee in a non-entrepreneurial position in someone else’s venture. The income to be earned by joining someone else’s venture can be measured against the imagined income from trying to set up an own venture, and so too can the non-pecuniary features of the alternatives of either assuming the entrepreneurial role or that of the employee. This means that all entrepreneurial projects are involved in a contest which, because there is no own market for evaluating such projects, takes place in the labour market.

Cognition and the theory of the firm

185

Everyone who participates in it by considering her/his own business idea obtains some information about how promising their own imaginings are compared to those of others. The contest not only sorts the vast number of entrepreneurial visions into realized and non-realized ones but, by the same token, sorts people into entrepreneurs and employees. Leaving the non-pecuniary aspects aside, the agents who take on the entrepreneurial role are those who believe that they are able to pay more than the other agents would expect to earn after paying them. The sorting, and thus the genesis of the firm organization, reflects a multilateral opportunity cost assessment in which everyone participates on the basis of her/his subjective calculations. These subjective calculations (and thus the individual comparisons made in the implicit contest) may, of course, turn out to be wrong. For this reason, the results of the sorting process are not necessarily stable. This means that the genesis of the firm organization may already contain the nucleus of later change, not only in the trivial sense of firms shutting down after failing to generate profits. The subjective assessments are confirmed if the imagined entrepreneurial income and the wages and non-pecuniary benefits offered to those who decide to become employees can indeed be earned. This either requires a business conception which proves to be sufficiently profitable or an entrepreneur who proves to be in command of sufficient cognitive leadership skills – or both. Even if these conditions are met, however, some results of the sorting process may not be stable. The subjective opportunity costs of being hired by someone else are expressed by the imagined entrepreneurial income that could be earned by pursuing an own venture. They can rise because learning and experience on the job creates revaluations of the (non-realized) business conception and the prospects of an own venture. If this happens, an agent who has just become an employee in a non-entrepreneurial position may want to leave the firm and set up something of her/his own. If the firm wishes to prevent this, it would have to transfer the agent to a position in which a wage increase sufficient to keep up with the changed opportunity costs can be paid and the agent can nonetheless profitably be employed. In this case, the competitive sorting process of entrepreneurial capacity continues within the firm and does not necessarily affect the firm organization. A substantial increase in the earnings of an employee may often only be possible if the agent can be given a position in which her/his entrepreneurial capacity can be utilized. The agent could, for instance, be entrusted to pursue his own business conception as an entrepreneur employed within the existing firm. However, such a move is not without risk. The more entrepreneurial tasks are assigned to an agent in a firm organization, the better prepared the agent may become for breaking away and starting an own business. (An employee in a non-entrepreneurial position, by contrast, has to devote her/his mental activities, and thus attention and effort, to carrying out other mental tasks. Hence, less of her/his attention

186

Ulrich Witt

is available for inquiring more deeply into her/his own entrepreneurial ideas.) In fact, when breaking away, a former employee may trigger further change within the firm organization which (s)he leaves in the form of a ‘fissioning’ phenomenon. Those who leave and set up their own firm may be able to offer competitive wages and non-pecuniary rewards to members of their former firm with whom they share conceptions and/or social models and may thus attract them to the new enterprise.1

Towards a developmental theory of the firm organization The entrepreneurial creation of a firm organization is the initial episode in a longer process of organizational development. The static and comparative static perspective of the transaction cost approach to the firm tends to detract attention from the fact that there are systematic developments which can be observed to unfold over the history of an organization (Rathe and Witt 2001). In the empirical literature, such developmental regularities have often been associated with the life-cycle metaphor (e.g. in Quinn and Cameron 1983), but an analytic foundation has rarely been attempted. A first step towards a more elaborate theoretical basis is to keep track of the organizational development and the various factors that may influence it by distinguishing different organizational states and the transitions between them. Let us therefore return to the entrepreneur’s founding of the firm organization and the organizational states that may result here. These depend on two things: the entrepreneurial choice of a regime (which could be modelled as a fairly conventional decision problem) and the factors leading to success or failure of the chosen regime (which, because of their heterogeneity, are more complex to analyse). In making her/his choice in the light of her/his subjective imaginings, the alternatives available to the entrepreneur are defined by a regime of cognitive leadership on the one side, or a regime of governance on the other. The latter choice amounts to waiving the attempt to gain cognitive leadership and to immediately subject the employees’ performance to entrepreneurial monitoring just as is assumed as the regular case in Alchian and Demsetz’s (1972) monitoring approach. The entrepreneur’s subjective estimates of what her/his chances of succeeding are with the one or other regime reflect a self-assessment. In the case of the cognitive leadership regime, this self-assessment refers to the entrepreneur’s own social skills and the intrinsic quality of the business conception. In the case of the governance regime, it refers to the entrepreneur’s own monitoring capacity. If the entrepreneur rates her/his social skills and/or the business conception highly, there will be a comparatively higher estimate of the probability of succeeding with cognitive leadership and vice versa. Accordingly, it may be argued that the entrepreneur tries to implement either a cognitive leadership regime or a governance regime, depending on her/his self-assessment.

Cognition and the theory of the firm

187

What happens thereafter – the actual consequences of the founding of the firm – hinges on several factors which can usefully be structured by distinguishing between two different questions. The first is whether the performance of the firm is sufficiently successful to enable it to grow. The second question relates to what happens inside the firm organization if it does grow. With respect to the first question it is clear that actual performance strongly depends on factors exogenous to the firm’s entrepreneurial regime (e.g., the co-evolution of the firm’s markets). As the product lifecycle literature has it – life is easier, and profits are easier to earn, in booming markets with increasing rates of growth of total sales. It is increasingly less easy when growth rates decrease or even become negative. Many other contingencies with similar influences on growth and/or decline of the firm can be easily imagined, and all of them seem to be closely correlated with profitability. Note, however, that profitability differences do not necessarily translate into corresponding differences in changes of the size of the business and the firm; that is, its growth, stagnation, or decline. Profits can be used for many purposes other than investing in expanding the business, and there are, at least within bounds, several means other than the accumulation of profits for financing a firm’s growth. Moreover, growth of a business can always be achieved at the expense of profitability. Growth resulting from using such a strategy cannot be compared to growth accomplished while maintaining a given profitability. Financing strategies, growth strategies, and many other such strategies which influence growth clearly are endogenous to the firm. Lacking more specific information, it seems justified, for the present expository purpose, to treat all these influences as if they were not systematically correlated with the initial choice of an entrepreneurial regime. With regard to the second question note that, after personnel have been hired, the entrepreneur’s behaviour within the firm organization is no longer a matter of simple strategic choice between more or less welldefined alternatives. Rather, it is a matter of daily struggle for cognitive leadership under one regime, and for staying in control of the bureaucracy under the other. As long as the entrepreneur follows a cognitive leadership regime, the spontaneously and continuously produced results of the social learning process within the firm organization, and thus success or failure and the duration of the respective outcomes, hinge on the actually demonstrated social skills of the entrepreneur and/or the intrinsic quality of her/his business conception.2 If governance has been chosen, success or failure and the duration of the respective outcomes hinge on whether, and how long, the entrepreneur is able to keep control of the bureaucracy. If growth of the business does occur, and if the firm organization expands in terms of the number of staff employed, then the crucial question is how long the chosen entrepreneurial regime will remain successful. This question leads to identifying the organizational states which may follow the

188

Ulrich Witt

initial ones and the possible hypotheses about why and when transitions between the organizational states occur. The hypothesis suggested here is that, sooner or later, a critical size of staff will be reached beyond which a further expansion of the firm size overcharges the entrepreneur’s capacities which (s)he has previously demonstrated. The critical size of staff is likely to differ between entrepreneurs and businesses, and may be argued to be smaller on average for a cognitive leadership regime than for a monitoring regime. If a monitoring regime has initially been chosen – that is, an attempt to achieve coordination through detailed directions, regulations, authorization, and tight control – it is known that such a regime will cause frictions and be slow and costly in terms of time resources. The larger the firm, the more these negative effects tend to lower its efficiency. Once the critical size is exceeded, the entrepreneur’s effectiveness in monitoring fails to ensure the minimum level of organizational coherence and efficiency necessary to maintain a profitable operation of the firm. Hence, the organizational state in this case results in disorder. It is characterized by stagnation, decline, and eventually exit of the firm – that is, nonsustainable organizational development. In the case of the cognitive leadership regime, exceeding the critical size means that the entrepreneur loses control of the informal agenda which supports her/his business conception and some supportive social models of behaviour. Other conceptions start to disseminate to a significant extent. The employees tend to switch to cognitive frames and corresponding social models of behaviour which can compete with the entrepreneurial business conception or invite opportunistic reflections. Attention is drawn to figuring out how they can take advantage of the unspecified nature of the employment contract by pursuing their own separate interests. Work effort tends to go down. As a consequence of all this, the firm organization is likely to perform in a significantly less coherent, and hence less efficient, way. Profitability is negatively affected, as is the potential for further growth. A critical stage like this is indeed often reported for organizations which once were start-up firms and whose founder-entrepreneur has become of age. As in the previous case, profitable operation of the firm organization may no longer be possible. An organizational state of disorder may result which again leads to stagnation, decline, and, eventually, the exit of the firm. Facing (or, perhaps, only fearing to converge to) a state of organizational disorder, the entrepreneur may, of course, be induced to react so that, with a growing size of staff, a transition from the initial organizational regime to an organizational state of disorder and decay does not necessarily occur.3 One way of responding to a crumbling monitoring regime could be to create a control hierarchy. This means delegating the entrepreneurial task of monitoring to possibly several layers of subordinate monitoring agents – the typical organizational form of the large, mature, and bureau-

Cognition and the theory of the firm

189

cratic firm. However, such a reorganization usually entails more than proportionately increasing administrative costs, the well-known ‘managerial diseconomies of scale’ (Mueller 1972). Therefore, whether or not such a division of the monitoring task inside the firm organization (an organizational state of an hierarchically extended monitoring regime allows a sustainable development hinges on two things: whether further growth of the firm’s business can be generated and whether this growth allows a compensation, or even over-compensation, of the increasing slack by realizing economies of scale and scope, or by being able to reap monopolistic profits. Where this is not the case, or where managerial governance declines even with a hierarchical organization because of uncontrollable opportunism, the further development of the firm is again likely to lead to stagnation, decline, and even exit. In the case where a cognitive leadership regime reaches its limits with growing staff size, the entrepreneur may also decide to fight the increasing incoherence and declining work motivation. One remedy that may be tried is a regime switch.4 If (s)he considers the critical size of staff for a governance regime larger than for a cognitive leadership regime, an attempt can be made to replace the fading regime of cognitive leadership with its organizational culture of loose hierarchical ties and its (also fading) intrinsic work motivation by a governance regime for running the firm organization. This means introducing a tight entrepreneurial monitoring of the employees’ performance. Since a transition from an entrepreneurial regime of cognitive leadership to a regime of monitoring curbs individual creativity and the intrinsic motivation in problem-solving (see Williams and Yang 1999), such a transition also means sacrificing a substantial advantage of the cognitive leadership regime, along with its positive effects on profitability and growth. If an (attempted) regime of cognitive leadership does not lead directly to failure and decay, but instead can be transformed into an effective monitoring regime, the further development of the firm organization is then the same as described above, with the transition from a monitoring regime to either disorder and decay or an hierarchically extended monitoring regime and its contingencies.

Conclusions While the empirical and historical evidence in modern capitalism clearly points to firm organizations being constantly in flux – new ventures are set up inside and outside existing firm organizations and, as time goes by, expand, decline, or exit – such change is only insufficiently dealt with in the (comparative-) static Coasian and post-Coasian theories of the firm. In these approaches, the main goal is to explain the ‘nature’ of the firm. This is meant, on the one hand, to identify the functions served by the institutional form of a firm organization compared to the coordination of the division of labour via market contracts, and, on the other hand, to

190

Ulrich Witt

determine the dividing line between the realm of the market and that of the firm (‘make or buy’). As is well known, the core argument in the explanation is savings in transaction costs which become feasible through choosing the one form of coordinating the division of labour over the other. The fact that an entrepreneurial division of labour needs to be envisioned in the form of a business conception in the first place is neglected, just as is the problem of how to induce the members of the firm to follow the entrepreneur’s business conception in their individual problem-solving activities rather than any other goals and interpretations. In the preceding sections of this chapter it has been argued that an approach which explicitly acknowledges the cognitive background of firm organizations can provide several new insights. First, it contributes to understanding why firms are chosen over markets for pursuing entrepreneurial plans. Cognitive coordination on the entrepreneurial business conception within a firm organization is a prerequisite for obtaining a high degree of motivation and coherence and efficiency in jointly pursuing a business venture and in solving, in a decentralized way, the unanticipated problems which arise for any such venture over time. By the very fact that several entrepreneurs with usually rival business conceptions are involved in a division of labour via ordinary market contracts, markets cannot ensure the same degree of cognitive coherence in seeing through a business venture as firm organizations. Second, the cognitive approach suggested here allows several contingencies for successfully organizing the division of labour within firms to be identified. Hypotheses can be derived from these contingencies which shed new light on both the genesis of firms and the developmental changes often observed to take place in firm organizations in historical time. As a consequence of interactive learning among organization members and/or of the growth of the firm, the coordinating power of the underlying business conception may fade. Among the developments which have been derived in this chapter, a particularly significant one is the transition of a cognitive leadership regime congenial to the entrepreneurial venture during its founding period and times of rapid expansion to a firm organization in which inspiration and persuasion are replaced by tight directions and control – in short, a governance regime. Despite its intrinsic motivational frictions and its disadvantage in problem-solving capacity, a governance regime may be able to manage the further growth of business provided it is able to exploit increasing returns to scale available through the sheer size of the operations. Indeed, as experience teaches, hierarchically organized managerial governance may guide the firm organization to the mature state of the large (often multi-nationally active) corporation or trust.

Cognition and the theory of the firm

191

Notes 1 Such a development has indeed been observed to occur on a rather regular basis in some highly innovative industries (see Ziegler 1985). 2 As explained before, while the entrepreneur is in control of the formal agenda setting in the firm, the informal agenda, and thus the socially shared cognitive frames and social models of behaviour, emerge in a spontaneous manner from the activities of all firm members. There is thus, in the perspective of the socialcognitive approach, no element of strategic choice involved in how the employees arrive at their cognitive attitudes. Rather, these attitudes result from social learning. 3 One possible, and often observable, entrepreneurial reaction which will not be considered further here is to withdraw from the business. Particularly in the case of an owner-entrepreneur there may be a temptation to put up the firm for sale and to ‘cash in’ on the efforts of building up and expanding a business once the organization has grown to a size at which the entrepreneur’s capacity to exert cognitive leadership is exhausted. 4 Another one, discussed in Witt (2000), is the attempt to subdivide cognitive leadership among employed entrepreneurs.

References Alchian, A.A. and Demsetz, H. (1972) ‘Production, Information Costs, and Economic Organization’, American Economic Review, 62: 777–795. Anderson, J.R. (1990) Cognitive Psychology and Its Implications (3rd edn), New York: Freeman. Bandura, A. (1986) Social Foundations of Thought and Action – A Social Cognitive Theory, Englewood Cliffs, N.J.: Prentice-Hall. Chandler, A.D. (1990) Scale and Scope – The Dynamics of Industrial Capitalism, Cambridge, Mass.: Harvard University Press. —— (1992) ‘Organizational Capabilities and the Economic History of the Industrial Enterprise’, Journal of Economic Perspectives, 6: 79–100. Coase, R.H. (1988) ‘The Nature of the Firm: Origin, Meaning, Influence’, Journal of Law, Economics and Organization, 4: 3–47. Denzau, A. and North, D.C. (1994) ‘Shared Mental Models: Ideologies and Institutions’, Kyklos, 47: 3–31. Foss, N.J. (2001) ‘Evolutionary Theories of the Firm: Reconstruction and Relations to Contractual Theories’, in K. Dopfer (ed.), Evolutionary Economics – Program and Scope, Boston: Kluwer Academic Publishers, 317–355. Langlois, R.N. (1992) ‘Transaction-cost Economics in Real Time’, Industrial and Corporate Change, 1: 99–127. Langlois, R.N. and Robertson, P.L. (1995) Firms, Markets, and Economic Change. A Dynamic Theory of Business Institutions, London: Routledge. Menger, K. ([1883] 1963) Problems of Economics and Sociology, Urbana: University of Illinois Press. Mueller, D.C. (1972) ‘A Life Cycle Theory of the Firm’, Journal of Industrial Economics, 20: 199–219. Penrose, E.T. (1959) The Theory of the Growth of the Firm, Oxford: Basil Blackwell. Quinn, R.E. and Cameron, K. (1983) ‘Organizational Life Cycles and Shifting

192

Ulrich Witt

Criteria of Effectiveness: Some Preliminary Evidence’, Management Science, 29: 33–51. Rathe, K. and Witt, U. (2001) ‘The “Nature” of the Firm – Static vs. Developmental Interpretations’, Journal of Management and Governance, 5(3–4): 331–351. Rizzello, S. (1999) The Economics of the Mind, Cheltenham: Edward Elgar. Vanberg, V. (1998) ‘Menger, Carl (1840–1921)’, in P. Newman (ed.), The New Palgrave Dictionary of Economics and the Law, Vol. 2, London: Macmillan, 635–641. Williams, W.M. and Yang, L.T. (1999) ‘Organizational Creativity’, in R.J. Sternberg (ed.), Handbook of Creativity, Cambridge: Cambridge University Press, 373–391. Williamson, O.E. (1979) ‘Transaction-Cost Economics: The Governance of Contractual Relations’, Journal of Law and Economics, 22: 233–261. Witt, U. (1998) ‘Imagination and Leadership – The Neglected Dimension of an Evolutionary Theory of the Firm’, Journal of Economic Behavior and Organization, 35: 161–177. —— (2000) ‘Changing Cognitive Frames – Changing Organizational Forms: An Entrepreneurial Theory of Organizational Development’, Industrial and Corporate Change, 9: 733–755. Ziegler, C.A. (1985) ‘Innovation and the Imitative Entrepreneur’, Journal of Economic Behavior and Organization, 6: 103–121.

8

Knowledge, understanding and the epistemology of innovative firms1 J. Stanley Metcalfe

Introduction My purpose in this chapter is not to survey scholarly claims about the origins, causes and consequences of innovation, the relation between innovation and the accumulation of knowledge, or the nature of management of the innovation process. Rather it is a statement of different viewpoints concerning the relation between innovation, the growth of knowledge, an evolutionary theory of economic and social change and our understanding of strategic management. The common ground of all these relations is the heterogeneity of the behaviour of individuals and organisations and the origins of this diversity in the heterogeneity of knowledge (Nelson and Winter 1984; Witt 1993; Nelson 1991). The issues addressed are particularly relevant to the theory of the firm and its capacity to acquire and articulate knowledge. Central to my approach to these matters is the idea that capitalism in equilibrium is a contradiction in terms. Like Schumpeter (1944), I argue that we understand modern capitalism most completely as a process of creative destruction, which never can be at rest by virtue of the manifold incentives and opportunities it provides to challenge established positions of economic influence. In this process, technology and organisation play a central role. Indeed they are evolving phenomena par excellence, the rate and direction in which they evolve being deeply dependent on the way in which a market economy and market institutions co-ordinate this diversity of behaviour to produce recognisable and ordered patterns of economic change. Capitalism is restless because knowledge is restless, and because the institutions of an open-market economy embody the incentives to challenge established positions and facilitate those challenges by making resources and custom available on equal market terms. Thus, what is distinctive about modern capitalism is the connection between the market process and the growth of knowledge. The purpose of this chapter is to throw a little light on this very complex question.

194

J. Stanley Metcalfe

Varieties of innovation research For many years now, the complexity of innovation phenomena has been documented in a wide range of scholarly studies. Each study typically takes its own frame of reference and unit of analysis, which can range across the genesis of individual innovations, the development of whole technologies interpreted as a sequence of related innovations; the behaviour of innovating organisations, typically firms; or the activities of individual innovators. All have found their mark in the literature, and it will help to begin with a small number of examples. In their path-breaking work, Carter and Williams (1957) undertook many case studies of innovations in British firms, which led them to a characterisation of the progressive firm in terms of no less than twenty-four dimensions. In this, they pre-figured much of the subsequent interest of innovation scholars. They drew attention to the importance of a firm’s external connections with the world of science and technology, the quality of its training and recruitment activities, and the degree to which selling policy and technical services to customers supported its innovative efforts. They also drew attention to the importance of trained, lively and receptive management and judged that the technically progressive firm would be of good quality in general. In reaching their conclusions, they had also noted the wide differences between firms in their innovative performance and the apparent lack of a clear relationship between progressiveness and performance. Nonetheless, they felt confident enough to conclude in a way that is thoroughly modern in its ring. The progressive firm linked its R&D closely with production and sales policies, it enjoyed close connections with suppliers and customers and it paid close attention in its recruitment policy to attract good minds. Success, they opined, built on success and, conversely, backwardness had a tendency to be self-perpetuating, the lowquality firm having neither the resources nor the reputation to attract and keep creative minds. The studies of Langrish et al. (1972) and Freeman et al. (1972) came to remarkably similar conclusions, in relation to the role of external connections and the internal connectedness of R&D with marketing activities. The recent study by Kunkle (1995), on the comparative records of GE and RCA at developing a commercial electron microscope in the United States, provides a clear-cut example of the Carter and Williams thesis. A different approach is taken in Georghiou et al. (1986) where the focus is not on single innovations but on sequences of related innovations, developed by a firm to explore a particular technological capability, to satisfy new applications and enhanced customer needs in existing applications. Each of the firms had been awarded the Queen’s Award for Innovation in the UK. Comparing cases, what appeared to be significant was the different scope for improvement within particular design configurations, together with the range of incentives to improvement, which in many cases

Knowledge, understanding and epistemology

195

could be linked to competitive or complementary developments in other technologies. Writing almost contemporaneously with Carter and Williams, Jewkes et al. (1969), in their study of fifty-seven important inventions, had also concluded that the forces making for technological creativity were too complex and intricate to provide any simple summary. While recognising that large and small firms would have comparative advantages in different types of innovation they also emphasised the continuing importance of the lone inventor whose motives and persistence could not be reduced to the simply economic. Like Carter and Williams they clearly saw and identified the great diversity which characterises innovative activity and drew the obvious implication that no one channel or institutional form could be expected to provide an ideal framework for innovation. Almost three decades later, Freeman (1994) has written a comprehensive survey of the innovation literature that draws upon many of the themes established by these earlier writers. Now, of course, the evidence is more impressive and he is able to address a wide variety of sources, distinguishing between the roles of external and internal learning in the innovation process. Now the progressive firm has become the learning firm, but the broad messages remain the same. Consider next an approach based upon case studies of organisational performance at innovation: Margaret Graham’s (1986) study of RCA’s attempt at inventing the consumer videorecorder – an attempt that ultimately failed. RCA had a strong record as a consumer electronics innovator and had the self-image of a technological leader, which it embedded in a separate research centre set up to drive the knowledge base free from pressures associated with operating units. Within RCA, multiple research teams followed different technological paths, so posing problems for the internal selection environment with conflict emerging between fundamental discipline-driven work and practical problem-solving activity. This tension defined the relation between the research centre and the consumer electronics division, and it is a tension familiar to the organisation of R&D. The route chosen by RCA was one of ten or so tried internationally by companies in the 1960s and 1970s, and it settled for a disk-based technology. The two technologies that ultimately triumphed, VHS and Beta-Max, were based on magnetic tape technology, design analogies to the cassette recorder not the record player. After a number of abortive product launches, the company made its final choice of design and technology in 1979 and launched this in the market in March 1981. By April 1984 the product was withdrawn: Japanese competitors had established an insurmountable market advantage based upon a technology that RCA had rejected as infeasible. Several interesting lessons follow from this study of a major, failed innovation in a very large company. The first is the combination of the management of the creative process, which generates the range of possible

196

J. Stanley Metcalfe

paths to innovation, with the management of the selection process, which ultimately concentrates effort around the chosen option. These coupled processes have to channel the tensions between unchallenged R&D driving the long-term knowledge base of the company in isolation from other aspects of strategy and the shorter-term customer- and problemdriven needs of the operating divisions. For whatever reason, RCA defined the ‘wrong agenda’ and placed its bets on the ‘wrong’ options. The technologies that triumphed were themselves subject to a fierce competitive process with the VHS design driving out the early market leader – the Beta-Max design. As Cusumano et al. (1992) show, the success of the VHS option did not depend on any technological superiority within the family of tape-based technologies. Rather, it followed from the ability of the companies concerned to accumulate the complementary marketing and productive assets necessary to build market positions more quickly than their rivals. This is, of course, the point about the importance of complementary assets that Teece (1987) has emphasised. What is also significant about these cases is the long timescale of the competitive process and the choices made by different firms in favour of radically different designs and conceptions of the likely market. From a different perspective, Henderson (1994) has explored the theme of strategic diversity in the pharmaceutical industry, focusing on the development of cardio-vascular drugs. In this sector, the crucial competitive issue is the ability to design compounds to produce specified effects on the cardiovascular system. The traditional knowledge base in this area was built around the technology of diuretics but, following a number of scientific breakthroughs, a range of new possibilities became available. The British company ICI successfully exploited the Beta-blocker route while other companies explored quite different design paths principally around calcium channel blockers and ACE inhibitors. Each of these design paths had a different rationalisation and provided the companies concerned with the opportunity to develop a family of closely related designs to control particular body mechanisms. A second theme, which Henderson explores, is the ability of companies to integrate quite different bodies of knowledge and their disciplines to achieve competitive advantage. In particular, she draws attention to the links that companies enjoyed with the external science knowledge base and the way in which their internal organisation integrated the wide range of disciplines required to take a drug from design to market. From our perspective, what is interesting about this study is that it also shows how companies can take very different and nonsubstitutable paths to serve the same market. They conjecture different things and what they conjecture depends on their interaction with external knowledge, together with the capacity to integrate it with internal knowledge. Historians of technology have provided quite different approaches to the study of individual innovations, where the unit of analysis is usually a

Knowledge, understanding and epistemology

197

well-defined artefact. Vincenti’s study (1990) of flush riveting in the construction of metal-skinned aircraft is a good case in point. This apparently simple innovation required an immense amount of trial-and-error learning before success was attained, with several companies following different routes to a solution, sharing that knowledge in a variety of ways and finally settling upon a uniform design solution. Science played no direct role in reaching this design; according to Vincenti it was reached entirely as a result of empirical engineering, the progressive search for a design by sequential trial and error, with the resulting knowledge not easily codifiable or communicable. Thus in building the industry knowledge base there was an element of collective invention with the movement of personnel and an important role being played by a group of key suppliers, the tool manufacturers. McBride’s study (1992) of the development of warship propulsion technology following the introduction of the Parsons turbine at the turn of the century provides another perspective. Three rival solutions to the problems, which arose in coupling a high-speed turbine to the ship’s propellers, were developed – reduction gearing, fluid drive and electric drive. Each had its own distinctive knowledge base. The adoption of electric drive brought a new entrant into the ship construction industry, General Electric, and, for a short period, this technology met the strategic needs of the US navy to improve the operating range of the Pacific fleet. Two factors led to the eclipse of this technology: (1) the need for faster ships to counter the growth of the Japanese fleet, and (2) improvements in metallurgy and machinery technology which jointly led to the re-adoption of the rival technology of reduction gearing. Here we see the significance of the processes by which rival options are selected and developed, together with the changes in relative technical and economic performance that occur in this process of competition. A final literature deals with the ability of firms to change technological direction in response to technological threats. Cooper and Schendal’s study (1976) was one of the first to explore this question. They found systematic support for the view that such changes are very difficult to accomplish; a corollary of which is the importance of new entrants in establishing radically different designs, a theme explored in greater depth in Utterback (1994). Technologies within organisations seem to generate their own inertia. New technology threatens skills and established accumulations of human and knowledge capital, and there is a clear tendency for management to emphasise the shortcomings of new technologies and downplay their potential advantages. When firms do attempt to develop the new technology, strategy and resources appear to be divided with a continuing commitment to the old technology, and often this division is reflected in organisational separation of the old from the new. It is this divided loyalty that reflects the difficulties of building new skills and strategic perspectives appropriate to the new technology.

198

J. Stanley Metcalfe

A subsequent study by Cooper and Smith (1992) further explores this theme and points to the multiplicity of technological and market strategies followed by new entrants and established firms. Of the twenty-seven threatened firms they studied, only seven made a successful transition to a new technology and built a strong competitive position. Clark (1994) provides an interesting case of this problem of technological rejection in a study of Bell Laboratories’ response to magnetic recording technology. In the 1930s, Bell built a considerable knowledge base and technological capability that enabled it to construct recording machines superior to those of the German competitor firms. Nonetheless, ATT’s management refused to sanction commercial innovation because they considered that recording would create a loss of privacy and reduce the demand for telephone calls. It was part of the ATT strategic model to control the entire telephone system and to prevent the attachment of devices, the use of which would be outside of its control. Consequently, when post-1945, the market for tape machines for telephone recording purposes was developed, Bell had lost its lead and was unable to respond to a market opportunity that was supplied by rival firms. We see in all these studies a number of themes central to the study of innovation in an evolutionary context. Put simply, resources and incentives to innovate do not appear to be a problem for industries as a whole, although they may be a problem for individual firms. Rather, the issues appear to centre on questions of opportunities and capabilities, and what is of central importance here is the fact that different competing firms follow a wide range of technological and strategic routes to compete in the same market. The ability to generate sequences of linked innovations appears to be of particular significance. As Nelson (1991) has emphasised, there are non-discretionary differences in firm behaviour at innovation, which cannot be explained in terms of economic factors. Individual firms coalesce around particular but different innovation trajectories, and it is these differences that provide the basis for an evolutionary theory of competition. Thus the central questions for innovation research are not whether firms are similar but whether and why they are different. The similarities cannot be the basis of competitive advantage – only the differences can play that role. Thus, how and why differences in behaviour arise is what must be explained. In the long term, every market position is vulnerable, and this is perhaps the most important lesson to be learnt from Schumpeter’s emphasis on creative destruction.

Rationality, imagination and innovation One of the great problems in understanding restless capitalism is the conjunction of innovation and rational decisions, for the two are related to quite different ways of thinking. For a theory of knowledge accumulation in the firm the connection between rationality and innovation strategy and

Knowledge, understanding and epistemology

199

performance is the key issue to be resolved. This is a complex matter, and no serious student of the innovation process has ever put much faith in the extreme idea of instrumental rationality, which is otherwise so prevalent in modern economy theory. Nonetheless, as the studies cited in the previous section indicate, nor is innovation a random process; it is managed and managed for a purpose. Surely, there is a fruitful middle way between Olympian grandeur and nihilistic despair. Let us start by accepting a far less restrictive version of rationality as reasoned behaviour: the directed, intentional behaviour of firms seeking competitive advantage by committing resources to innovation. Such behaviour is motivated, it involves striving (Winter 1984) and entails no more than decision-making and thus choice which seeks to do the best with the resources at the command of the firm or innovating business unit. The arguments in favour of reasonable rationality as distinct from Olympian rationality are I believe compelling, and they help us understand why firms come to innovate in such widely different ways. Olympian rationality, it will be recalled, requires a full specification of all the relevant options from which choices are to be made; a full evaluation of how each option contributes to the objectives of the rational decision-maker, including options whose consequences lie entirely in the future, and a full specification of the relevant constraints on the field of choice. In the field of innovation, none of these Olympian prerequisites seem remotely plausible; neither what is considered best nor the resources considered available are free from ambiguity. Thus, one must confront the complexity of innovative activity, its future orientation and inherent degree of non-quantifiable uncertainty in relation to events which must be considered unique (Shackle 1961). Indeed, to the extent that innovation is reducible to mathematical risk, there is a good case for saying that this cannot involve innovation at all. As the historian Joel Mokyr puts it, ‘innovations are singletons’. When problems become too complex to be well defined, let alone solved analytically, one is reduced to judgement and the guiding hand of experience (Lane et al. 1995). Complexity implies that there is no single metric to reduce choices to a comparable basis, that there is computational indeterminacy in translating between options and objectives, and an ineluctable range of ignorance – which means that decisions can only be made on the basis of expectation and hunch. Of course, many situations are complicated without necessarily being complex; what seems crucial to this distinction is that complex problems admit insights not answers, and that knowledge of the parts does not constitute knowledge of the whole (Bettis and Prahalad 1995). My conclusion from this discussion is that to the extent that optimisation occurs at all (i.e. reasonable behaviour), it does so within local domains in which the choice set and the evaluation of that choice set is idiosyncratic and historically contingent. Of itself, this is quite enough to underpin the diversity of firm behaviour, which is

200

J. Stanley Metcalfe

the chief empirical characteristic of the innovation process in modern capitalism. There is much more to bounded rationality than the notion that it is simply rationality with extra constraints. The much more fundamental issue is the question of the exercise of imagination and the construction of the innovation choice sets. Just what are the options for change? The idea that innovation is the ‘imagined deemed possible’ surely cannot be bettered as a statement of the essence of the problem (Shackle 1986). Langlois (1990) for example, along with many others in the Shackle (1961)/Simon (1955) tradition, has pointed out the fundamental problems that arise in compiling the list of potential choices, problems relating to the creative powers required to construct the relevant choice set. Within evolutionary epistemology, this is equivalent to the question of ‘Where guesses come from’ (Stein and Lipton 1984), a problem in the generation of variety and the subsequent selection of options from that range of possibilities (Campbell 1987). Two themes are particularly important here: the processes which give rise to the winnowing of ideas in a pre-adaptive sense, and the limitations on the generation of possible ideas via principles of guided variation. Stein and Lipton place particular emphasis on mechanisms for pre-adaptation, those heuristics and focusing devices, which shape the hypotheses, which are candidates for testing. They suggest that the heuristics can be of three general types: abstract rules or general principles of enquiry; concrete rules, specific ways of generating hypotheses in a discipline context; and exemplars, canonical solutions or concrete models of solutions to specific problems. The case of pre-stressed methods for bridge building provides clear examples of heuristics at each level. As Copp and Zanella (1993) write the story, the theory of bending moments provided the general principles of bridge design, while the concept of pre-stressing cables within concrete beams provided the specific route for generating hypotheses about how to offset loads acting on a bridge span. This generated a new way of building bridges with concrete beams containing pre-stressed cables, exemplified in a number of standard design features or models, which determined not only the form of the bridge (long, unsupported spans) but its manner of construction as well. What is important about this argument is that epistemic variations, which all innovations are at some point, while they may necessarily be blind in that their economic performance can only be anticipated, they are not random. The mental frameworks within which innovative ideas arise are heavily shaped by experience and by the limits imposed by the particular sets of heuristics in play. The search for novelty cannot be random for if it were it would fall victim to the tyranny of combinatorial explosion, there being too many possible combinations of human ideas to imagine and evaluate. If new ideas are to come, they come because search is

Knowledge, understanding and epistemology

201

directed to limited areas of the set of possibilities. In this sense innovations are never entirely novel, they are always prefigured in some of their dimensions. This emphasis on diversity of innovative behaviour raises the question of whether there can be a true model of the reality of any given innovative context, or whether there are as many divergent and valid representations as there are diverse imaginations of firms. Wilson (1990) tackles this question by distinguishing two approaches: where the generation of ideas is based on some model of reality, and where it is based on what he calls adaptive imaginary representations or fictional worlds, which simplify complex reality and provide a set of instructions (algorithms) on how to behave in particular circumstances. He places all models of rationality, including bounded rationality, in the first group. Such a fictional world limits and motivates behaviour and it is to be judged by its operational effectiveness, not by its relation to the truth – in the scientific sense appropriate to a model of reality. Thus if models of reality are ‘scientific’, adaptive imaginary representations are ‘technological’, like engineering knowledge they are tested by their utility, accumulated by trial and error, and they are the epitome of rule-guided behaviour. Technological design and development is more concerned with puzzle-solving than with hypothesis testing, more with verifying what works than with theoretical falsification. A technological puzzle is solved when the performance standards of an artefact are improved or become more predictable, not when a puzzle solution yields a better understanding of natural phenomena (Vincenti 1990). Moreover, technologies involve practice as well as knowledge, and this is why it is necessary to investigate the development of a technology in three dimensions: the growth of codified knowledge, the acquisition of tacit skills and the development and application of product and process artefacts (Layton 1974). It seems to us that managerial ideas about innovation and its organisation in the firm are more adaptive imaginary representations than models of reality, and in this they equate to technology and engineering, not science. History matters here because historical experience shapes the knowledge that is extracted from particular items of information. Hence the reason why firms so often get it wrong in their assessment of new technology: they appraise it through a grossly distorting mirror, the mirror of their past which in all likelihood stands ready to shatter in the face of significant innovation. In Schumpeter’s theory of innovation ([1912] 1961), it will be remembered that the entrepreneur is precisely distinguished by an ability to take leaps into the unknown, to act although ignorant of the consequences of action. Brian Arthur (1994) expresses the matter perceptively when he emphasises the resort to inductive reasoning and pattern recognition in the face of complexity. New information is evaluated by reference to previous experience, and this gives to the knowledge generation process an inherently path-dependent character.

202

J. Stanley Metcalfe

On the nature of knowledge and understanding In this section we explore the problem of knowledge and understanding as a basis for comprehending how different patterns of knowledge accumulation are reflected in different organisational structures. This is a problem central to the innovation process since it relates directly to the issue of what a firm can be said to know, if anything. Since few scholars would deny that economic and social change is contingent on the continued growth of knowledge, this is not a trivial issue. Who precisely is said to know more when we link the growth of knowledge to the development of the firm? The view taken here is foundationalist. Only individuals can know and what they know depends on perceptions, introspection, memory and inference; in short, experience allied with reason (Audi 1998). Individual knowledge is thus shaped, refined and continually moulded by the activities that individuals engage in during their lifetime and by the contexts that frame these activities. In this sense, individual knowledge is relative and, as the Austrian economists are wont to put it, ‘subjective’. What we know arises and develops partly in the context of our innate curiosity as human beings and partly via the stimuli provided by everyday experience of interaction with others. All these forms of knowledge are held in the subconscious recesses of our mind, they are essentially a matter of electro-chemical responses in the human brain with the consequence that mind is a store of latent potentialities. Indeed, we are conscious of knowing only when a problem or a discontinuity confronts us in our sense of the outside world. Personal knowledge is from this viewpoint a conversation with oneself, perhaps substantially unconscious. What we hold in our conscious mind at any point in time are organising principles and reference points from which to interrogate the subconscious when confronted by questions that capture attention (Witt 2002). It is our private way of making sense of the world, of distinguishing facts, of grouping related phenomena, of finding connections and of establishing cause and effect. Private knowledge is thus our ever-changing frame of reference for interaction with the wider world. As such, it is always incomplete and it is only because of this that learning and creative thought become possible. We can of course only conjecture a difference in knowledge against some existing frame of reference. Quite crucially for the following argument, these private states of mind are not accessible by any other individual mind and this carries with it an implication of great importance. Namely, that those processes, which as individuals we come to know, are augmented by social processes that permit exchanges of information (representations of knowledge) that are communicated between individuals such that they can lay claim to common understanding. The scarcity of individual mental capacity is to a degree overcome by the sharing of thoughts through the interchange of data and ideas. Through sense experiences of the knowledge of others, our

Knowledge, understanding and epistemology

203

individual states of knowledge become interdependent. In this way, private knowledge is correlated with but not identical to the private knowledge of others. An immediate implication of this distinction is that knowledge is always tacit, it is only information, the more or less imperfect representation of knowledge that can be codified in symbolic form. This extended reliance upon the testimony of others is one of the key factors in comprehending understanding as a complex system predicated on the knowledge of individuals, and indeed in comprehending the nature of capitalism as a knowledge-based system. It is also one of the most powerful reasons for organisation, the bringing together of minds to a common purpose. For it leads us directly to one of the most powerful of the ideas derived from Adam Smith – namely, the division of labour in the production as well as in the use of knowledge. Every economist is familiar with The Wealth of Nations and the story of the pin factory and the detailed specialisation within that work process. Yet within a few pages of this famous example, Smith turns to a far more powerful case, the case that underpins modern economies and indeed forms the test for a modern economy. For Smith applies the division of labour to the growth of knowledge, pointing to the role of those specialised philosophers and men of speculation, ‘whose trade is not to do any thing, but to observe everything; and who, upon that account, are often capable of combining together the powers of the most distant and dissimilar objects’ (Smith [1776] 1994: 11). Moreover, because the division of labour also applies to the philosophers, ‘Each individual becomes more expert in his own peculiar branch, more work is done upon the whole, and the quantity of science is considerably increased by it’ (ibid.). In this way, Smith reaches the core of the matter, identifying capitalism as a self-exciting system that has evolved and continues to evolve new methods of inventing. What Smith does not develop is how this growth of knowledge is to be co-ordinated, how individual knowledge is to be shared in the wider social context, for this will determine how individual knowledge will grow. What is the instituted process that achieves for knowledge activities what markets achieve for conventional productive activities? The answer from the innovation literature is that this process is framed by the organisation of the firm and the firm’s embeddedness in a wider matrix of knowledge sustaining organisations. To explore this answer further it is necessary to develop the distinction between personal knowledge and shared, interactive understanding in more detail. What Smith drew attention to is the individually idiosyncratic, specialised nature of personal knowledge; the corollary to this is that it is not only the use of knowledge but also the growth of knowledge that is embedded in a social process that is co-ordinated through appropriate patterns of social interaction. The problem here is clear: if knowledge remains private, it can inform private action but not social action. For social action to be possible, for actions to be mutually supporting and collaborative, it is necessary that private knowledge becomes public understanding to the

204

J. Stanley Metcalfe

requisite degree. The transmission of private knowledge into shared understanding is a socially distributed process and this process must depend on institutions for the sharing and common interpretation of flows of information. We can never say two individuals have the same knowledge, nor devise a way of establishing what they know. We can say instead that as individuals they have the same understanding in so far as they provide indistinguishable or at least closely correlated answers to the same question, or if they respond in indistinguishable ways to the same instructions. If information flow is to convey personal knowledge with sufficient accuracy to achieve commonality of understanding, then there must be common standards of communication, language or other forms of symbolic representation, and agreed standards for the justification of that which can be said to be known. Moreover, there must also be shared interpretative frames, theoretical schema to judge the content of information, otherwise private knowledge cannot develop into collective understanding. As Nelson puts it, there must be ‘social technologies’ to make testimony possible (Nelson 1999). In this regard, institutions and organisations matter in two fundamental ways in relation to the connection between knowledge and understanding. First, they constitute the means to store and communicate information in general and the means to support particular patterns of interaction, ‘who talks to whom with what frequency and with what authority’, in a society. This is the question of language, commensuration and symbolic representation in general. Different patterns of interconnection imply different distributed patterns of understanding and thus different paths for the growth of knowledge in specific contexts. Second, institutions and organisations embody the rules, the standards of socially agreed belief, that are the means to accumulate justifiably true knowledge in relation to science, technology, as well as organisation and social discourse. It is the institution of understanding in common that makes economic and social life possible while simultaneously constituting a powerful engine for the differential growth of personal knowledge. North (1990) is only partially correct in arguing that institutional rules constrain behaviour by facilitating the growth of common understanding. These same rules are to a degree enabling and facilitating in that the spread of understanding opens up opportunities for the further growth of private knowledge. To summarise the argument thus far, human interaction generates directly and indirectly a flow of information between individuals who, at best, treat that information as a representation of the knowledge of others, and interpret it through the confrontation with their own sense of knowing. In the process, the disjunctures that arise are a powerful stimulus to new thoughts. Information flow may change the knowledge states of the recipients, but there can be no expectation that the change of knowledge will be complete, that it will be identical for all recipients, or that it will be

Knowledge, understanding and epistemology

205

accepted. Indeed, it is the fact that information is not interpreted in uniform fashion that is essential to the continued growth of knowledge and thus understanding. A world in which all private knowledge was identical would be a world in which the problem of knowledge had ceased to exist. To speak of knowledge in general is to speak of the union of everything that is known privately, while to speak of understanding is to speak of that which is shared, that which is held to a degree in common. Like individual knowledge, understanding is an open system, it is emergent, it can grow combinatorially fast, at least in local domains, and it has no rest points or stable, invariant attractors. It can undergo subtle changes as information percolates across networks of relationships or it can undergo more radical, sweeping changes in a relatively short period and may take understanding into entirely new dimensions. To this degree understanding is unstable. Yet the kinds of understanding we develop and share are not unconstrained. The path of understanding is, we suggest, chreodic; it is channelled primarily by the particular nature of the organisational and institutional context. The conventions as to what information is made public, and who could communicate with whom about what, are deeply important for the growth of knowledge. Institutions, conventions and social standards can suppress the generation of understanding just as they can enable it. We can find here the unpredictability of knowledge accumulation – that is, the uneven nature of knowledge accumulation and innovation – and its corollary, the unpredictability of the surface forms of capitalism in terms of what is produced and consumed. From this viewpoint, it is natural to emphasise the role of personal creativity in the growth of understanding. Inventive genius associated with individuals or teams is crucial to the progress of knowledge in all its forms. However, creativity alone is not sufficient. What is personal at source must be capable of being placed in and spread within the public domain, and here the development and embodiment in organisations of information and communication technologies have been of paramount importance in permitting the creation, storage and transmission of symbolic records. Yet this technological view leaves the account incomplete. Equally crucial has been the invention and adoption of standards or norms to distinguish reliable from less reliable knowledge. The process of establishing error, of identifying mistakes, is absolutely essential to the growth of understanding. Criteria for falsification and rejection have provided the critical edge that combats the problem of superfecundity, the problem of being unable to distinguish which of the too numerous rival courses of action to follow. In regard to science, we enquire of the truth of the relation between conjecture and natural fact. In regard to technology, we enquire whether the device works in the environment in which it is intended such as to achieve the desired effect. In regard to business, we ask does the plan achieve the profitability required to justify its continuation. Without these instituted

206

J. Stanley Metcalfe

and thus shared winnowing processes, it is not at all clear how knowledge and understanding can grow. Thus, all knowledge and all understanding reflect multiple processes of trial and error of variation and selection. We would agree with Campbell (1987) when he argues that all growth of knowledge is predicated upon a process of blind variation and subsequent selection. It should be noted that ‘blind’ does not mean ‘random’; rather, it means that the validity of new knowledge can never be known in advance since those tests of validity always lie in the future. These considerations tell us a great deal about the unique properties of capitalism as a knowledge-based system and of the role of the firm as the locus of innovative activity. Capitalism is a system for generating business experiments based on the accumulation of scientific, technical, organisational and market knowledge. Business conjectures create sequences of new problems to be solved, give meaning to entrepreneurship, and give to the firm as organisation the unique role of combining the multiple knowledge elements that are needed to innovate successfully. In being a problem-solving system, any organisation such as a firm is necessarily a problem-generating system. Hence the restless nature of firms and whole economies: capitalism can never be at rest because understanding and knowledge are never at rest and never can be; an ‘equilibrium of knowledge’, is an oxymoron. By this notion of restlessness, we intend to convey two separate meanings: first, that economic actors are to a degree ill at ease, they can never be sure that their position will not be challenged; and second, that they are always searching for new situations and the route to these is the growth of knowledge. Of course, it is important to emphasise that the growth of understanding cannot be random without further processes to focus that randomness to good effect. Unaided, random systems do not evolve – they drift. We make rapid economic progress precisely because the underlying processes of variation are guided, because they explore only limited regions of the space of possibilities. Nor can the growth of understanding be entirely deterministic for it involves choice, judgement and creativity in the sense of the expansion of thought and action into new dimensions and spaces. For this reason alone, uncertainty is an unavoidable element in the picture for only when the space of possibilities is closed can probability judgement and its formal calculus be entertained. Hence, the importance of the rationality issue referred to above. Not only are the winnowing processes referred to above crucial, but the possibility of the recombination of ideas and concepts is essential to the cumulative growth of personal knowledge and shared understanding, for what is important about combinatorial processes is that they build on memory and thus upon experience. Only then is it possible to entertain the idea of cumulative learning and cultural transmission. What seems to be unique about capitalism as a knowledge-based system is that it has evolved a system of knowing and understanding that is in reality a system for standing on the shoulders of giants.

Knowledge, understanding and epistemology

207

The capabilities approach We turn now to the knowledge-based view of the firm, and it will be clear that this is not an easy concept to deal with. It is a familiar line of reasoning that the firm is defined by what it knows; its knowledge determines what it can do and how. In recent literature this has become coterminous with the capabilities view of the firm and the knowledge-based theory of the firm (Nonaka and Takeuchi 1995; Itami 1987). Both of these approaches are of considerable interest in any study of the sources and conduct of innovation-related activities: for the essence of innovation is that it requires a change of knowledge within the firm (Kogut and Zander 1992). However, in light of our previous argument, we need to pause at this point and enquire more carefully into the idea that organisations can know. Taken literally this would imply that they can enjoy sense experiences, that they can reflect, that they possess memory and that they can reason and infer. Now one immediate objection to this is the claim made above that knowledge is an attribute of individuals and individuals alone. However, epistemologists also recognise that what individuals believe also depends upon the testimony of others; that it is possible, indeed essential, in the presence of the division of knowledge, to share information – that is to say, representations of what is known by individuals. Firms are organisations in which the testimony of other team members matters crucially, and thus firms embody instituted, shared practices such as language and trust and agreed procedures that permit the sharing of reliable knowledge. Thus, to understand the ‘knowledge of the firm’ one must know how that firm is organised and how its patterns and language of communication are articulated. Arrow (1974) refers to this organisation-contingent aspect of the extension of understanding in terms of the communication codes that are defined by particular organisations. Moreover, and this is exactly Penrose’s point, capabilities depend upon team activity in which the knowledge and skills of the individuals are transformed into the synthetic understanding of the organisation. Thus the organisation, in all its aspects, becomes an operator for creating the collective from the individual. Since these phenomena are often less than transparent to the external observer (and are certainly complex, social processes), this has given rise to the idea of causal ambiguity. In short, the link between capabilities and competitive performance is not readily decipherable, even perhaps by the management of the firm to which its operations are partly a black box (Rumelt 1984). Of course, this raises particular difficulties when a firm comes to choose which other imperfectly understood organisations to engage with in cooperative arrangements to generate new capabilities (James and Metcalfe 1999; Lado et al. 1997). This is a matter far broader in its impact than a concern with innovation and its management; it reaches to the core of the distinction between firms

208

J. Stanley Metcalfe

and markets and the nature of the competitive process. Egidi (1996) makes this clear in his claim that firms and markets are not alternative ways of achieving the same end. They are complements, not substitutes, even though the boundary between them may be a matter of conscious choice in Coasian fashion. Market co-ordination is about the selfconsistency of plans, not the creation and implementation of plans; that is the role of the firm. McNulty (1984) makes the same point when he insists that the unique province of the firm is to decide what is produced, its quality and its method of production. As he points out, this does not depend, per se, on team production and it is as valid for the single-person firm as it is for any other. It is the firm as the locus of the transformation process; that is the role that markets cannot fulfil. However, McNulty’s point is strengthened when we enquire why the division of labour leads to team production. The point about ‘production’ is that it requires the synthesis and combination of many different kinds of knowledge and thus the integration of teams of specialised (and semi-ignorant) individuals. This is what we mean by organisation. Correspondingly, there is a minimum scale to the firm, defined by the multiple kinds of understanding it needs to integrate in order to function effectively. Notice, also, that this does not rule out the firm using information generated outside of its boundaries. The external organisation of the firm and the ability to turn external information into internal knowledge, absorptive capacity, are crucial aspects in understanding the modern firm (Coombs and Metcalfe 2000). The nature of this synthetic capability cannot be independent of the design of the firm and, as Bausor (1994) points out, design is an unfinished process. This gives the firm a unique characteristic: to be the locus for the combinatorial combination of multiple kinds of knowledge in the execution of its productive purpose. The implications of this view of the ‘sentient firm’ are reflected most obviously in the developing capabilities and resource-based perspectives of the firm (Foss and Knudson 1996; Montgomery 1995). As with many other contributions to this literature, the starting point is Edith Penrose’s (1959) view of the firm as an administrative unit, which generates bundles of productive services from bundles of physical and human productive factors. This perspective presupposes organisation, and purpose, and the existence of a theory of the particular business to provide the framework, which justifies and evaluates the operation of that business.2 Each of these elements plays a role in the capabilities perspective, as we shall explore below. Richardson (1972) provides the second point of departure. He presents firms as sets of complementary, interdependent activities, which for their operation require knowledge, expertise and skills in their performance. In other words, capabilities are the abilities to carry out specific actions. Richardson begins by distinguishing broad activities, such as production, marketing, research and development, and progressively refines and sub-divides these so that they relate increasingly to the execution of

Knowledge, understanding and epistemology

209

specific, narrow tasks. Alternatively, one can think of capabilities at different levels from top strategic management, the top competent team (Eliasson 1990; Castanias and Helfat 1991), down through various levels of operation to the ‘shop-floor’. The point is that capabilities are bundles of behaviours, they are pluralistic, multi-dimensional, interdependent, and they differ according to the function and level at which they are identified. For example, in her influential book Wellsprings of Knowledge, LeonardBarton (1995) distinguishes core capabilities from supplemental and enabling capabilities, and goes on to distinguish physical systems, skills, managerial systems and cultural values as elements that shape capabilities. Small wonder that their precise measurement at this general level proves to be so elusive. I now explore the nature of capabilities more carefully, beginning with a distinction between resources, routines, organisation and intent. Some of the major contributions to the literature have espoused a resource-based view of the firm to explain firm heterogeneity and emphasise the importance of a lack of inter-firm resource mobility (Barney 1991; Amit and Shoemaker 1993). We can explain this more clearly by distinguishing production factors, employees, capital and intermediate goods, and organisational structures from the factor services or resources, which are derived, in Penrose fashion, from the factors. Now the point is that the resources are not intrinsic properties of the factors; what each productive resource flow is depends upon the other resource flows it is combined with, the context in which that combination takes place, and the purpose behind that combinative process. The same skilled employee may be very productive in one function, and quite unproductive in another; in one firm, she may work successfully, in another firm she may not. It is the ability to work together that matters, and this is a matter of collective understanding. Consequently, while the production factors are divisible and tradable, the collective bundle of interdependent, complementary services is not; it is a unity premised on understanding and not knowledge, which, if it is to be traded in the market for corporate control, must be traded as an entity. Here it is important to recognise that the firm, tout court, is not always the appropriate unitary entity that is traded. The more significant concept is the business unit, that particular organisation charged with articulating a particular set of transformation processes addressing a specified product market. This is why we observe far more trading in business units in the market for corporate control than we do trading in entire companies. Indeed, an active market in corporate control is an extremely important complement to the process of capability formation; it is vital if the right incentive structure is to be created to support the often highly uncertain process of capability formation It should be clear that the collective capability represented by the bundle of resources is dependent on the way in which the firm organises its many activities. The organisation is an operator for translating the

210

J. Stanley Metcalfe

individual resources into a collective bundle – the capability. One need not stress here how important this dimension can be, or that it has formal and informal components in terms of the social capital of patterns of trust, reputation and authority, which are created continuously within the firm (Erikson and Mikkelson 1996). Furthermore, the focus on services and activities leads naturally to the question of what is done by whom and how; that is to say, the question of routines (Nelson and Winter 1984). Routines are stabilised patterns of behaviour but of widely differing degrees of specificity. Some routines are precise templates for action, specifying to the ‘nth’ degree what must be done in each set of conceivable circumstances. Others, typically those associated with higher management, may offer far greater degrees of interpretative flexibility, some indeed may be routines for defining new routines, or for improving existing ones. This assembly of concepts promises a rich theory of the internal workings of the firm and a basis for linking this with competitive advantage. Fundamental to all of this is the idea that it is understanding that is the common denominator to tie the capability into a co-ordinated unity. Members of the firm know; the organisation turns that knowledge into collective understanding. However, we also need to recognise the final element in the capabilities perspective – that of intent and purpose; for it is intent, combined with capabilities, that gives rise to productive opportunities (Loasby 1994). Now intent implies a ‘theory of the business’, what shall be done and how, a framework which allows the various resource services to be combined to a purpose. Such a framework, a paradigm, a hardcore business programme, an image or a dominant logic identifies productive opportunities, guides action and decision, provides tests of performance and acts as a filter in relation to the accumulation of new internal or external knowledge (Fransman 1994). In determining which questions are asked, the framework necessarily defines which questions are off the agenda. Here we find one of the strong themes in the capabilities literature: that firms are subject to inertia – it is not only physical and organizational capital that is sunk but also human intellectual capital, as the members of the firm develop their distinctive patterns of activity. The theory of business generates constraints as well as opportunities: as with all knowledge, a competence in a particular activity brings with it a trained incapacity to think beyond the region of specialist expertise. Thus a capabilities perspective speaks not only as to why firms are different but also as to why they develop in different ways, and how that development is channelled in specific directions and is unresponsive to even major shifts in the competitive environment. All this, it will be recognised immediately, is a natural consequence of specialisation and the division of labour. The acquisition of expertise in one domain forecloses its acquisition in some other domains, which is part of the reason why such investments are risky and subject to unplanned obsolescence. If one accepts this argument about organisation and collective under-

Knowledge, understanding and epistemology

211

standing, one is of course in danger of ignoring too much the role of the knowledgeable individual. This would be a mistake. The knowledge base of the firm is necessarily the union of what its individual members know; firms are composed of individuals and indeed any serious discussion of entrepreneurship emphasises that point (Hughes 1971; Douglas 1987; Millard 1990). Therefore, this line of argument is not intended to deny the importance of the individual or of knowledge but rather to emphasise that individuals always operate in an organisational context. However distinctive the individual may be, the context is still important. Clearly, in closely coupled organisations there is far less scope for individual creative behaviour than in more open, loosely coupled organisations. It is a familiar theme in the management literature as to where one draws the balance, close coupling being beneficial for efficiency and control but loose coupling being beneficial for innovation and adaptation (Loasby 1991). It is here that the Richardson and Penrose themes come together and complement our earlier discussion. The process by which understanding is gained inevitably generates imbalances in the knowledge of the firm’s members, imbalances that disrupt or fall short of the requirements of a close complementarity of capabilities required for the operation of the firm. Knowledge is acquired which no longer fits with the existing capabilities and if this slack is to be exploited it can only be done by attending to the gaps in understanding which prevent its classification as a capability. By contrast, the imbalance can be interpreted, not as an under-utilised capability but as a ‘missing’ capability needed for the optimal exploitation of other capabilities which are present. The restless firm finds itself continually faced with emergent opportunities to bring together complementary but discriminate bodies of expertise, to create capabilities where imbalances existed previously. The crucial point, though, is that it is not axiomatic that internal generation and exploitation of the capabilities is always the most effective option. Thus the options of internal and external acquisition become different but complementary routes to the same end; namely, to exploit imbalances in knowledge, skill and expertise. The issue is one of careful judgement, of the appropriate choice of external and internal methods of accumulation, and of the effective integration of the two methods. This connects to the idea of the distributed and decentralised accumulation of competences in the economy. That is, to the division of labour in the accumulation of knowledge and expertise and the co-ordination of those accumulation processes in systems of interacting organisations, as in the innovations system perspective (Nelson 1992; Carlsson 1995; Edquist 1997). In sum, factors external and internal to the firm drive the process of capability formation. The activities of all expertise-generating organisations creates for each of them underutilised expertise and thus the opportunities to engage in co-operative arrangements to exploit that expertise and to develop any further expertise necessary to the fulfilment of the objectives of the co-operative agreement.

212

J. Stanley Metcalfe

Issues and questions Having devoted so much time to these general themes, the reader will rightly expect me to draw some implications for the study of innovation and its consequences. What are some of the lessons? What are some of the corresponding questions? A first point is that what constitutes a bundle of capabilities cannot come from the properties of a firm alone but from the way in which that firm stands in a competitive relation with the core capabilities of rival firms. All capabilities are relative, one firm is only ‘better’ by virtue of comparison with a firm which is ‘worse’. Capabilities are not absolutes, and this is why different capabilities give rise to differential rents in the competitive process (Winter 1995; Peteraf 1993; Montgomery 1995). Is it then possible to identify and evaluate capabilities as attributes of the single business? To what extent are they sectoral, to what extent are they national? Second, and following the same theme, what constitutes valuable knowledge in a firm is not independent of the organisational context in which that knowledge is articulated. Even when faced with the same external information different firms come to understand different things. Thus, to an important degree, knowledge is a latent capability: in some contexts, it may be the basis for market-leading performance, in others for performance that is not best practice. I am inclined to think that this difference cannot be traced to individual resources or routines but rather to the way in which everything is brought together: to the integrating properties of the organisational operator and to the intent of the firm. Do we really know what makes the signature of one firm different from that of any other? It is clear that the capabilities theme leads to the firm as an experimental organisation, successful by virtue of its ability to generate a sequence of innovative conjectures that create market advantage. The set of ideas, which is generated internally, together with the methods by which options are selected for development, are key aspects of behaviour. It is difficult enough to make one innovation, even more so a sequence or a shift to a new design configuration. Industrial history is littered with firms who failed to make the necessary evolutionary steps and fell by the wayside, first relatively and then absolutely as they were forced out of the market (Utterback 1994; Georghiou et al. 1986). The fundamental fact to comprehend is that innovation-based competitive advantages are transient, permanently threatened by creative destruction. From this perspective, we can comprehend many of the tensions that figure in the day-to-day formulation and implementation of technology strategy; they all impinge upon the firm as an experimental organisation. Should R&D be centralised or decentralised? Are the needs of existing customers an accurate guide to future innovation options? How is an innovation by a competitor

Knowledge, understanding and epistemology

213

to be appraised? What can be learned from benchmarking a rival organisation? Finally, I want to say a little about the clear danger in an ‘internalist approach’ to the creativity of the firm, the danger in overlooking that the business unit is embedded in a wider matrix of organisations which support the generation of technology (Coombs and Metcalfe 2000). Constant’s (1980) idea of the community of practitioners is still perhaps the best way to sum this up. Thus in Freeman’s (1994) survey the external connections of the firm with a wider knowledge base are given considerable emphasis, and this is also the theme which underpins the idea of innovation systems and technology support systems. However, this is not a new idea; industrial districts, Marshallian external economies, and informal knowledge trading (Von Hippel 1988) are familiar examples in which capabilities spread beyond the boundary of the individual organisation. Of course, this is a reflection of how knowledge accumulates when the generation of knowledge has become subject to a division of labour. Hence, the processes of knowledge accumulation internal to the firm must be augmented through interorganisational arrangements, such as joint R&D ventures and mergers and acquisitions. The questions become more complex where we consider the impact of mergers and acquisitions on capabilities and strategy. While comparatively few changes in corporate ownership are specifically to acquire technology, they all carry implications for the innovation process. How is due diligence of intangible technological assets to be carried out? How is the R&D capability to be reorganised within a unified selection environment? If there are technological synergies, how are they to be turned from prospect into reality? Each of these is familiar to managers involved in the bringing together of formerly distinct businesses, each with its own pattern of understanding.

Conclusion My purpose in this chapter has been limited to considering three themes which, taken together, I believe are having a major impact on the way that scholars approach the study of innovation. These themes were rationality and imagination, private knowledge and organised understanding, and capabilities in relation to the accumulation and use of knowledge. Together they lead to an epistemological view of the firm and the market (Metcalfe and Boden 1994); to the firm as a creative experimental organisation embedded in a wider network of interactive knowledge-generating relationships. The firm is what it believes it is. It does what it understands, and what it understands depends on what its members know and on how they interact through organisation. It can only innovate by changing what it believes, and understands. Most fundamental of all, it is a view that underpins our understanding of the differential innovative behaviour of

214

J. Stanley Metcalfe

firms and the essential evolutionary dynamic, which defines modern competition. It is a view which I believe demands a great deal more attention both conceptually and empirically, and which is fundamental to comprehending the nature of innovation processes in an increasingly integrated and restless world economy.

Notes 1 I wish to thank Rod Coombs and Ronnie Ramlogan for allowing me to draw on our joint work in writing this chapter. 2 Penrose refers in this context to the image of the business opportunity; others have proposed similar framing concepts (e.g. the dominant logic, proposed by Bettis and Prahalad 1995). The point about any framing concept is its durability and relative imperviousness to the information it generates and its capacity to connect together different kinds of information.

References Amit, R. and Schoemaker, P.J.H. (1993) ‘Strategic Assets and Organizational Rent’, Strategic Management Journal, 14: 32–46. Arrow, K. (1974) The Limits of Organisation, New York: W.W. Norton. Arthur, W.B. (1994) ‘Inductive Reasoning and Bounded Rationality’, American Economic Review, 84 (May): 406–411. Audi, R. (1998) Epistemology, London: Routledge. Barney, J. (1991) ‘Firm Resources and Sustained Competitive Advantage’, Journal of Management, 17: 99–120. Bausor, R. (1994) ‘Entrepreneurial Imagination, Information and the Evolution of the Firm’, in R. England (ed.), Evolutionary Concepts in Contemporary Economics, Ann Arbor: University of Michigan Press. Bettis, R.A. and Prahalad, C.K. (1995) ‘The Dominant Logic: Retrospective and Extension’, Strategic Management Journal, 16: 5–14. Campbell, D.T. (1987) ‘Blind Variation and Selective Retention in Creative Thought as in other Knowledge Processes’, in G. Radnitzky and W. Bartley (eds), Evolutionary Epistemology, Theory of Rationality and the Sociology of Knowledge, New York: Open Court. Carlsson, B. (ed.) (1995) Technological Systems and Economic Performance, Dordrecht: Kluwer Academic Publishers. Carter, C. and Williams, B.R. (1957) Industry and Technical Progress, London: Oxford University Press. Castanias, R.A. and Helfat, C.E. (1991) ‘Management Resources and Rents’, Journal of Management, 17: 155–171. Clark, M. (1994) ‘Suppressing Innovation: Bell Laboratories and Magnetic Recording’, Technology and Culture, 34: 516–538. Constant, E.W. (1980) The Origins of the Turbojet Revolution, Baltimore, Md.: Johns Hopkins University Press. Coombs, R. and Metcalfe, J.S. (2000) ‘Organising for Innovation: Co-ordinating Distributed Innovation Capabilities’, in N. Foss and J. Mahnke (eds) (2000), Competence, Governance and Entrepreneurship, Oxford: Oxford University Press.

Knowledge, understanding and epistemology

215

Cooper, A.C. and Smith, C.G. (1992) ‘How Established Firms Respond to Threatening Technologies’, Academy of Management Executive, 6: 55–70. Cooper, A.C. and Schendel, D. (1976) ‘Strategic Responses to Technological Threats’, Business Horizons, 19: 61–69. Copp, N.H. and Zanella, A.W. (1993) Discovery, Innovation and Risk, Cambridge, Mass.: The MIT Press. Cusumano, M.A., Mylonadis, Y. and Rosenbloom, R.S. (1992) ‘Strategic Manoeuvring and Mass Market Dynamics: The Triumph of VHS over Beta’, Business History Review, 66: 51–94. Douglas, S.J. (1987) Inventing American Broadcasting 1899–1922, Baltimore, Md.: Johns Hopkins University Press. Edquist, C. (1997) Systems of Innovation: Technologies, Institutions and Organisations, London: Pinter. Egidi, M. (1996) ‘Virtuousness and Adverse Selection in Economic Organisations’, IIASA, mimeo, Laxenberg. Eliasson, G. (1990) ‘The Firm as a Competent Team’, Journal of Economic Behaviour and Organization, 13: 273–298. Erikson, B. and Mikkelson, J. (1996) ‘Competitive Advantage and the Concept of Core Competence’, in N.J. Foss and C. Knudson (eds) Towards a Competence Theory of the Firm, London: Routledge. Foss, N.J. and Knudson, C. (eds) (1996) Towards a Competence Theory of the Firm, London: Routledge. Fransman, M. (1994) ‘Information, Knowledge, Vision and Theories of the Firm’, Industrial and Corporate Change, 3(3): 713–757. Freeman, C. (1994) ‘The Economics of Technical Change’, Cambridge Journal of Economics, 18: 463–514. Freeman, C. et al. (1972) Success and Failure in Industrial Innovation, Science Policy Research Unit, Brighton, Sussex: University of Sussex. Georghiou, L., Metcalfe, J.S., Gibbons, M., Ray, T. and Evans, J. (1986) PostInnovation Performance, London: Macmillan. Graham, M.B.W. (1986) The Business of Research: RCA and the Video Disk, Cambridge: Cambridge University Press. Henderson, R. (1994) ‘The Evolution of Integrative Capability: Innovation in Cardio Vascular Drug Discovery’, Industrial and Corporate Change, 3: 607– 630. Hughes, T.P. (1971) Elmer Sperry, Baltimore: Johns Hopkins University Press. Itami, H. (1987) Mobilizing Invisible Assets, Cambridge, Mass.: Harvard University Press. James, A. and Metcalfe, J.S. (1999) ‘Capabilities and the New Theory of the Firm’, in N. Foss and P.L. Robertson (eds), Resources Strategy and Technology, London: Routledge. Jewkes, J., Sawyers, D. and Stillerman, R. (1969) The Sources of Invention (2nd edn), New York: Norton. Kogut, B. and Zander, U. (1992) ‘Knowledge of the Firm, Combination Capabilities and the Replication of Technology’, Organization Science, 3: 383–396. Kunkle, G.C. (1995) ‘Technology in the Seamless Web: “Success” and “Failure” in the History of the Electron Microscope’, Technology and Culture, 36: 80–103. Lado, A.A., Boyd, N.G. and Hanbu, S.C. (1997) ‘Competition, Co-operation, and

216

J. Stanley Metcalfe

the Search for Economic Rents: A Syncretic Model’, Academy of Management Review, 22: 110–141. Lane, D., Malerba, F., Maxfield, F. and Orsenigo, L. (1995) ‘Choice and Action’, Sante Fe Institute (95-01-004). Langlois, R.N. (1990) ‘Bounded Rationality and Behaviourism: A Clarification and a Critique’, Journal of Institutional and Theoretical Economics, 146: 691–695. Langrish, J., Gibbons, M., Evans, W.J. and Jevans, F.R. (1972) Wealth From Knowledge, London: Macmillan. Layton, E.T. (1974) ‘Technology as Knowledge’, Technology and Culture, 15(1) (January): 31–41. Leonard-Barton, D. (1995) Wellsprings of Knowledge, Cambridge, Mass.: Harvard Business School Press. Loasby, B.J. (1991) Equilibrium and Evolution: An Exploration of Connecting Principles in Economics, Manchester: Manchester University Press. —— (1994) ‘The Organization of Knowledge and the Organization of Industry’, mimeo, University of Stirling. McBride, W.M. (1992) ‘Strategic Determinism in Technology Selection: The Electric Battleship and U.S. Naval–Industrial Relations’, Technology and Culture, Vol. 33, No. 2 (April): 248–277. McNulty, P. (1984) ‘On the Nature and Theory of Economic Organisation: The Role of the Firm Reconsidered’, History of Political Economy, 16: 233–253. Metcalfe, J.S. and Boden, M. (1994) ‘Paradigms, Strategies and the Evolutionary Basis of Technological Competition’, in Swann, P. (ed.), New Technologies and the Firm: Innovation and Competition, London: Routledge, 83–102. Millard, A. (1990) Edison and the Business of Innovation, Baltimore, Md.: Johns Hopkins University Press. Montgomery, C. (1995) Resource-Based and Evolutionary Theories of the Firm, Dordrecht: Kluwer Academic Publishers. Nelson, R.R. (1991) ‘Why No Firms Differ, and How Does it Matter’, Strategic Management Journal, 12: 61–74. —— (1992) ‘National Innovation Systems: A Retrospective on a Study, Industrial and Corporate Change, 2: 347–374. —— (1999) ‘On the Uneven Evolution of Human Know-How’, mimeo, Columbia University. Nelson, R.R. and Winter, S. (1984) An Evolutionary Theory of Economic Change, Cambridge, Mass.: Harvard University Press. Nonaka, I. and Takeuchi, H. (1995) The Knowledge-Creating Company, Oxford: Oxford University Press. North, D. (1990) Institutions, Institutional Change and Economic Performance, Cambridge: Cambridge University Press. Penrose, E. (1959) The Theory of the Growth of the Firm, Oxford: Basil Blackwell. Peteraf, M.A. (1993) ‘The Cornerstone of Competitive Advantage: A Resource Based View’, Strategic Management Journal, 14: 179–191. Richardson, G.B. (1972) ‘The Organization of Industry’, Economic Journal, 72: 883–896. Rumelt, R. (1984) ‘Towards a Strategic Theory of the Firm’, in R. Lamb (ed.), Competitive Strategic Management, New York: Prentice-Hall. Schumpeter, J.A. ([1912] 1961) The Theory of Economic Development, New York: Oxford University Press.

Knowledge, understanding and epistemology

217

—— (1944) Capitalism, Socialism and Democracy, London: Unwin. Shackle, G.L.S. (1961) Decision, Order and Time in Human Affairs, Cambridge: Cambridge University Press. —— (1986) ‘The Origination of Choice’, in I. Kirzner (ed.), Subjectivism, Intelligibility and Economic Understanding, London: Macmillan. Simon, H.A. (1955) ‘A Behavioral Model of Rational Choice’, Quarterly Journal of Economics, 69: 99–118. Smith, A. ([1776] 1994) The Wealth of Nations (Cannan edn), New York: The Modern Library. Stein, E. and Lipton, P. (1984) ‘Where Guesses Come From: Evolutionary Epistemology and the Anomaly of Guided Variation’, Biology and Philosophy, 4: 33–56. Teece, D.J. (1987) ‘Capturing Value from Technological Innovation: Integration, Strategic Patterning and Licensing Decisions’, in H. Brooks and B.R. Guide (eds), Technology and Global Industry, Washington: National Academy Press. Utterback, J.M. (1994) Mastering the Dynamics of Innovation, Cambridge: Harvard Business School Press. Vincenti, W.G. (1990) What Engineers Know and How They Know It, Baltimore, Md.: Johns Hopkins University Press. Von Hippel, E. (1988) The Sources of Innovation, Cambridge, Mass.: The MIT Press. Wilson, D.S. (1990) ‘Species of Thought: A Comment on Evolutionary Epistemology’, Biology and Philosophy, 5: 37–62. Winter, S. (1984) ‘Schumpeterian Competition in Alternative Technological Regimes’, in R. Day and G. Eliasson (eds), The Dynamics of Market Economies, Amsterdam: North-Holland. —— (1995) ‘Four Rs of Profitability: Rents, Resources, Routines, and Replication’, in C. Montgomery (ed.), Resource-Based and Evolutionary Theories of the Firm, Dordrecht: Kluwer Academic Publishers. Witt, U. (1993) Evolutionary Economics, London: Edward Elgar. —— (2002) ‘Competition for Scarce Attention – Mass Media, Information Dissemination and Economic Choices’, mimeo, Max Planck Institute for Research into Economic Systems, Jena.

9

Managing exploitation and exploration Bart Nooteboom

Introduction In the literature on organisational learning a distinction is made between first and second order learning (Hedberg et al. 1976; Fiol and Lyles 1985) or, equivalently, between ‘single loop and double loop’ learning (Argyris and Schön 1978). The first is learning to do existing things better (more efficiently) and the second is learning to do new things. This is linked with the notion of ‘parametric’ change (Langlois and Robertson 1995) as opposed to ‘architectural’ change (Henderson and Clark 1990). Also related to this, March (1991) and Holland (1975) distinguished between ‘exploitation’ and ‘exploration’. In order to survive in the short term, firms need to exploit their present resources (or competencies or abilities) efficiently, and to survive in the long term firms need to develop novel competencies, to anticipate or create future market conditions. This combination of exploitation and exploration is arguably the main challenge for management, but it is a paradoxical task. To a greater or lesser extent, depending on the type of product, market, technologies and types of knowledge and competencies involved, exploitation requires fixed standards, routinisation and tight co-ordination, while exploration requires a loosening of structural ties and conditions. In this chapter I aim to contribute to the solution of this paradox. For that I make use of a recent book (Nooteboom 2000). Exploration entails discovery, which is subject to radical uncertainty, in the sense of Knight (1921). In other words, it goes beyond risk, which is associated with a known, closed set of possible alternatives to which one can attach a probability distribution. The set of options to choose from is open, and often options are discovered or created after, not prior to, action. Options are often options to discover further options. This requires a logic or heuristic of ‘abduction’ (Peirce 1957; Holland et al. 1989): how do we explore options that are unknown? How do we arrive at new hypotheses that have some chance of viability? Of all the novel ways of doing things that we can think of, which should we try, and how do we find out what other, as yet unknown options there are? We might be tempted by this uncertainty to see search as blind. And,

Managing exploitation and exploration

219

indeed, one way to proceed would be to engage in random novel combinations, and doubtless some of this is going on. But we are thinking animals that make inferences about the future on the basis of our experience with the past. Consider, for example, the famous scenario planning by Shell oil company. It entails the development of contingency plans on the basis of the analysis of alternative policies under different possible futures. Robustness of elements of policy under different possible futures provides a reasonable guess for contingency plans. However, while we can think of many logically possible future worlds, we not only lack knowledge on their likelihood but we have no way of knowing whether we have thought of all possible futures, and we cannot be certain that the futures we have thought of contain the actual future. In particular, the future is difficult to predict because actions will have unforeseeable consequences, and there will be strategic reactions to our actions from others. We are playing games whose participants, strategies and pay-offs are revealed only as the game is played, and then shift in the process. The future will be different from any of the ones imagined, but nevertheless one may have developed a platform for viable strategies, with capabilities in place to execute them. Thus, discovery goes beyond search among existing options, to include the creation of new options. We need to solve the problem of abduction: how can we make steps into the unknown, in exploration, while preserving existing resources in such a way that exploitation is maintained? How do we set about creation with a minimum of destruction? What is the optimal process of discovery? I will propose a heuristic to move from present competence to novel competence, while surviving in the process. But first I specify the theory of knowledge and learning that I use, and I consider the environment in which firms strive to survive.

Learning For knowledge I take a social constructivist, interactionist view. The term ‘knowledge’ here is a broad one, and denotes any mental activity, including perception and value judgements. People perceive, interpret and evaluate the world according to mental categories (or frames or mental models) which they have developed in interaction with their social and physical environment. This is enabled but also constrained by the construction of neural nets, building on bodily functions developed in a shared evolution, so that there is a basic structural similarity of cognition between people (Lakoff and Johnson 1999). However, this construction of cognition takes place on the basis of interaction with the physical and social environment, which varies between people. This entails that perception, interpretation and evaluation are path-dependent and idiosyncratic to a greater or lesser extent. Different people see the world differently to the extent that they have developed in

220

Bart Nooteboom

different social and physical surroundings and have not interacted with each other. In other words, past experience determines ‘absorptive capacity’ (Cohen and Levinthal 1990). In social science this approach is linked with the views of G.H. Mead (‘symbolic interactionism’). In developmental psychology it is linked, up to a point, with the work of Piaget and, more fully, with the work of Vygotsky (1962). In cognitive science it is linked with the emerging, non-mainstream view of ‘situated action’ (as opposed to the mainstream ‘computational-representational’ view – see Shanon 1988, 1990, 1993; Hendriks-Jansen 1996). In economics it is linked with the notion of subjectivism in the Austrian school: different people not only have different preferences, but ‘different minds think different things’. The crux of this view is, as proposed by Piaget and Vygotsky, that intelligence is internalised action and speech, and that both knowledge and meaning are context-dependent. This context-dependence links with the Austrian school in economics, and particularly Hayek’s view that a variety of ‘local’ knowledge is distributed across heterogeneous contexts. I propose that categories develop from interaction, and this is how competition, or markets more generally, constitute a Hayekian ‘discovery process’. Summing up, the term ‘constructivist’ indicates that intelligence is internalised action. The term ‘social’ or ‘interactionist’ indicates that since one cannot ‘climb down from one’s mind’ to assess whether one’s knowledge is properly ‘hooked on to the world’, the variety of perception and understanding offered by other people is the only source one has for correcting one’s errors. As discussed in Nooteboom (1992) an implication of this view for the theory of the firm is that in order to achieve a specific joint goal the categories of thought of the people involved must be aligned to some extent. Different people have a greater or lesser ‘cognitive distance’ between them (Nooteboom 1999a). This yields the notion of the firm as a ‘focusing device’, to reduce cognitive distance; that is, achieve a sufficient alignment of mental categories, to understand each other, utilise complementary capabilities and achieve the goal. Organisations develop their own specialised semiotic systems: language, symbols, metaphors, myths, rituals. This is what we call organisational culture. This differs between organisations to the extent that they have accumulated different experiences, in different industries, technologies and markets. This connects with the idea, in the organisation literature, that the crux of the firm is to serve as a ‘sensemaking system’ (Weick 1979, 1995), ‘system of shared meaning’ (Smircich 1983) or ‘interpretation system’ (Choo 1998). I propose that this yields a more fundamental reason for firms to exist than the reduction of transaction costs, although transaction costs are also part of the story. One interpretation of entrepreneurship, which links with Schumpeter’s notion of the entrepreneur as a charismatic figure, is that it is his central task to achieve this: to align perceptions, understandings and goals. Note that alignment of cognitive categories need not entail identity. As dis-

Managing exploitation and exploration

221

cussed in Nooteboom (1999a) there is a trade-off between cognitive distance, needed for variety and novelty of cognition, and cognitive proximity, needed for mutual understanding. In fact, different people in a firm will to a greater or lesser extent introduce elements of novelty from their outside lives and experience, and this is a source of both error and innovation. A second implication is that by the need to achieve a focus there is a risk of myopia: relevant threats and opportunities to the firm are not perceived. To compensate for this, people and firms need complementary sources of outside intelligence to utilise ‘external economy of cognitive scope’ (Nooteboom 1992). Here again the trade-off arises between cognitive distance, for the sake of novelty, and cognitive proximity, for the sake of understanding and utilisation of complementarity. This perspective is well suited to the prevalent idea in the literature on innovation systems that innovation derives primarily from interaction between firms (Lundvall 1985, 1988, 1993). The present theory yields a prediction that is opposite to classical transaction cost economics: with increasing uncertainty, in terms of volatility of technology and markets, firms should not integrate activities more, as transaction cost theory prescribes, but less, because the need to utilise outside complementary cognition is greater. Here, the prediction is that firms will engage less in mergers and acquisitions and more in intensive alliances at some cognitive distance, but with sufficient durability and intensity to achieve mutual understanding and co-operation. Boundaries of the firm are determined, in part, by the need for cognitive proximity, next to reduction of transaction costs

Selection and firm strategy Perhaps it is useful, up to a point, to take an evolutionary perspective (Nelson and Winter 1982). Firms strive to survive in the ‘selection environment’, consisting of markets and other institutions. However, the question is how independent the selection environment is from the behaviour it is supposed to select. What if behaviour, in turn, creates and selects the institutions that are part of the selection environment? This is related to the controversy, in the management literature, between the ‘positioning’ view versus the view of ‘strategic choice’. The positioning view, attributed in particular to Porter (1980, 1985), derives from the ‘structure–conduct–performance’ perspective in industrial economics: market structure determines conduct, and conduct determines performance. The underlying assumption, as generally in mainstream economics, is that technology and demand are given and that firms find themselves in an established field of competitive forces in which they should find an appropriate niche. This perspective has been criticised from the ‘competence’ perspective for its neglect of ‘strategic choice’ and entrepreneurial abilities

222

Bart Nooteboom

to transcend competition for existing scarce resources by the creation of novel resources and to distinguish a firm from its competitors by means of firm-specific competencies. The difference between these perspectives is sometimes exaggerated. Of course a firm cannot create any environment and any competencies it likes. It will need to make entrepreneurial use of windows of technological and institutional opportunity, and will need to overcome internal and external obstacles to change. Nevertheless, the notion of strategic intent and the scope for entrepreneurial shifts of technology and preferences to alter the field of competitive forces yield a useful shift of perspective on the nature of the selection environment in evolutionary economics. The issue of the role of institutions, and the extent to which they can be modified by the behavioural units that they select, is close, if not identical, to the issue of structure and agency in sociology (Giddens 1984; Archer 1995). How do social and cultural structures condition (enable, constrain) actions, and how do these actions reproduce, elaborate or transform those structures? Of course, in biological evolution also, forms of life affect the environment in which they are selected, by foraging, creating refuse, etc., and they may to some extent select their niche by migration. There is the notion of co-evolution: two species co-evolve if they exert selection pressures upon each other. We find this when species prey on each other or live in symbiosis. The notion of co-evolution is useful, up to a point, but in the context of human systems it is also seriously misleading. Perhaps it does make sense to say that the selection environment of markets and (other) institutions co-evolves with firms or routines or competencies. However, in socioeconomic evolution the influence of firms on selection by competition is of an entirely different, non-biological, cognitive and linguistic order. Here we are dealing with invention, communicative interaction, social construction of selection conditions, alliance formation, mergers and acquisitions, and political manoeuvring. To put it more precisely, in biological co-evolution the environment sets the selection conditions for the species, and the species affects the selection conditions for its environment. In human systems, however, the species (a form of organisation, or routine, or competence) not only affects the selection conditions for the institutions in its environment, but by cognitive and communicative ingenuity may dodge or directly affect the selection conditions that the environment sets for it. That is a fundamental difference, and it makes biological analogies suspect.

Cycle theory of innovation In the organisational and economic literatures there is a stream of thought that suggests that innovation proceeds according to a ‘cycle’ with two stages. An initial stage of volatility, with the creation of Schumpeterian

Managing exploitation and exploration

223

‘novel combinations’, and a later stage of consolidation, with ‘dominant designs’ (Abernathy 1978; Abernathy and Utterback 1978; Abernathy and Clark 1985) and efficient production systems that employ economies of scale and routinisation. The cycle is generally held to imply a shift from product to process innovations, as product forms settle down and competitive pressure shifts to efficient production. Note that standardisation and utilisation of scale economies do not necessarily imply ‘Fordist’ production. Standards can allow for a wide scope of differentiated production and still be standards, requiring a certain amount of control and co-ordination. Take, for example, the consultancy firm Arthur Andersen, which is proposed as a paradigm example of a flexible firm. Its consultants are highly autonomous, employing their individual knowledge, skill and creativity to provide custom-made advice. But even there, attempts are made to safeguard professional standards and consistent quality. Scale is exploited by requiring consultants to contribute their experience to a common pool, and to work together, which requires a certain amount of standardisation of definitions and procedures. Almost any type of efficient production will require some amount, no matter how limited, of routinisation and standardisation, of actions, output, skill, knowledge or information (Mintzberg 1983). The life cycle theory of innovation has been complemented with the product life cycle theory of internationalisation (Vernon 1966). According to this theory, the consolidated innovation, which originated in countries with advanced technology and demand, is ‘generalised’; that is, carried to less developed countries with lower wages in order to further extend the market and fight the competition by a further decrease of costs. The life cycle theory of innovation suggests that the first, volatile stage of novel combinations requires decentralised, disintegrated organisational forms such as industrial districts of small, independent firms (Piore and Sabel 1983), or firms with a decentralised ‘organic’ structure, while the stage of consolidation requires a more integrated, bureaucratic structure. In other words: the degree of organisational integration depends on the stage in the innovation cycle. This connects with a long tradition in the organisational literature to propose that stable, predictable environments require integrated, ‘mechanistic’, bureaucratic forms of organisation, while volatile, varied and unpredictable environments require disintegrated, ‘organic’ forms of organisation (Burns and Stalker 1961; Chandler 1962; Emery and Trist 1965; Thompson 1967; Lawrence and Lorsch 1967). In the present discussion, ‘integrated structure’ is synonymous with ‘mechanistic’ and ‘bureaucratic’ structure, and ‘disintegrated’ structure is synonymous with ‘organic’ structure. It is useful to distinguish between integration of ownership (in the sense of claims to residual profit) and integration of decision-making. The most extreme form of organisational disintegration is autonomous units, under separate ownership, who engage in pure, arm’s-length market transactions. Separately owned units can

224

Bart Nooteboom

become more or less integrated, in the sense of losing autonomy of decision-making and freedom of choice, by more or less constraining governance of inter-firm relations. These can entail more or less long-term, more or less detailed contracts; mutual dependence by dedicated investments or exchange of hostages, or other switching costs; routinisation; building of trust. Next, units can become integrated under unified ownership, within a single firm. Within the firm, units can be more or less integrated in terms of centralisation and hierarchical control. This depends on the tightness, rigidity and type of co-ordination. For types of co-ordination one might use Mintzberg’s (1983) typology of co-ordination by direct supervision, standardisation of action, of output, of skills or knowledge. The cycle theory has met with empirical contradictions. Among other things, often process innovation precedes rather than follows product innovation. But my main objection to the cycle, perhaps related to the empirical anomalies, is that this cycle is not really a cycle. A genuine cycle leads back to the beginning. Like evolutionary theory, existing theory tells us how exogenously generated novel variety settles down. The origins of novelty remain a mystery, and that is precisely what we would like to understand. How does the discovery process work? Later, a distinction was made between ‘competence enhancing’ versus ‘competence destroying’ innovations (Abernathy and Clark 1985). The former favour incumbent firms and the latter new entrants. This was confirmed, for example, in a study of technological discontinuities in the US cement, airline and minicomputer industries by Tushman and Anderson (1986). However, while this theory of technological discontinuities in ‘punctuated equilibria’ suggests an explanation of how the discontinuities come about, such explanation is not forthcoming; they are simply assumed to arise somehow, and remain as exogenous as they were in earlier theory. Thus Gersick (1991) states: The definitive assertion in this paradigm (i.e. punctuated equilibria) is that systems do not shift from one kind of game to another through incremental steps: such transformations occur through wholesale upheaval. The discussions (in previous sections of the article) should help to explain why incremental changes in system’s parts would not alter the whole. As long as the deep structure is intact, it generates a strong inertia, first to prevent the system from generating alternatives outside its own boundaries, then to pull any deviations that do occur back into line. According to this logic, the deep structure must first be dismantled, leaving the system temporarily disorganised, in order for any fundamental changes to be accomplished. Next, a subset of the system’s old pieces, along with some new pieces, can be put back together into a new configuration, which operates according to a new set of rules. (Gersick 1991: 19)

Managing exploitation and exploration

225

This is unsatisfactory for several reasons. First, we should get our sources right. According to Gould and Eldredge, the originators of the theory of punctuated equilibrium, punctuation is not to be confused with saltationism: the idea that change arises from some discontinuous transformation (Eldredge and Gould 1972; Gould 1989). It arises from allopatric speciation: small populations, isolated at the periphery of the range of their parents develop into a new species. There is no discontinuity in this; just a development that is abrupt only on a geological timescale. Second, if apparent discontinuities occur this does not eliminate the need to explain where they come from. Third, the crucial question is when the ‘deep structure’ breaks down, and, above all, how one finds out what ‘subset of old pieces’, and what ‘new pieces’ one should ‘put back together’, and according to what ‘new configuration’. Thus, again, we need to find a procedure of abduction; a procedure of exploration which indicates how novelty comes about, in such a way that it can be combined with exploitation, in order to survive on the way to novelty.

A heuristic of discovery Tomas Kuhn (1970) proposed that a certain amount of conservatism in theory is rational: counter to what Popper was supposed to have prescribed it is not rational to drop any investment, including investment in theory, whenever the first indication (‘falsifier’) arises to prove that it is not perfect. In fact, scientists engage in solving puzzles within the purview of ‘normal science’, within its dominant ‘paradigm’, until the cumulative weight of anomalies becomes ‘excessive’, and then novelty generally comes from outside. In fact, Popper agreed that a certain amount of theoretical tenacity is rational, because ‘otherwise we will never find out where the real strengths of our theories lie’ (Popper [1970] 1976: 52). Expanding on these insights I proposed that, like crime, discovery is guided by motive, opportunity and means (Nooteboom 1992, 1999b). One needs an accumulation of unsatisfactory performance to generate motive; to overcome one’s own inertia or that of others in an organisation. In markets, one also needs an opportunity of demand and/or technology. And one needs insights into what novel elements to obtain from what source and how to incorporate them in present competence. I propose that one can obtain such conditions only by moving one’s present competencies across a variety of contexts (‘generalisation’), adapting them to local conditions (‘differentiation’), seeking interaction with novel conditions and people, to adopt elements of novelty from them (‘reciprocation’). That is how we obtain motive, opportunity and means for change. Generalisation often needs to be preceded by ‘consolidation’, to find out what precisely it is that one will transfer to novel contexts, and to do so efficiently. This often requires the codification of knowledge that at first

226

Bart Nooteboom

was tacit. This entails abstraction: the loss of context-specific knowledge in order to escape from the narrow context and provide a basis for generalisation. In other words: it is in part due to loss of information, to escape from context, that we are able to learn from application in a variety of new contexts. Application requires re-contextualisation, where explicit knowledge is embedded again in local, tacit knowledge. This introduces elements for possible novel combinations. Transfer to novel contexts often requires standardisation for the sake of division of labour and co-ordination. The need for this depends on how systemic the activity or technology involved is. Novel combinations, by grafting elements from outside practice, yield syncretic structures that induce pressure for more radical ‘architectural’ innovations. I will specify this in more detail later. This, I propose, is how a discovery process in markets might work. The basic principle is an alternation of variety of form and variety of context: variety of form is reduced and replaced with variety of context that generates novel variety of form. Discovery proceeds by applying knowledge in novel contexts, where we encounter unexpected rival or complementary elements of knowledge that provide the material for novel combinations. For technology novel context entails a new area of application, and for products it entails a novel market or market segment. This cycle is illustrated in Figure 9.1. Note that the cycle is proposed as a heuristic, i.e. a rule that generally tends to contribute to the goal of preserving exploitation while conducting exploitation. It is not proposed as an inexorable march of logic that is necessary, uniform and universal. It is not necessary: innovation can occur on the basis of purely random trial and error. That, however, is likely to be associated with excessive waste and failure. The cycle is not inexorable: CHAOS

Exploration opening variety of content

Novel combinations Reciprocation

Consolidation

Closing variety of content Exploitation

Differentiation

Generalisation INERTIA

Opening variety of context exploitation and exploration

Figure 9.1 Cycle of exploitation and exploration

Managing exploitation and exploration

227

development may get stalled. In particular, after consolidation development may get stuck in inertia. People and organisations may be unable to unlearn and to surrender the routines and skills they have developed and which are demonstrably efficient. In markets, however, such inertia is vulnerable to new entry of entrepreneurs, unless they are barred from entry. Thus firms may decline due to inertia or may be jolted out of it by threats of competition. In the stage of radical innovation, novel combinations may get lost in a chaos of competing systems that fail to lead to any dominant design. In the stage of consolidation, the importance of fixed standards for the sake of efficient production and generalisation depends on the type of product, technology, knowledge and market. The whole process of discovery may even be unintentional: generalisation is generally not driven by the will to learn but by an expansionist, imperialist drive to obtain profit or power from applying successful practice to wider contexts. There are many contingencies in the process, and some will be discussed later. Consider product change in more detail. As products are offered in novel markets, one has an opportunity to find out where its limitations lie, in lack of fit with newly discovered characteristics of demand. That yields motive and opportunity for ‘differentiation’. A trade-off is involved here between the opportunity in terms of a higher profit margin for tailor-made products and the possible opportunity cost of loss of economy of scale. This trade-off depends on the type of product, technology and market. Next, one may find out how in those respects in which one’s own product fails, the competing product or comparable products perform better. That yields the means of ‘reciprocation’. This is logically equivalent to metaphor in language: one practice is seen in the light of another. Nonaka and Takeuchi (1995) provide the example where in the development of a bread-baking machine inspiration is sought from the skill of an expert cook. Alternatively, one may find novel opportunities in local supply of labour, materials or technology to improve the product or its process of production. A famous example is how Henry Ford’s idea of an assembly line in car manufacturing was inspired by the procedure, at a mail-order company, in which boxes on a conveyor belt passed successive stations, to be filled according to order slips. Syncretic add-ons of outside elements, in reciprocation, labour under one or more of the following problems. Complexity of ad hoc add-ons increase architectural complexity (‘spaghetti’) in syncretic structures, which yields problems of co-ordination and decreasing returns from further add-ons. Duplication of parts in different places of the architecture forgoes opportunities for economy of scale. Above all, initial success of novel elements can be achieved only in niches where they fit into established structures that constitute the local selection environment. Such a structure may be the architecture of the practice itself, or of structures of use, or superordinate structures of distribution channels, legal acceptance,

228

Bart Nooteboom

vested interests, etc. This explains why often novelty first emerges in other niches than the ones where it is later most successful. From an evolutionary perspective, we might see this as the ‘allopatric speciation’ that yields punctuated equilibrium, along the lines proposed by Gould. As success emerges in the niche, pressures arise for more radical architectural changes, again in the product itself or its structures of use, or the superordinate architectures, to allow the novelty to fully realise its potential. Here the niche which served for the incubation of novelty is expanded, and novelty creates its own selection environment. Such more radical architectural innovation, on different levels of structure, creates confusion, creative destruction and a great deal of uncertainty. This is where a novel ‘technical trajectory’ or ‘techno-economic paradigm’ may arise (Dosi 1982, 1984; Dosi et al. 1988; Freeman and Perez 1989). Such architectural change is not random: one indication for it is to design architecture such that novel elements that were proved useful in the preceding stage of reciprocation can better realise their potential. But multiple interests are at stake, and strategic interaction can have unpredictable outcomes. Here we are back at the beginning of the cycle: a process of consolidation is needed. Completion of the cycle explains, among other things, that while process innovation may follow product innovation, the reverse can equally be the case. Note that the cycle appears to solve the problem of exploitation and exploration, at least in part. By applying current competencies in novel contexts we preserve exploitation, needed to survive in the process of discovery, while at the same time contributing to exploration: the accumulation of motives, opportunities and means for change. A criterion for search is to look for novel contexts that are sufficiently close to maintain exploitation and sufficiently distant to offer significant novelty. This connects with the trade-off between cognitive distance and proximity discussed before In earlier work (Nooteboom 1999b, 2000) I employed the notion of scripts to develop a hierarchy of innovations and institutions, on the basis of a hierarchy of structures. A script is an architecture of nodes that represent component activities in more or less rule-guided behaviour. Substitutions into nodes represent alternative ways to perform component activities. Scripts may denote mental concepts or procedures (Abelson 1976; Shank and Abelson 1977), organisational routines, primary processes of production and distribution, industry supply chains. Component activities in an (organisational) script are based on (cognitive) subscripts, and (organisational) scripts substitute their output into superscripts (supply chains). Parametric innovation with respect to a given script entails a novel substitution into a given node; a minor architectural innovation is a reversal of orders of nodes or the creation of alternative branches for different contexts. A more radical architectural innovation is

Managing exploitation and exploration

229

the complete reconfiguration of nodes reconstituted from different parallel scripts.

A cycle of integration and disintegration As in the classical innovation cycle, the extended cycle is associated with integration and disintegration of organisational forms; that is, with the strengthening and loosening of linkages of co-ordination. Generally, after a movement of integration in the process of consolidation and generalisation, there is a movement of disintegration, first gradual and later more radical, in the stages of differentiation and reciprocation. This is needed to deal with variety of information, in the generalisation to different contexts (cf. Chandler 1962; Stinchcombe 1990), and to allow for flexibility of differentiation and novel combinations. Disintegration can take several forms: decentralisation within the integrated firm, increased autonomy for subsidiaries, outsourcing, management buy-outs, spin-offs, sell-offs, novel entry of independent firms and the formation of industrial districts. These predictions are in line with empirical research which has found that innovation is associated with greater ‘complexity’ (which here would be called ‘variety’) and organic structures (here associated with ‘disintegration’, i.e. decentralisation) (Hage 1998). The cycle of (dis)integration is illustrated in Figure 9.2. In the cycle of innovation, how, more precisely, do the stages connect with (dis)integration? Taking into account the role of distance: novel combinations are promoted by a constellation of separate, relatively small, weakly connected, spatially proximate units in complementary activities (‘industrial districts’ or autonomous units in large firms). In such constellations, a number of requirements are satisfied. Sufficient cognitive Industrial districts Loose intra-firm networks

ENTRY

EXIT

Mergers/acquisitions

Novel combinations

Alliances

Reciprocation

Consolidation

DISINTEGRATION

INTEGRATION

Differentiation

Generalisation

Independent subsidiaries

Figure 9.2 Cycle of integration and disintegration

Tight intra-firm networks

230

Bart Nooteboom

proximity is needed to be able to understand each other, and trust is needed to do without complex, detailed, costly, constraining contracts, and to contain risks of spill-over. They are achieved on the basis of shared norms and values of conduct, an efficient reputation mechanism, the ‘shadow of the future’ from expected dealings with each other in the future, shared routines (Nooteboom 1999a). Sufficient cognitive distance is achieved by variety in activity and experience. There is sufficient spatial proximity to allow for frequent and varied contacts, and for intensive interaction in partial joint production, needed for the transfer of tacit, procedural knowledge, which is characteristic of the early stages of innovation. Here, competition is not on price but on novelty. Opportunity is also related to the absence of disadvantages of disintegrated structures that arise at other stages of development. Small, independent units are not so good at orchestrating many parts of a system to innovate in tune (Teece 1986, 1988; Chesbrough and Teece 1996), but since we are dealing here with radical innovations, in ‘novel combinations’, which break up existing systems, that yields no obstacle. Tacit, procedural knowledge has the disadvantage of lacking the basis for formalised procedures with documented communication that is needed for large-scale production, with specialisation in different departments. But at the early stage of innovation both the opportunity and the need for large scale are absent: no opportunity because the market is still small; no need because due to initial monopoly pressure from competition is weak. Large, integrated firms can survive or indeed create the discontinuities of novel combinations by means of decentralisation of highly autonomous divisions or even individual ‘intrapreneurs’. This is discussed in more detail in a later section. But there are limits to the variety that can be created and sustained in a large firm. How can one foresee the kinds of variety that might become relevant? In the extreme case, to create that variety the large firm would have to engage in practically everything, allowing for any combinations, and what then remains of the notion of an organisation? It seems necessary to also maintain a readiness to mop up successful small innovators, in order to tap into a variety of independent firms that would be hard to reproduce within the firm. And to benefit from their advantages of integration large firms must also maintain a capability for systemic alignment, with strong ties, in the later stages of consolidation and in the stage of generalisation. In this way it is conceivable that a large firm combines the best of two worlds. While it is not easy to perform this balancing act, it is conceivable, and indeed appears to be achieved by firms such as the 3M company and INTEL. However, an illustration of how difficult this is given by the recent federalisation of IBM, which was instituted to compete with more flexible, specialised and independent firms. In the stage of consolidation, with the search for a dominant design, it is important that there is flexibility to try out various combinations and forms, and that misfits are efficiently weeded out. Here also lies the

Managing exploitation and exploration

231

strength of the variety and idiosyncrasy of small, independent units, and the fact that misfits cannot be kept alive by cross-subsidisation from successful products in a portfolio. Here we run into a second restriction to the mimicry of industrial districts by large firms: the efficiency of the elimination of failures becomes doubtful due to the possibility of propping them up with cross-subsidisation. In that sense they are not efficient from a societal perspective, but that of course still leaves the possibility for their existence. Depending on the selection environment of markets and institutions such practices are not necessarily weeded out. Examples of small firms running ahead in commercialisation are: semiconductors and computer-aided design (Rothwell and Zegveld 1985), micro-computers (Langlois and Robertson 1995) and self-service retailing (Nooteboom 1984). In the stage of generalisation, after consolidation, integrated structures are better at large-volume production and distribution of novel products in wider world markets. A dominant design has emerged. Tacit, procedural knowledge has been developed into declarative, documented knowledge, which allows for transfer across larger distances. Standards allow for disintegration while maintaining fit across interfaces in chains of production. At the same time, increase of scale is feasible with the growth of demand, and is necessary to reduce costs due to increasing competition, as patents wear out and imitation increases. Competition has shifted from novelty to price. This favours larger production units, integrated distribution channels, spreading of risks, access to finance and the umbrella of a brand name, on the basis of penetration into extended markets. This favours a larger, more international and more integrated firm. Integrated structures are also better at the development of more coherent systems of connecting technologies, distribution systems, industry structures, supporting infrastructure, technical and commercial standards, yielding the configuration of a novel techno-economic paradigm (Freeman and Perez 1989). Next, as generalisation turns into differentiation and reciprocation, comparative advantage shifts again to greater variety, in more autonomous divisions, subsidiaries or independent firms, to give room for the generation of variety by reciprocation, in preparation for the next round of more fundamental innovation. Differentiation of products and processes also contributes to an escape from pure price competition between identical products that developed from generalisation. Small firms, or independent units within large firms, are better at product differentiation in niche markets where they do not run into disadvantages of small scale and can benefit from flexibility and proximity to customers.

232

Bart Nooteboom

Contingencies It was noted before that the cycles of discovery and (dis)integration are not offered as inexorable, necessary and uniform. A complete discussion of contingencies that affect the cycles is beyond the scope of this chapter, and I can consider only a few (cf. Nooteboom 2000). The speed of the cycle, for example, depends on what level we are investigating: individual, firm, industry, techno-economic paradigm. Higher-level cycles contain many iterations of cycles on lower levels. On the level of idea formation by individuals the cycle can be quite fast: it can turn around in an hour or a day. Product cycles vary enormously. In financial services and some fashion goods the cycle can be a year; in cars, computers and machine tools two to four years; in major construction projects five to seven years; in pharmaceuticals and telecom infrastructure ten to fifteen years (Quinn 1992). The key question is not only how long the cycle is, but whether there is a mismatch between the cycle for the product and the cycle for production. Generally, production systems have long cycles if they are embodied in large sunk investments in the form of dedicated hardware (such as factories), and short cycles when they entail craft production with tools that can easily be replaced or professional work on the basis of knowledge or skill that can easily be updated. If the cycle of the production system is long, due to a large fixed sunk cost in hardware, and the product cycle is short, there may be a problem. This problem can be solved at least to some extent if novel products can be made by novel assemblies of components, according to an enduring technology of assembly, or if the production system has the flexibility to adapt product forms in small batches, because it is programmable. According to the study of technological discontinuities in the cement industry by Tushman and Anderson (1986) it took thirteen years to move from the Rotary kiln to the Edison long kiln (1896–1909) and sixty years to move to the Dundee kiln with process control. In the airline industry it took twenty-two years to move from the generation of the Boeing 247, Douglas DC-2 and DC-3 (with the DC-3 as the dominant design, in 1937) to the era of the jet aircraft, with the Boeing 707 (1959), and next ten years to the wide-body jets, with the Boeing 747 (1969). In the minicomputer industry it took only two years to move from transistors to integrated circuits (in 1964), and seven years to move to semiconductor memories. The speed of movement to integrated circuits derived from the strong pressure to eliminate the constraints that limited the realisation of the potential of semiconductors imposed by the assembly of different components of different materials. The duration and prominence of different stages within the cycle can also vary considerably. This depends, for example, on the intensity of competition and its pressure for change. It also depends on how important

Managing exploitation and exploration

233

market fluctuations, product differentiation and economy of scale are, and of what type the economy of scale is. Product differentiation depends on the type of product and customer. In fashions in prosperous countries product differentiation is essential. Engineering economy of scale is crucial in process industries. Economy of scale in the use of information technology has declined enormously with the advent of micro-computers and user-friendly software. In some markets there is enormous economy of scale in marketing, and in brand name, advertising and distribution. When product differentiation is crucial, and economy of scale is limited or absent, and no major discontinuities in production technology occur, industrial districts can last for long periods of time. This is to be expected in fashion goods, where automation is difficult in some activities involved, such as cutting and assembling clothes, which yields a limit to economy of scale, while product differentiation and speedy response are crucial. On the other hand, when economy of scale is crucial and the product is hardly differentiable, large, tightly integrated companies can persist for a long time. Illustrations of the longevity of industrial districts are found in abundance in Italy (Malerba 1993): in fashion, shoes, furniture. These satisfy the conditions: great importance of differentiation, low production economies of scale. As indicated by Malerba, a problem may be that information technology may to some extent be competence destroying, because for effective use it cannot be simply attached to existing production but requires its redesign. It may require a change of production scripts. Another consideration is that, as indicated above, there can be large effects of scale in marketing, in distribution and brand name. The prediction would be that in fashion-oriented industrial districts there is, or will be, a tendency for the emergence of central, and perhaps dominant, parties that provide this marketing. This is confirmed by the case of Benetton. Here the economies of scale in marketing and brand name are combined with economy of scale in the provision of the ICT network which co-ordinates flexible, differentiated production with speedy and efficient response to shifts in fashion. An illustration of the longevity of large, tightly integrated firms is the oil industry. This satisfies the conditions of large economy of scale, in the exploration and production of crude oil, the spread of political risk in having access to crude oil at diverse locations, and the refining and distribution of oil, while the product is hardly differentiable. The prediction is that those firms will be in trouble when product differentiation does appear, as is happening in the chemical industry. Combination of oil with chemicals made sense from the perspective that the feedstock for most chemicals is a derivative from oil (naphta), as long as the market was dominated by bulk products that fitted the large-scale process mode of production. But the rationale is disappearing in the short term due to the emergence of low-volume, high added-value chemical specialities and

234

Bart Nooteboom

‘designer chemicals’, which require a shift from undifferentiated bulk production far from the market to close market interaction for low-volume specialised products. In fact, there have been intensive debates in the Shell oil company on whether the organisation should effect a turnaround in this direction. It seems that the resistance to this has prevailed, probably rightly so: the transformation would be too big.1 The expectation now is that Shell will divest its chemicals division. In the longer term, however, threats emerge also for oil: solar energy, which entails an entirely different production technology and distribution structure; hydrogen fuel, which does not fit the installed base for distribution of oil. Some markets have a dual structure: a large segment for standardised products and small niche markets for specialised, differentiated products. Examples are clothing and shoes. For the first segment one would expect more integrated and for the latter more disintegrated structures. And this is indeed what is found.

Firm strategy We have seen that exploration and exploitation can be reconciled along the cycle of discovery: one can explore while maintaining exploitation. But if exploitation requires an integrated organisational structure and exploration a disintegrated one, how are organisations to be structured if they need to combine exploitation and exploration? Often the growth path of new small firms coincides with part of the discovery cycle. In the literature on the growth of the firm a well-known hurdle arises when the innovating entrepreneur has to delegate responsibility, systematise and formalise the organisation after the innovation proves its worth. In the terms of Witt (1998): it will have to shift from ‘cognitive leadership’ to ‘governance’ (and back again). Contrary to what Witt claims, cognitive leadership does not always yield better performance than governance. It performs worse in the systematisation, rationalisation and increase of scale associated with the stage of generalisation. Take-overs, alliances, spin-offs and break-ups of firms help to overcome such problems of transition between the stages of the discovery process. Volberda (1998) identified several ways to solve the paradox of exploitation and exploration. One is separation in place: one part of an organisation engages in exploitation, another in exploration. There is horizontal and vertical separation. In horizontal separation one division or department, typically R&D, preferably in collaboration with marketing, engages in exploration, and another, typically production, engages in exploitation. The problem here is, of course, how to govern the interface. There is the perennial frustration of marketing that production is ‘not willing’ or ‘not competent’ to deliver what market opportunities call for, and the equally perennial frustration of production people that marketing people are too dense to appreciate what is technically feasible and cost-

Managing exploitation and exploration

235

effective. Vertical separation can go two ways. Management yields scope for exploration in the firm by allowing people who interact with the market and with sources of technology, labour and inputs to utilise the opportunities they meet, and management tries to maintain sufficient coherence to prevent waste of duplication and mismatch. Alternatively, management lays claim to choices of direction and content, and coordinates staff in the execution of their vision. This theme has been dealt with by many organisational scholars (e.g. Thompson 1967). In economics it was discussed by Aoki (1986), who made a distinction between a horizontal and a vertical ‘information structure’. In the vertical structure management co-ordinates workshops but is incapable of adequate monitoring of emerging events in markets and technologies. In the horizontal structure production decisions are co-ordinated among semi-autonomous shops, who can better respond on the spot to emerging events. Aoki proposed that the former is typically American (the ‘A firm’), and the latter typically Japanese (the ‘J firm’). In the A firm there are clear job specifications, and standard operating procedures. Problem-solving is relegated to supervisors, repairmen, engineers. In the J firm, duties are not specified in detail, workers rotate across jobs so that they become familiar with a wide range of activities, as a basis for horizontal co-ordination. Decentralisation is also carried across the boundaries of the firm to suppliers, who are given more scope for initiative. The weak spot of this arrangement is that in spite of rotation the insight needed for effective co-ordination may be too limited. And they may have divergent strategic orientations that are at odds with the firm’s focus on core competencies or activities. Separation can also take place between different organisations, and then we might call it the strategy of specialisation. A firm focuses on a specific stage, of efficient exploitation or of exploration, and connects with other firms that offer complementary stages. It continually shifts its portfolio of activities; phasing them in and out as they enter and leave the stage in which it specialises. A prominent example is the pharmaceutical industry, where the large pharmaceutical companies provide efficient production, marketing and distribution, and biotechnology firms explore the novel product forms. Another example is industrial districts. Some firms are specialised in R&D or other forms of experimentation with novel combinations; some are specialised in consolidation and production; some in large-scale and distant marketing, distribution, exports; some in incremental improvements and differentiation. One may also accept that organisations are formed only temporarily, as the need arises. This is what one sees in building consortia, for example. Another separation is separation in time: exploitation occurs at one time, and exploration at another. This yields the ‘oscillating’ (Burns and Stalker 1961) mode, with a to-and-fro between loose and open to tight and homogeneous, and back again. This is very difficult to achieve.

236

Bart Nooteboom

Organisational change requires restructuring of organisational scripts, involving a redistribution of people across tasks and a reconstitution of tasks, goals, motives, perspectives and shared meanings. In industrial districts it takes entry and exit of firms and the building of new network relations. Such developments tend to take a long time, especially if they require a change of the ‘deep structure’ of organisational culture, such as basic categories of perception, interpretation and evaluation (Schein 1985). Restructuring of systems of production, supply and distribution also take a long time. Increasingly, the problem of inertia lies in organisational culture and distribution systems rather than production technology. Furthermore, how does one operate this when different products or technologies are in different stages of development? Nonaka and Takeuchi (1995) recommend a ‘hypertext’ organisation, by analogy to windows processing on computers. This is like a flexible form of matrix organisation: cross-functional and cross-departmental groups are formed ad hoc, according to the opportunity at hand, like opening windows on the computer. A firm may have a basic structure oriented towards either exploitation or exploration, and forms temporary task forces for the other task. An example is the practice at the ‘Central Book House’, in the Netherlands. It buys and supplies books to the book trade. It has three core divisions, which need to be tightly integrated: 1 2 3

The assortment of books, which must offer a high degree of differentiation to customers, in terms of composition and size of packages. Logistics for efficient and fast delivery. Information technology to enable 2 and to collect and provide strategically vital information as to what books are read by whom, where, at what price, for 1 and for customers (both publishers and shop keepers).

Clearly, efficient exploitation is crucial. But so is exploration, to keep up with developments in reading habits, technologies of information processing and dissemination, and changes in publishing. Due to the systemic coherence of the three divisions, innovation must occur in step. To achieve this while maintaining efficient exploitation, the following solution was found. The organisation is basically exploitation oriented. But there is a permanent R&D team of eight people, recruited at different moments (two new people every half year), from the three different departments, and moving out after two years, not going back to their own division of origin but one of the others. Moving out from exploration, back into exploitation, they carry responsibility for implementing innovations they helped to develop. One can see the advantages. One is that exploration is based on experience in exploitation, and is conducted with a view to the expected responsibility for implementation to be taken on later. The dif-

Managing exploitation and exploration

237

ferent divisions each contribute both to exploration and its implementation, to protect systemic coherence. There is turnover in the team to maintain variety and ongoing influx of experience from exploitation. This is a case of combining exploitation and exploration by spatial separation with both horizontal and vertical co-ordination. It looks like the hypertext form recommended by Nonaka and Takeuchi (1995), with the important difference that the cross-functional team for exploration is always in place, albeit with a varying composition, and the production departments know in advance when they will lose and gain people. Are there problems? It seems that the assumption in the system is that the cycle of innovation is two years. If it turns out to be longer, a person moving out of the R&D team has nothing new to implement in the division he joins – at least not an innovation to which he contributed. Is this an incentive or a disincentive to see to it that something new comes out within the two years? What if people are valuable but do not want to move out of exploration and back into exploitation, and if their competencies support that preference? What guarantees that one will always find people who have the skills and interest in both exploitation and exploration? What guarantees are there that rivalries between the divisions will not be carried into the development process? Is this prevented by the prospect of having to move into another division after the two years in exploration? Does this form make the organisation ‘immortal’? Can it survive a shift to electronic publishing and ordering through the Internet? One may also try to escape from the cycle by acting as an orchestrator of activities of other companies. ICT increasingly yields the opportunities for this. Quinn (1992) gives the example of a company in custom-made ASICS (Application Specific Integrated Circuits). They interface with clients directly by means of ICT to determine functional specifications. They then employ their own specialised software to convert this into photo masks, which are sent by ICT to a company in Japan for etching, next to a company in Korea for dicing and mounting, next to Malaysia for assembly, from where the chip is flown directly to the customer. A similar example in sports shoes is Nike. Another example is Benetton: it also performs the task of orchestration, by means of ICT, of a decentralised network of individual producers and retailers.

A new industrial revolution? Let us reconsider the need for standardisation for the sake of efficient production, and the resulting need for integration. Perhaps it is useful to see the present revolution in the organisation of firms in terms of the disappearance of standards and integration. Then the cycle of innovation might collapse. Can we do without standardisation and durable organisational structure? Can exploration be instant and continuous? Can we do without integration, and retain ongoing disintegration? Bennis (1969) has already

238

Bart Nooteboom

predicted the death of bureaucracy, because all environments become turbulent, permitting only ‘organic’, disintegrated structures (quoted in Buchanan and Boddy 1992: 35). From the analysis it follows that this is conceivable when efficient production (exploitation) does not require standardisation, scaling up and division of labour, and when the product requires customised differentiation from the very beginning. This is approximated most closely in professional services, such as accountancy and consultancy. But even there the paradox of exploitation and reciprocation appears, though in a more limited fashion. As already noted in the discussion of the Arthur Andersen company, even there professional standards must be set and kept, and consistent quality must be guaranteed across different locations of a multinational customer. And measures must be taken that different consultants make use of each other’s experience in order not to reinvent wheels all the time. This requires incentive systems for consultants to volunteer their experience to a common pool. That requires that they be judged and promoted at least in part on the basis of such contributions, weighted by their usefulness, measured by the extent that colleagues make successful use of them. But such a common pool would require a certain minimal amount of standardisation of concepts and procedures in a thesaurus. In the previous section I noted the strategy of orchestration to escape the cycle. However, rather than refuting the cycle of innovation, this shows how the orchestrator conducts exploration by flexibly exploiting the productive competencies of different companies, in shifting configurations, and thereby tries to escape the need and the dangers of inertia. The risk of inertia due to standardised, more or less fixed systems for efficient production is hived off to other players. But even here, the focal, orchestrating firm must be careful to both maintain and develop his core competence of orchestration. When standards are not embodied in hardware that represents a large sunk cost (such as cables and switches in telecommunication systems), but in software (as in communication by radio), and it is possible at low cost to translate between different standards by slotting in translation software, then there would seem to be no need for any dominant design. Competing, differentiated standards may remain differentiated from the start, in ongoing differentiation and reciprocation.2 However, that would imply that the competing standards would not be appropriable, and how then do firms obtain the reward for their investment and risk taking? This conundrum is in fact with us: some firms freely distribute their system via the Internet. One explanation is that in doing so they quickly obtain a large user base, which gives information on usage and preferences, down to individual users, which gives a basis for the addition of further added value services geared to individual preferences. It is, so to speak, not the technical system but the customer base that becomes the core competence.

Managing exploitation and exploration

239

An example is the famous case of American Hospital Supplies (AHS). They started with dedicated hardware in the form of terminals which captivated customers. This captivity was broken by new intermediaries who provided an interface for linkage with other suppliers. But AHS had meanwhile added services, partly based on their accumulated knowledge of client wishes and procedures, and thereby created ongoing customer captivity. Does this invalidate the cycle of innovation? Perhaps it does. But the logic underlying the cycle still helps to analyse the conditions for such instant exploration.

Notes 1 Personal communication from Ernst Homburg, who is conducting a historical study of the development of the chemical industry. 2 This insight emerged in a discussion at the Netherlands Institute for Advanced Study (NIAS), with Henrik Glimstedt.

References Abelson, R.P. (1976) ‘Script Processing in Attitude Formation and Decision Making’, in J.S. Carroll and J.W. Payne (eds), Cognition and Social Behavior, Hillsdale, N.J.: Lawrence Erlbaum Associates Publishers, 33–45. Abernathy, W.J. (1978) The Productivity Dilemma: Roadblock to Innovation in the Automobile Industry, Baltimore, Md.: Johns Hopkins University Press. Abernathy, W.J. and Utterback, J.M. (1978) ‘Patterns of Industrial Innovation’, Technology Review, 81 (June/July): 41–47. Abernathy, W.J. and Clark, K.B. (1985) ‘Innovation: Mapping the Winds of Creative Destruction’, Research Policy, 14: 3–22. Aoki, M. (1986) ‘Horizontal vs. Vertical Information Structure of the Firm’, American Economic Review, 76(5): 971–983. Archer, M.S. (1995) Realist Social Theory: The Morphogenetic Approach, Cambridge: Cambridge University Press. Argyris, C. and Schön, D. (1978) Organizational Learning, Reading, Mass.: Addison-Wesley. Bennis, W.G. (1969) Organizational Development: Its Nature, Origins and Prospects, Reading Mass.: Addison-Wesley. Buchanan, D. and Boddy, D. (1992) The Expertise of the Change Agent, New York: Prentice-Hall. Burns, T. and Stalker, G.M. (1961) The Management of Innovation, London: Tavistock. Chandler, A. (1962) Strategy and Structure: Chapters in the History of American Industrial Enterprise, Cambridge, Mass.: The MIT Press. Chesbrough, H.W. and Teece, D.J. (1996) ‘When is Virtual Virtuous? Organizing for Innovation’, Harvard Business Review, Jan.–Feb.: 65–73. Choo, C.W. (1998) The Knowing Organization, Oxford: Oxford University Press. Cohen, W.M. and Levinthal, D.A. (1990) ‘Absorptive Capacity: A New Perspective on Learning and Innovation’, Administrative Science Quarterly, 35: 128–152.

240

Bart Nooteboom

Dosi, G. (1982) ‘Technological Paradigms and Technological Trajectories: A Suggested Interpretation of the Determinants and Directions of Technical Change’, Research Policy, 11: 147–162. —— (1984) Technical Change and Industrial Transformation, London: Macmillan. Dosi, G., Freeman, C., Nelson, R., Silverberg, G. and Soete, L. (1988) Technical Change and Economic Theory, London: Pinter. Eldredge, N. and Gould, S.J. (1972) ‘Punctuated Equilibria: An Alternative to Phyletic Gradualism’, in T.J.M. Schopf (ed.), Models in Paleobiology, San Franciso: Freeman, Cooper & Co., 82–115. Emery, F.E. and Trist, E.L. (1965) ‘Causal Texture of Organizational Environments’, Human Relations, Feb.: 21–32. Fiol, C.M. and Lyles, M.A. (1985) ‘Organizational Learning’, Academy of Management Review, 10(4): 803–813. Freeman, C. and Perez, C. (1989) ‘Structural Crises of Adjustment, Business Cycles and Investment Behaviour’, in G. Dosi (ed.), Technical Change and Economic Theory, London: Frances Pinter. Gersick, C.J.G. (1991) ‘Revolutionary Change Theories: A Multi-level Exploration of the Punctuated Equilibrium Paradigm’, Academy of Management Journal, 16(1): 10–36. Giddens, A. (1984) The Constitution of Society, Cambridge: Polity Press. Gould, S.J. (1989), ‘Punctuated Equilibrium in Fact and Theory’, Journal of Social Biological Structure, 12: 117–136. Hage, J. (ed.) (1998) Organizational Innovation, Dartmouth: Ashgate. Hedberg, B.L.T., Nystrom, P.C. and Starbuck, W.H. (1976) ‘Camping on Seesaws: Prescriptions for a Self-designing Organization’, Administrative Science Quarterly, 21: 41–65. Henderson, R.M. and Clark, K.B. (1990) ‘Architectural Innovation: The Reconstruction of Existing Product Technologies and the Failure of Established Firms’, Administrative Science Quarterly, 35: 9–30. Hendriks-Jansen, H. (1996) Catching Ourselves in the Act: Situated Activity, Interactive Emergence, Evolution and Human Thought, Cambridge, Mass.: The MIT Press. Holland, J.H. (1975) Adaptation in Natural and Artificial Systems, Ann Arbor: University of Michigan. Holland, J.H., Holyoak, K.J., Nisbett, R.E. and Thagard, P.R. (1989) Induction; Processes of Inference, Learning and Discovery, Cambridge, Mass.: The MIT Press. Knight, F. (1921) Risk, Uncertainty and Profit, Boston, Mass.: Houghton Mifflin. Kuhn, T.S. (1970) The Structure of Scientific Revolutions (2nd edn), Chicago: University of Chicago Press. Lakoff, G. and Johnson, M. (1999) Philosophy in the Flesh, New York: Basic Books. Langlois, R.N. and Robertson, P.L. (1995) Firms, Markets and Economic Change, London: Routledge. Lawrence, P. and Lorsch, J. (1967) Organization and Environment, Boston, Mass.: Harvard Business School. Lundvall, B.A. (1985) Product Innovation and User–Producer Interaction, Aalborg: Aalborg University Press. —— (1988) ‘Innovation as an Interactive Process – from User–Producer Inter-

Managing exploitation and exploration

241

action to National Systems of Innovation’, in G. Dosi, C. Freeman, C. Nelson, R. Silverberg and L. Soete (eds), Technology and Economic Theory, London: Pinter. —— (1993) ‘User–Producer Relationships, National Systems of Innovation and Internationalization’, in D. Foray and C. Freeman (eds), Technology and the Wealth of Nations, London: Pinter. Malerba, F. (1993) ‘The National System of Innovation: Italy’, in R.R. Nelson, National Innovation Systems, Oxford: Oxford University Press, 230–260. March, J. (1991) ‘Exploration and Exploitation in Organizational Learning’, Organization Science, 2(1): 101–123. Mintzberg, H. (1983) Structure in Fives: Designing Effective Organizations, Englewood Cliffs, N.J.: Prentice-Hall. Nelson, R.R. and Winter, S. (1982) An Evolutionary Theory of Economic Change, Cambridge: Cambridge University Press. Nonaka, I. and Takeuchi, H. (1995) The Knowledge Creating Company, Oxford: Oxford University Press. Nooteboom, B. (1984) ‘Innovation, Life Cycle and the Share of Independents: Cases from Retailing’, International Small Business Journal, 3(1): 21–33. —— (1992) ‘Towards a Dynamic Theory of Transactions’, Journal of Evolutionary Economics, 2: 281–299. —— (1999a) Inter-firm Alliances: Analysis and Design, London: Routledge. —— (1999b) ‘Learning, Innovation and Industrial Organisation’, Cambridge Journal of Economics, 23: 127–150. —— (2000) Learning and Innovation in Organizations and Economies, Oxford: Oxford University Press. Peirce, C.S. (1957) Essays in the Philosophy of Science, Indianapolis, Ind.: BobbsMerrill. Piore, M. and Sabel, C. (1983) ‘Italian Small Business Development: Lessons for US Industrial Policy’, in J. Zysman and L. Tyson (eds), American Industry in International Competition: Government Policies and Corporate Strategies, Ithaca, N.Y.: Cornell University Press. Popper, K.R. ([1970] 1976) ‘Normal Science and Its Dangers’, in I. Lakatos and A. Musgrave (eds), Criticism and the Growth of Knowledge, Cambridge: Cambridge University Press, 51–58. Porter, M.A (1980) Competitive Strategy, New York: Free Press. —— (1985) Competitive Advantage, New York: Free Press. Quinn, J.B. (1992) Intelligent Enterprise, New York: Free Press. Rothwell, R. and Zegveld, W. (1985) Innovation and the Small and Medium Sized Firm, London: Frances Pinter. Schein, E.H. (1985) Organizational Culture and Leadership, San Francisco, Calif.: Jossey-Bass. Shank, R. and Abelson, R. (1977) Scripts, Plans, Goals and Understanding, Hillsdale, N.J.: Lawrence Erlbaum Associates Publishers. Shanon, B. (1988) ‘Semantic Representation of Meaning: A Critique’, Psychological Bulletin, 104(1): 70–83. —— (1990) ‘What is Context?’, Journal for the Theory of Social Behaviour, 20(2): 157–166. —— (1993) The Representational and the Presentational, New York: Harvester/ Wheatsheaf.

242

Bart Nooteboom

Smircich, L. (1983) ‘Organization as Shared Meaning’, in L.R. Pondy, P.J. Frost, G. Morgan and T.C. Dandridge (eds), Organizational Symbolism, Greenwich, Conn.: JAI Press. Stinchcombe, A.L. (1990) Information and Organizations, Berkeley, Calif.: University of California Press. Teece, D.J. (1986) ‘Profiting from Technological Innovation: Implications for Integration, Collaboration, Licensing and Public Policy’, Research Policy, 15: 285–305. —— (1988) ‘Technological Change and the Nature of the Firm’, in G. Dosi, C. Freeman, R. Nelson, G. Silverberg and L. Soete (eds): Technical Change and Economic Theory, London: Frances Pinter. Thompson, J.D. (1967) Organizations in Action, New York: McGraw-Hill. Tushman, M.L. and Anderson, P. (1986) ‘Technological Discontinuities and Organizational Environments’, Administrative Science Quarterly, 31: 439–465. Vernon, R. (1966) ‘International Investment and International Trade in the Product Cycle’, Quarterly Journal of Economics, 80: 190–207. Volberda, H.W. (1998) Building the Flexible Firm, Oxford: Oxford University Press. Vygotsky, L. (1962) Thought and Language, edited and translated by E. Hanfmann and G. Varkar, Cambridge, Mass.: The MIT Press. Weick, K.F. (1979) The Social Psychology of Organizing, Reading, Mass.: Addison-Wesley. —— 1995 Sensemaking in Organisations, Thousand Oaks, Calif.: Sage. Witt, U. (1998) ‘Do Entrepreneurs Need Firms?’, Paper presented at the Workshop ‘Recent Developments in Austrian Economics’, Max-Planck Institut zur Erforschung von Wirtschaftssystemen, Jena, 7–8 August.

10 The epistemic foundations of social organizations A game theoretical approach Christian Schmidt

Introduction It could seem strange at first sight to use game theory to understand structural changes and institutional organizations of economies. Beyond the well-known objectives of a retrospective bias in reading past economic phenomena through a recent intellectual construction, two major criticisms immediately come to mind. First, game theory is supposed to be derived mainly from a strictly individualistic methodology based on a narrow definition of rationality. Second, the line of reasoning is largely dominated by the backward induction, a mental procedure where individual players are assumed to choose their strategy from the beginning to the end of the game. If game theory was actually built in the spirit of pure individualistic rationality and developed in a stable static perspective, we could legitimately question its relevance for studying structural and institutional changes in economy. The aim of this chapter is to show that these assumptions, which are more generally attached to game theory and commonly accepted by scientists, are not correct on the grounds of historical, as well as analytical arguments. Consequently, game theory opens up two interesting ways for understanding institutional changes which will be briefly explored. On one side, a fresh look at such basic notions as ‘standards of behaviour’ and ‘social orders’, which can already be found in von Neumann and Morgenstern’s Theory of Games and Economic Behavior (henceforth TGEB), gives analytical foundations for identifying various institutional systems through different solution concepts. On the other side, a dynamic approach to interactive systems exists, at least from Volterra’s works on modelling animal population struggles. Such a dynamic version of games was reinforced even in 1950 by Nash’s suggestions about an alternative dynamic interpretation of his well-known equilibrium-solution concept. Following these directions can definitely improve our knowledge of the processes of economic and social changes. The analytical part of the chapter is devoted to the first perspective.

244

Christian Schmidt

Some historical fallacies on game theory Nobody can object that the players of a game in its technical sense are individual decision-makers1 who choose freely their strategy (or more generally their movements). In addition, they are supposed to be selfish maximizers for the sake of simplicity.2 Are such very broad assumptions sufficient to conclude that the original framework of game theory is no more than an extended and sophisticated, or at variance oversimplified (Spohn 1982), version of the standard individualistic paradigms of neoclassical economics? We do not think so. From the very beginning, and in spite of some misinterpreted statements, the scope of game theory is not to be found in the entities themselves (the decision-makers) with their individual characteristics (preferences, ability to maximize selfish utilities, etc.), but rather in the content of their relations during the interactive process, precisely labelled as the ‘game’. Several pieces of historical evidence show that the founders of game theory were aware of this novelty.

Interactive decision processes and individual decision-making Mixed strategies occupy a central place in games of strategy. In his two brief notes concerning what he called ‘games that involve chance and the skill of the players’, Borel already introduced the idea of mixed strategy for solving the three examples of strategic games which he raised (Borel 1921, 1924). Von Neumann, in his 1928 seminal paper, suggested a disputable, but very interesting interpretation of the use of this notion for such a purpose. Let us quote von Neumann who is reasoning, like Borel, in the restricted format of a two-person, zero-sum game: At the beginning of the game, S1 (the player) is not asked to choose one of the numbers 1, 2, . . . 1, he has only to specify 1 probabilities. (von Neumann [1928] 1959: 31) Thus, he specifies the intuition behind the raison d’être of such an artifice in the following terms: If S really wants to get a particular strategy X, he can specify P(x1)  1, P(x2)  0 (for x1  x2). On the other hand, he is protected against his adversary ‘finding him out’; for, if e.g. P(x1)  P(x2)  
nobody (not even himself) can predict whether he is going to choose 1 or 2! (ibid.: 23) And he concludes on this point:

Epistemic foundations of organizations 245 In specifying the rules of behaviour for the players it becomes imperative to reconsider the element of ‘hazard’. The dependence on chance (the ‘statistical element’) in such an intrinsic part of the game itself (if not of the world) that there is no need to introduce it artificially by way of the rules of the game itself (if not of the world): Even if the formal rules contain no trace of it, it still will assert itself. (ibid.: 26) What does von Neumann’s interpretation of choosing mixed strategy really mean? Beyond the explicit rules of the game, game theory proposes rules of behaviour for the players. Such rules of behaviour are much more complex than simple individual maximization. Von Neumann suggests that the purpose of each player could be to avoid being ‘found out’ by the other. According to this purpose, ‘hazard’ is transformed into behaviour for the two players. If both accept these rules, the problem raised by their interaction in several game situations has a solution. This is the final justification for their common acceptance of the prescriptions. Therefore, hazard must be understood more as an implicit institution of what von Neumann called at that time ‘games of strategy’. Let us note that von Neumann’s strategic interpretation of hazard was not shared by Borel, for whom nobody can mimic hazard (Borel 1939). However, this does not refute our view on the real foundations of game theory. Borel does not dispute the necessity of rules of behaviour for the players of a strategic game. He only contests that such rules can be derived from hazard. One can even say, in retrospect, that he was partially right on this point. Indeed, when we extend the domain of the game from zero-sum to non zero-sum, hazard no longer justifies the mixed strategies corresponding to its solution. But it does not follow, however, that this set of mixed strategies cannot be understood as rules of behaviour derived from a social solution. Fifteen years later, the social characterization of interactive decisions are specified, and their consequences, both on the content and on the foundations of game theory, are developed in TGEB. Such an insistence on underlining the economic specificity of the interactive processes is probably due to Morgenstern and must be connected to his original position in the Austrian intellectual configuration (Schmidt 2001b). Anyway, from the very beginning of the book, a comparison is developed between a ‘Robinson Crusoe economy’ and a ‘social exchange economy’ in order to delimit the difference between the individual and the inter-individual approaches to economic phenomena (TGEB: 9–12). Three main features emerged from this comparison, and they are summarized in Table 10.1. Several interesting consequences are to be derived from Table 10.1. The difference in the two systems leads to a quite separate treatment of

246

Christian Schmidt

Table 10.1 Comparison between the Robinson Crusoe and social exchange economies Robinson Crusoe economy

Social exchange economy

The problem

Robinson Crusoe economy faces an ordinary maximum problem

No player faces a maximum problem but a peculiar and disconcerting mixture of several conflicting problems due to other players (TGEB)

The system

Robinson Crusoe operates in a world whose variables are either under his control (tastes, preferences, etc.), or completely out of his control (the occurrence of independent states, etc.)

Each player co-operates in a world where almost all the variables are only partially controlled by himself (states of the game, and their associated outcomes)

The expectations Robinson Crusoe makes expectations on the difference ‘states’ of the world

Each player makes expectations on the other players’ actions (and, through the reflexivity property, other players do the same on his action)

expectations in the two cases. As the players’ world is only derived from the interactions of all the participants, there is no hypothesis corresponding to the rational assumption of an equi-probability of the states of the world in game situations. Therefore, a statistical modus operandi for the decision-maker does not exist in a social exchange economy and must be replaced by a social modus operandi of a totally different type. So, as the players of a game make expectations on others, they must make expectations on other expectations . . . and so on. The core of a game situation is to be viewed as an intricate system of embedded mutual expectations, and, thus, of interdependent beliefs. Such a social dimension cannot be reduced to individual decision-makers, even if the states of the game are the final result of the free will of individuals.3 A careful reading of TGEB reveals that von Neumann and Morgenstern were perfectly aware that the analysis of such interactive economic and social processes is not a simple extension of individual decisionmaking plus additional technical difficulties. Their conclusion does not leave any doubt on the matter. The grounds for difficulty lie not on the field of those social relationships we have mentioned before, although we do not question their significance. But rather they arise from the conceptual differences between the original Crusoe’s maximum problem and the more complex problem sketched before.

Epistemic foundations of organizations 247 We hope that the reader will be convinced by the above that we face here and now a conceptual – and not merely technical – difficulty. And it is this problem which the theory of ‘games of strategy’ is mainly devised to meet (TGEB: 12) In spite of its clarity, this warning message has been fully understood only recently. Due to the well-known TGEB appendix devoted to the first axiomatic treatment of utility, the majority of game theorists up to the 1980s focused on the connection between game theory and the models of the expected utility through the Bayesian rules of revision. Therefore, game theory was mainly pictured as an attempt to transform problems of interacting players into a special case of individual decision-making under risk (Luce and Raiffa 1967). Harsanyi was even more precise on that point. He considered rational behaviour in game theory as a direct generalization of the Bayesian rationality under uncertainty. Indeed, in a game situation each player can evaluate his expectations about the decision of another player in terms of subjective probabilities and revise his belief in a way consistent with the Bayesian rules. Thus, game theory is to be interpreted as an extension of the Bayesian rational approach of individual decision-making (Harsanyi 1977). This standard view was largely shared by the economists and the theoreticians of decision-making during more than thirty years, with a few exceptions, including Binmore (1990, 1993). More reflection shows that the relation between the specificity of interactive situations covered by game theory and the statistical Bayesian rules of individual beliefs revision is hardly obvious. As was already noted by the authors of TGEB, players’ knowledge was not mainly concerned with substantive matters but rather with others’ knowledge. Aumann, although he himself utilized the Bayesian approach as a technical tool for a tractable subjective treatment of probability in game theory, became convinced that a system where knowledge and belief refer to more than one individual person requires its own logical foundations which cannot simply be derived from those of individual decisions (Aumann 1999a). This recognition of such a social foundation for interactive situations studied by game theory has only recently re-opened the way to new investigations on the epistemic background of this theory. The dynamic dimension of interfacing The static nature of game theory reasoning is derived from two features most generally attached to the concept of strategy. Being defined as a set of all the players’ movements chosen from the very beginning to the end of the game, the strategy appears as timeless. Furthermore, the wellknown procedure of ‘backward induction’, which is purely speculative and

248

Christian Schmidt

out of time, reinforces this picture, due to its close relation with players’ rationality in game theory. If choosing rationally implies proceeding ‘backwards’, there is apparently no room for time in strategic games. But, once again, some historical evidence questions the relevance of this common view. During the academic years 1928–1929, Borel invited his friend, the great mathematician Vito Volterra, to give several lectures on mathematical models of dynamic biology at the Institute Henri Poincaré, of which he was the director. An enlarged and revised version of those lectures were published later in French under the title Leçons sur la théorie mathématique de la lutte pour la vie (1931). Such a book is generally considered as a major precursor of evolutionary games (Maynard Smith 1982). The comparison between Borel’s mathematical treatment of what he called ‘la théorie du jeu’ between two players (Borel 1921, 1924) and Volterra’s models of animals fighting (1931) is really impressive. Starting with integral linear equation systems with a skew symmetric kernel where aij  aji, both authors utilized the possible permutations offered by such systems from which they derived their main results. But whereas Borel’s topic is the confrontation between two sets of ‘manners of playing’, Volterra’s study concerned the interrelations between animal species in a crude opposition predator/prey.4 In addition, the problem to be raised is not the same in the two cases. Borel’s research is to find a precise rule for choosing a manner of playing which competes against all the opponent’s manners of playing. Volterra’s matter is to determine some dynamic laws of animal species’ fluctuations. But as Borel rightly observed in his first note: Numerous problems can thus be reduced to the study of integral equations with a skew symmetric kernel. This kernel depends on the conventions of the game, whilst the diverse forms of the integral equations depended on the problems posed. (Borel [1921] 1953: 99) Indeed, the properties of the matrices’ coefficients can be studied either to determine the numerical solution, as for Borel’s problem, or for analysing the nature of the dynamic stability, as for Volterra’s question. This common mathematical support reveals the dynamic facet which necessarily complements the classical domain of game theory. Its best illustration is provided by a very simple example which can be found in Borel’s as well as in Volterra’s works translated in their respective fields of research (see Figure 10.1). For Borel, it pictures the parlour game ‘paper’, ‘rock’, ‘scissors’ (Borel 1924: 215). But it also portrays the situation studied by Volterra between three species, A, B, C, where A beats B, B beats C, but C beats A (Volterra 1931: 63). This mathematical dress leads Borel to the con-

Epistemic foundations of organizations 249 A

B

C

A

0


1

1

B

1

0


1

C


1

1

0

Figure 10.1 Borel’s and Volterra’s problem

clusion that it is the best rule for playing choosing by chance on the grounds that p1  p2  p3  1/3. It also allows Volterra to demonstrate that his stylized system of three animal species tends to a unique dynamic equilibrium which is asymptotically stable under some additional restrictions. The fact that the same formal properties are used for studying these two situations and solving the problem they raised does not mean that ‘paper’, ‘rock’, ‘scissors’ and the three species are two identical games. On the contrary, it clearly shows why a dynamic perspective of interactive phenomena also belongs to the domain of the mathematical theory of games from its very beginning. Surprisingly, we have no indication that Borel and Volterra were aware at that time of such a similarity (Schmidt 2001b). The connection between these two complementary dimensions of a game was clarified twenty years later, thanks to Nash and his renewal approach to game theory through non-cooperative games. Nash’s basic idea was a new mathematical concept of solutions, which became the well-known Nash equilibrium. The definition of non-cooperative games was only derived from it to provide an interpretative support to this concept, initially defined in its pure logical acceptance (Nash 1950b). Thus, the concept of equilibrium understood in a non-cooperative game gives rise to two different interpretations. A first interpretation, called by Nash ‘mass action’, refers to populations of individuals who accumulated empirical evidence about consequences of the pure strategies at their disposal during the development of the game. According to it, the mixed strategies represent the average individual’s behaviour in each population corresponding to an equilibrium point. In a second interpretation, the equilibrium point is the result of a rational prediction of the behaviour to be expected of rational players in the game (Nash 1950a). Thus the first

250

Christian Schmidt

one is connected to the notion of a stable distribution of strategies, whilst the second included a shared knowledge of mutual rational expectations. Let us observe, by the way, that a large number of players (population) is not a precondition for the dynamic interpretation of the equilibrium as a solution of a game: The population (according to the mass-action interpretation) needs not to be large if the assumptions still hold. There are situations in economics or international politics in which, effectively, a group of interests are involved in non-cooperative games without being aware of it. (Nash 1950a) At first glance, stability’s conditions in the dynamic interpretation seem to be a substitute for rationality in the static one. But things are much more complex. The information and the kind of knowledge are not the same in both cases. As for the stability, we know now that the conditions of stability for a Nash equilibrium are not sufficient to guarantee the asymptotic stability towards one equilibrating trajectory of a dynamic game. Once again, but for a different purpose, the previous example of the game ‘paper’, ‘rock’, ‘scissors’ provided an illustrative example where the Nash equilibrium does not necessarily correspond to an ESS (equilibrium stable strategy) which only guarantees the system against unpredictable cycles when A → B → C → A . . . (Maynard Smith 1982: 20). Every ESS is a Nash equilibrium, but the converse is false. The main difference between ‘dynamics’ for evolutionary games and ‘static (or strategic) games’ is mainly a question of interpretation. While strategies are related to populations in the evolutionary games, they are associated with individuals in strategic games. Therefore, evolutionary games require an additional assumption for relaying individuals to populations. The biological concepts of ‘phenotypes’ and Darwin ‘fitness’ easily solve this question for animal populations. They cannot be directly transposed to human societies. Therefore, the formal black box of ‘dynamic replicator’ necessitates considerable work to give rise to relevant interpretation of social phenomena. Anyway, game theory is no more intrinsically static in its content than individualistic in its topic.

From solution concepts of a game to social organizations The historical background of game theory offers serious arguments for restoring the institutional dimension of its topic. The landscape today, however, is not so clear due to several evolutions more or less generated by Nash’s very innovative ideas. First of all, the initial domain of game theory has been split into two separate fields, namely, the cooperative and the non-cooperative games.

Epistemic foundations of organizations 251 At first glance, the institutional dimension seems more obvious in cooperative games, where the solutions are defined in terms of coalitional organizations which give rise to institutional interpretations. Such evidence disappears in non-cooperative games where the players are assumed to operate in a totally decentralized world, free of any kind of institutions.5 On another hand, the evolutionary games have been developed from a non-cooperative approach as an extension of the dynamic interpretation of the Nash equilibrium. By contrast, the analytical framework of cooperative games remained static up to now. Thus, the link between the dynamics of interactive social processes and one or the other solution concept becomes less visible. The implication of another evolution, namely Aumann’s investigation on the epistemic foundation of interactive situations, is much more complex for our purpose. Of course, Aumann stresses the logical distinction between the knowledge of an individual and the ‘common knowledge’ of a set of individuals. Between these two extremes, Aumann points out the specificity of intermediate situations, such as, for example, the mutual knowledge between players. He also outlines the specificity of the knowledge that an individual can have on the others. Aumann’s canonical model, although straightforward, remains slightly disappointing. Such a knowledge, as would be a dictionary, does not provide any factual information to the players, but it is supposed to be commonly known by the process of its construction (Aumann 1999b). If this syntax approach to the means of knowledge operators is formally convincing, its interpretation is not so clear. What is the actual status of players’ knowledge of a solution concept? Traditionally, it is assumed to be common knowledge between the players, on the disputable grounds that the solution is a component of the rules of the game. Such an assertion postpones the question which concerns the knowledge of the rules of the game. In an informal discussion of his canonical model, Aumann takes the example of a chess game. According to this model, a proposition like ‘player 1 and player 2 played a game of chess’ implies that player 1 knows that player 2 won or knows that player 2 did not win and that this implication is commonly known by the players (Aumann 1999a: 293). But this contradicts the metaphor of the letters of an alphabet (or the words in a dictionary). Therefore, the knowledge of the rules of the chess game can hardly be reduced to some kind of dictionary. Can the same result be applied to the solution concepts in game theory? Part of the answer can be found in Aumann and Brandenburger’s paper where they investigate the conditions on players’ knowledge and beliefs for a Nash equilibrium (Aumann and Brandenburger 1995). The two authors demonstrate that: (1) the knowledge of an equilibrium must only be a mutual knowledge, at least in the case of a two-person game;6 and (2) the model of belief which is attached to the game is commonly known by the players. There is some difficulty in combining these two statements

252

Christian Schmidt

into a clear unified construction. Anyhow, Aumann’s work remains, until now, limited to the Nash equilibrium solution. For other solution concepts, nothing has really been done in that direction. In spite of these ambiguous signals, we persist in thinking that game theory is developing a relevant analytical framework for understanding the institutional foundations of social organizations. In order to establish the point, we will revisit the notions of ‘accepted standards of behaviour’ and ‘established social order’ picked out from TGEB in the light of new insights. These notions will be used as guidelines for discussing various approaches to the question proposed by different researchers, including myself. ‘Accepted standard of behaviour’ and ‘established social order’ von Neumann and Morgenstern provided in TGEB some interesting suggestions, which have been largely neglected by their successors, through the notions of ‘standard of behaviour’ and ‘established social order’. At first sight, such notions seem to be elaborated only to give an intuitive interpretation for the mathematical solution of a game. A careful reading of the book reveals their much more important role in von Neumann and Morgenstern’s intellectual construction. Indeed, they first appear in chapter 1 devoted to a re-interpretation of ‘the economic problem’ in the new language of game theory (TGEB: 30–44). But they became imperative for the understanding of what they call the exact form of a solution (TGEB: 263) and support each attempt to extend the initial framework of the theory (TGEB: 471, 499, 501). A footnote in chapter 1 gives the reason why the ‘standard of behaviour’ can be used as a key for understanding the necessary social acceptance of the solution by the players. After recalling that the rules of the game are considered as given, it adds in a rather elliptic way: ‘We suggest . . . that the reader forgets temporarily the analogy with games and thinks entirely in terms of social organisation’ (TGEB: 41, n. 1). What does it mean? In parlour games, the solution concept is implicitly included in the rules of the game which are completely known by the players. Therefore, the solution concept of the game is given with its rules. The situation is not the same in the social world, where the rules of the game (i.e. the order of movements, the numbers of the sequences, etc.) are free from a definite concept of solution. Thus, in parlour games the solution does not require an interpretation because the players adhere to the solution when they accept the rules of the game. In social games, on the contrary, the solution is not directly derived from the rules of the game and, therefore, must be found by the theoretician. This is not sufficient, however, to implement it. In addition, the players must accept the solution proposed by the theoretician. In that perspective, the solution is to be attractive for the players in order to become a social organization.

Epistemic foundations of organizations 253 Let us abandon the metaphor of parlour games to consider a more abstract game with a unique solution. Such a solution has to be interpreted for the reasons previously given. But its interpretation does not raise a specific problem, due to its unique associated standard of behaviour. A two-person zero-sum game is a good illustration. Its solution, derived from the famous theorem Maxmin  Minmax, is intimately linked to a wellknown standard of behaviour, namely, the Maximin criterion. In other words, the interpretation of the solution is self-evident in this case. This explains, in retrospect, why von Neumann ([1928] 1959) and Borel did not care for it. A new problem actually emerges when the solution concept is not given by the rules of the game, and when either the solution concept gives rise to different solutions, or when various solution concepts are consistent with the same data of the game. Following the two authors of TGEB, we will start with the first case. The relation between the standard of behaviour assigned to the players by the solution of a social game and the social order established by the implementation of the solution looks simple. To perform the social organization, the players have to adopt a specific standard of behaviour. On another side, the players will adopt this standard if they accept the social order which will follow the recommended behaviour. At this point, von Neumann and Morgenstern make a disputable distinction between two categories of conditions on the social order(s) established by the solution. First of all, the solution of a game cannot lead to a self-defeating system. Second, it must satisfy some additional desiderata concerning, for example, the allocation of the final outcome between the players. Whereas the first condition is supposed to be natural and reflect the ‘order of the things’, the second conditions are derived from subjective considerations. This way of setting the question is closely dependent on the solution concept chosen in TGEB by von Neumann and Morgenstern, ‘the stable sets’ where the absence of dominated imputations characterized lato sensu the ‘inner stability’ of the corresponding organization. Such a stability can be viewed as an expression of the non-self-defeating condition. Unfortunately, von Neumann and Morgenstern’s ‘stable sets’ can lead to different solutions which implies, in TGEB’s terminology, various social orders and, consequently, several standards of behaviour. Such a multiplicity is just the consequence of the second conditions; that is, the diversity of social desiderata. But this leads to a result which is not quite convincing, as suggested by the authors themselves: Several stable standards of behaviour, of course, exist for the same factual situation: Each of these would, of course, be stable and consistent in itself, but in conflict with all the others. (TGEB: 266)

254

Christian Schmidt

No doubt the canvas sketched out by von Neumann and Morgenstern is a fruitful starting point. But due to the time, several of its components reveal some weaknesses. The distinction between the solution of a game and the solution concept from which this solution is derived was still not drawn. Von Neumann and Morgenstern’s initial quest is to discover a general solution for every game, but the solution they found was a set of different ‘imputations’. In order to give a sense to these imputations they associated a specific standard of behaviour to each of them. Let us try to extend von Neumann and Morgenstern’s ideas to other concepts, such as the Nash equilibrium. In most cases, a game possesses several Nash equilibria. Each of them, however, refers to the same standard of behaviour. If we consider that each equilibrium is a solution which belongs to the same concept, it seems correct to associate a definite standard of behaviour to the solution concept and not to its different solutions in a given game. That which is true for the Nash concept is also true for other solution concepts, including von Neumann and Morgenstern’s stable sets solution. Von Neumann and Morgenstern’s interpretation of the stability condition is also disputable. Obviously, any solution must have a minimum of stability to give rise to a social order. But one can wonder on what principle this stability is based. In a social situation, the stability most often results from the confidence of the players’ expectations on behaviour of the others. Incidentally, the two authors of TGEB surmise the existence of such a phenomenon. They notice that the ‘accepted standard of behaviour’ is a necessary condition to maintain players’ faith in the definite solution (TGEB: 266). They cannot analyse this more rigorously because they lack a model of the players’ knowledge. Once again, von Neumann and Morgenstern’s intuition is derived from the solution they have elaborated in TGEB where the stability is a direct implication of non-dominated imputations. A glance at other solution concepts proves that their stability has nothing to do with a natural ‘order of things’. The strategic stability of a Nash equilibrium only depends on the absence of an incentive to deviate from the strategy specified by the concept (Kohlberg and Mertens 1986). Even the Shapley value, which escapes a priori any stability conditions, can generate a social order ‘stable’ at minima through a self-fulfilling mechanism. These two examples show that, contrary to von Neumann and Morgenstern’s interpretation, the stability of a solution concept must be understood as a desiderata for the corresponding social order. Finally, von Neumann and Morgenstern are aware of a kind of vicious circle in the reasoning for relying on a standard of behaviour accepted by the players to the established social order. On the one hand, the existence of a social order depends on the implementation of a specific standard of behaviour by the players. On the other, a specific standard of behaviour

Epistemic foundations of organizations 255 must satisfy the social desiderata included in the expected social order to be accepted by the players. In order to avoid such a difficulty, von Neumann and Morgenstern utilize the distinction previously criticized between the stability conditions and the other social desiderata. Beyond the necessary stability of whatever social order, their investigation does not concern the social norms imbedded in the social order corresponding to a solution of the game. Thus they quickly close the debate in the following terms: Our problem is not to determine what ought to be happening in pursuance of any set of necessarily arbitrary a priori principles, but to investigate where the equilibrium of forces lies. (TGEB: 43). This position becomes hardly acceptable if the stability conditions are no more disconnected from other social norms. There is nothing to object to the very general statement that the solution of a game is more or less directly derived from the investigation of the balance of forces. But, according to von Neumann and Morgenstern, such a balance results from the implementation of standards of behaviour which are accepted by the players. On what grounds can the players accept this or that standard of behaviour, except the social norms directly or indirectly incorporated in the corresponding social order? Therefore contemporary game theorists must bravely face the problem raised by TGEB about the connection between the acceptance of the standard of behaviour by the players and the norms included in its corresponding social order. However, its investigation necessitates, on the contrary, the analyses of what ought to happen in pursuance of different sets of a priori principles. This legacy from von Neumann and Morgenstern can be developed in different directions. Three of them will be successively discussed. Greenberg’s social situations A first extension of von Neumann and Morgenstern’s seminal ideas is due to Greenberg. In an ambitious research programme, Greenberg has sketched out what he has called a ‘theory of social situations’ for challenging game theory (Greenberg 1990). Roughly speaking, a social situation pictures all the environmental features necessary for the players in order to reach a definite solution in the game theory terminology. Let us summarize how Greenberg revisits the old notions of ‘accepted standard of behaviour’ and ‘established social orders’ for his purpose: 1

The solution of a game is tantamount to a set of rules which lead to a set of final outcomes when they are applied by the players. According to this, the standard of behaviour is an operator who transforms

256

2 3

4

Christian Schmidt players’ behaviour to a game solution in a broad sense. Therefore it can be mathematically written as a mapping (G), where G is the domain of the game. Only solutions which are not self-defeated can refer to social orders. Consequently (G) must be stable in this very weak meaning. (G) in itself is not dependent on a specific solution, but the interpretation of a solution concept necessitates the definition of an associated social situation (S.S.), which takes into account the institutional environment and the players’ beliefs. Thus, a correspondence can be shown between a definite standard of behaviour and a specific solution concept thanks to the social situation. In formal terms, (S) ⇔ (S.S.), where S is the solution concept of the game G and S.S. its corresponding social situation. The players of a game are always free to accept or to reject a definite standard of behaviour. Accepting (S) implies accepting the social situation to be associated to the solution concept of the game. So, either this social situation is a correct idealization of the players’ actual situation and the players normally accept the corresponding standard of behaviour, or it is not the case and they reject this standard of behaviour. If their actual situation cannot be translated into any kind of social situation derived from a game theory solution concept, a new solution is to be found.

Greenberg moves from von Neumann and Morgenstern’s construction. With Greenberg, the standard of behaviour becomes the prescriptive version of the mathematical concept of a solution. In that spirit, the standard of behaviour offers an opportunity to unify almost all the different solution concepts of a game.7 On the other side, each solution concept gives rise to a social situation. This social situation induces constraints on the standard of behaviour and reveals, at the same time, a social order which can be established by the players. Greenberg’s notion of ‘social situation’ brings the link between the standard of behaviour to be accepted by all the players and the social order established by their acceptance. Unfortunately, a social situation à la Greenberg is a baroque notion which puts together two very different components: the institutional environment, on the one hand, which is given from outside, and the players’ beliefs, on the other, which are derived from assumptions about their cognitive ability. In Greenberg’s theory, the social situations provide a unique framework for the two purposes – namely, to describe the process for implementing a definite game solution concept and to picture the real interactive situation where the players operate. The first target is hardly reached. As for the second, the point remains questionable, due to the mixture of its components. Let us start with the weakness of the social situation as the procedure to be associated with a social concept. Nash equilibrium, for example, is

Epistemic foundations of organizations 257 transformed in what Greenberg calls a Nash situation by means of the three following conditions (Greenberg 1990: 89): C1  only simple players are allowed to deviate. C2  a deviating player is free to choose any strategy from his strategy set Zi. C3  the deviating player believes that all the players will stay put and pursue the same action they intend regardless of the action he chooses. Indeed, a Nash equilibrium is a state where no individual player has an incentive to deviate from his corresponding strategy. One can easily understand why this set of conditions imposed on the deviating players is the cornerstone of a Nash situation. C1 and C2 belong to a supposed Nashian environment. But the real meaning of C3 about the players’ beliefs is much more questionable, as shown by the game illustrated in Figure 10.2. What precisely does it mean for a player to deviate? According to the classical definition of a strategy, to deviate is to be understood as a part of a mental process in the deliberation for choosing a strategy. If, for instance, (A,a) is the starting point of player 1’s reasoning, he will mentally deviate from A to B because, thanks to C3, he believes that player 2 will stay with a. Thus he moves mentally in the game from the state (A,a) to the state (B,a). Let us imagine now that player 2, starting from (B,a) will also deviate from a to b for the reason previously given for player 1. Can we still assert that a deviating player will stay put on the other player’s same strategy? C3 entertains a confusion between two different kinds of player’s belief. Certainly, a deviating possibility in the Nash context of best reply to the 2

1

a

b

c

A

1,3

4,2

2,1

B

2,2

3,3

2,1

C

1,2

1,2

2,2

Figure 10.2 Nash situation example

258

Christian Schmidt

other’s strategy implies that the other’s strategy is supposedly given for each player during his mental deliberation. But this does not preclude the players’ expectations on the strategy chosen by the others (and expectations on the other’s expectations). More precisely, as long as players do not have any belief about the strategy which is chosen by the other players, they consider the strategies of other players as potentially given according to C3. Players, however, must have some beliefs on others’ behaviour when they play in a Nash situation. Let us come back to the example. According to C1, C2 and C3, player 1 knows that he must choose B against a, A against b and A, B or C against c. Suppose, moreover, a perfect and complete information. Player 1 also knows that player 2 must choose a against A, b against B and a, b, or c against C. Does player 1 necessarily induce from this knowledge the belief that player 2 will choose c, and, consequently, that he must himself actually chose C? A positive answer requires additional knowledge. If Nash equilibrium, as the solution of the game, is a mutual knowledge between players 1 and 2, then player 1 will choose C, because he will take for sure that player 2 will choose c (and symmetrically for player 2). What would happen if this mutual knowledge was dropped out? Player 1 will not choose C if he does not take for sure that player 2 will choose c because c is (weakly) dominated by A and B (and symmetrically for player 2). Therefore (C,c) which is the Nash equilibrium, will never be obtained.8 This belief is for this reason absolutely necessary and its roots can only be found in the knowledge of the mutual acceptance of this solution by the two players. Greenberg would argue that the Nash equilibrium, as the solution of this game, is already incorporated in the standard of behaviour consistent with the corresponding social situation. As, by hypothesis, there is also a relevant standard of behaviour for this social situation, the players are supposed to infer the Nash solution from their knowledge of the situation. The raison d’être of C3 is to assure the players that no state of the game, except the Nash equilibrium ((C,c) in the example) is stable in Greenberg terms, but nothing more. Does it imply that the standard of behaviour associated with a Nash situation must necessarily be ‘accepted’? The answer is obviously positive for a theoretician in an almost tautological way. It cannot be extended, however, to the players of the game because there is no reason that C3 generates the confidence for each player in a mutual acceptance of this solution.9 A major consequence of this deficiency is the impossibility of reducing a social order to a positive comparison between players’ real situations in an idealized ad hoc social situation.

Epistemic foundations of organizations 259 Moulin’s modes of cooperation Another tentative approach to attach the different solution concepts to institutional organizations is due to Moulin. Contrary to Greenberg, Moulin does not refer to von Neumann and Morgenstern’s notions. His starting point is a very broad definition of cooperation combined with a strict delimitation of its domain. According to the dominant economic tradition, Moulin’s cooperation is tantamount to a mutual assistance between selfish individuals (Moulin 1995: 4). He then differentiates three fundamental ‘modes of cooperation’ that he respectively calls ‘direct agreement’, ‘decentralized behaviour’ and ‘justice’. Each of these modes pick out a specific feature of the cooperation, as well as a type of institutional organization. Each mode of cooperation is modelled by a group of games. Their backgrounds are to be found in specific solution concepts. I propose to sketch out Moulin’s approach by associating the ‘direct agreement’ with the core solution, the ‘decentralized behaviour’ with the Nash equilibrium, and the ‘justice mode’ with Shapley’s value.10 From an abstract point of view, the simplest mode of cooperation is the result of a direct agreement between the players. More precisely, there is a set of individual players. Each player is able to negotiate freely with all the other players of the game in a face-to-face ‘bargaining process’. This process takes the form of a hierarchical system of coalitions and sub-coalitions, where a coalition is a set (or a sub-set) of agreeing players. As, for example, suppose three players, namely players 1, 2, and 3. Player 1 considers successively possible agreements with 2, leaving out 3, or with 3, leaving out 2, or with 2 and 3 together. The solution is final agreement which is accepted by all the players of a game. In such a context the ‘core’ is a serious candidate for a solution, because it singles out the set of possible agreements which cannot be challenged by any sub-coalition. In a world driven by direct agreements, the core frames stable social order(s), preserving individual rationality and satisfying to the Pareto optimality; three reasons to be accepted by the players. In spite of that, the core is neither always a fair solution at variance with the justice mode of cooperation, nor necessarily the best issue of a bargain, in the decentralized behaviour mode. This point is illustrated by the two following examples: Let us assume three players: 1, 2, 3. Each of them knows the outcome of all the possible coalitions of the game ({1,})  ({2})  ({3})  0 ({1,2})  ({1,3})  1, ({2,3})  0 ({1,2,3})  1 Player 1 can agree with player 2, or player 3, or both. He knows that an agreement between players 2 and 3 is not fruitful. If player 1 receives 1

260

Christian Schmidt

and the two other players nothing (the core solution), no player can object against this agreement. One can, however, legitimately contest this solution from a justice point of view. Indeed, without player 2 and player 3, player 1 will never obtain the maximal outcome. In a more subtle way, player 2 and player 3 provide together a necessary information to player 1 in his face-to-face bargaining. This can be agreed, but in the case of the direct cooperation process à la Moulin, player 1 and player 2 could hardly put forward these arguments. It would be different in the ‘justice mode’. In the second example, the following outcomes are associated with the coalitions: ({1})  ({2})  ({3})  0 ({1,2})  ({1,3})  ({2,3})  8 ({1,2,3})  9 The outcome of the great coalition dominates the outcomes corresponding to all the possible sub-coalitions. According to Pareto’s optimality, the core is the result of a face-to-face bargaining between the three players in the grand coalition. One can wonder, however, if each individual player cannot really obtain a better pay-off in playing differently. As already pointed out by Aumann and Dreze, two of the three players can benefit from negotiating together, dropping the third (Aumann and Dreze 1974). Thanks to the pure symmetry of the game, each of these players would reasonably expect a pay-off of 3 instead of 2, if they open the bargaining to the third player in the great coalition as in the core solution. What does this prove? If the core does not contradict the players’ individual rationality, the pay-offs which are derived from it do not necessarily coincide with the best strategic choice of all the players. This is right, but reasoning in these terms implies endogenizing the coalition formation in a strategic process which is out of the direct agreement context. Thus, with the strategic choice of a coalitional structure, we progressively move to another mode of cooperation, ‘the decentralization of behaviour’. The same kind of analysis can be developed with the Nash solution for the ‘decentralized behaviour mode’ of cooperation, and with the Shapley value (and other concepts of value) for the ‘justice mode of cooperation’. Two additional observations have interesting consequences for institutions. In the decentralized behaviour mode, whereas the players have nothing to know about the other players, they must perfectly know the exact results of the game and know that the other players also know them (Moulin 1995: 28). As previously noted, the Nash equilibrium operates as a legal code which is a public information to be read by every citizen. The justice mode, on the other side, requires implementing the solution. This requirement can be realized by an arbitrator or through a social mechan-

Epistemic foundations of organizations 261 ism which is accepted by all the players, as for example, the rule of fair division to share a cake (Steinhaus 1948). However, Moulin’s most important contribution to our topic is elsewhere. Moulin offers the opportunity to elaborate this notion of cooperation in developing the basic ideas which underline these three modes of cooperation. It refers actually to two different meanings – namely, to agree and to co-ordinate. An agreement between players necessarily entails a coordination between their actions, but the reverse is false. Players’ individual strategies can be co-ordinated without any agreement between the players through a supposed mechanism.11 Furthermore, the players either can agree directly upon a substantive matter, as in a contract (‘direct agreement’), or indirectly upon a norm included in the solution concept (‘justice’). As for coordination, it necessitates the existence of guidelines commonly understood by the players (‘decentralized behaviour’). These three modes of cooperation are not mutually exclusive. As, for example, two or three solution concepts related to different modes of cooperation can effectively give rise to the same result. In the well-known ‘stag-hunt’ game, one of the two Nash pure equilibria is also the core of the game viewing with cooperative spectacles. For Moulin, the main interest of this singularity is to open the way to a convergence between the different modes of cooperation. We do not share this position. The single fact that two solution concepts reach the same result in spite of their difference does not reduce the gap between their respective underlying social order. Let us come back to the stag-hunt game. By chance, one of the stag-hunt Nash equilibrium is also a cooperative situation. What does this mean? According to the specific configuration of the game (the pay-offs structure) the two players can co-ordinate their strategy in such a way that the outcome of the game is Pareto optimal and satisfies other conditions which belong to the direct agreement mode of cooperation. But this opportunity creates a new problem in the decentralized behaviour environment: how will the players choose between a coordination on the first equilibrium, which is ‘pay-off dominant’, or on the second equilibrium, which is ‘risk dominant’ according to Harsanyi and Selten’s terminology (Harsanyi and Selten 1988)? This second level of coordination problem has no room in the context of direct agreements. Indeed, a stag-hunt situation can be solved in following one of the other modes of cooperation (‘decentralized behaviour’ or ‘direct agreement’; see Schmidt 2001a). Now, if the players can agree on a cooperative solution it is not surprising that this solution also supports an implicit coordination. In some cases, as in the stag-hunt example, this coordination could take the form of a regulation by a Nash equilibrium. But, beyond the relation between ‘agreements’ and ‘coordination’, such a coincidence actually depends upon the data.

262

Christian Schmidt

Agreeing to agree or to co-ordinate? Greenberg, with his personal interpretation of the standard of behaviour, and Moulin, with his broad notion of cooperation, attempt, by different ways, to unify the various solution concepts of game theory. We will draw out, on the contrary, the background of their difference. Our purpose is to clarify the relation between the standard of behaviour ‘accepted’ by the players and related to each solution concept and their corresponding social order. A first step in that direction is to be found in the connection between the players’ acceptance of a standard of behaviour and a specific mode of cooperation à la Moulin, which is based on agreements and/or coordination. To say that all the players of a game accept a given standard of behaviour is quite the same thing as if they agreed on this standard of behaviour. But this terminological change is not purely semantic. Agreeing to a standard of behaviour implies for the players that their acceptance is a common language between them. Agreeing means here that player 1 accepts the standard and that he knows that player 2 accepts it also, and that he knows that player 2 knows that he accepts it . . . Consequently no player can disagree on an ‘accepted standard of behaviour’.12 In that sense, the first degree of agreement between the players of a game is the acceptance of a well-defined standard of behaviour. As initially specified by von Neumann and Morgenstern, the raison d’être of a standard of behaviour is to lead to an ‘established’ social order via a solution concept. The solution of the game refers to two principal ideas: on one hand, a kind of agreement between the players; on the other, a way to co-ordinate their individual decisions. Such a plain distinction must be refined. First, the two extreme situations of ‘pure agreement’ and ‘pure coordination’ are to be eliminated, because an agreement between individuals and a coordination of their actions are not, by themselves, the theoretical solutions of a game. As previously noticed, the agents become the players of a game when, prior to any decision or movement, they agree to a specific standard of behaviour. Players’ agreements that are derived from the accepted standard of behaviour belong to a second degree of agreements, at the same level as the coordinating processes. Second, there exists a large spectrum of combinations between agreements and coordination among the various solution concepts. We propose Figure 10.3 to classify, by their goal orientation, the different standards of behaviour associated with the various solutions. The first two columns of the figure are ‘agreement oriented’. They traditionally belong to the cooperative games. The two others are ‘coordination oriented’ and are considered as non-cooperative. But the difference between these four types of standard of behaviour is much more subtle. One can either make direct agreements (col. 1) or agree

Epistemic foundations of organizations 263 1 GoalTo achieve oriented possible standard agreements of behaviour

2

3

4

To To reveal To coordinate implement disagreement individual a solution points decisions including a social norm

Pure cooperations

Pure coordinations Solution concepts

– the stable – the The Nash sets Shapley bargaining – the core value solution . . . – the Owen value . . .

The Nash equilibrium

Figure 10.3 Goal orientation

indirectly through the adhesion to a social norm (col. 2). In a strict noncooperative interpretation, the disagreement points are nothing more than the revelation of a mutual incompatibility of the demands emerging from the players. But to co-ordinate the players’ strategies on an equilibrium through a process of best answers (col. 4) is not identical to force their demand to be compatible demands with fixed threats (col. 3). All the pieces of the puzzle are now available. Players have at their disposal some information on the situation where they operate (the outcomes of the different coalitions in a cooperative, the possible individual pay-offs in a non-cooperative game, according to the traditional distinction). This information data can be complete or incomplete, symmetric or asymmetric, between the players. More generally, with these available data, players can organize different economic orders corresponding to different games and their solutions. The ‘established’ order depends on players’ acceptance of a definite standard of behaviour rather than another which would also be consistent with the data. This acceptance is to be analysed as a prior agreement between the players. Such an agreement can be tacit or formal, as in voting procedures. Thanks to this agreement, the private knowledge of each individual approval becomes a common belief among the players. Figure 10.3 illustrates this general framework well. The data are given by the different outcomes associated with each coalition. The three players are supposed to possess all this information. Two different solutions, at least, are consistent with these data and their knowledge by the players. Each of

264

Christian Schmidt

them organizes a different social order, because the solution of its core gives the pay-offs profile (1,0,0) and the Shapley value (4/6, 1/6, 1/6) (see example p. 259). To reach one or the other solution, the players must first agree on their respective standard of behaviour (see Figure 10.3, cols 1 and 2). The relation between the goal and the pattern of organization vary from one standard of behaviour to another. With the core, for instance, the pattern of organizational rules (direct agreements in Moulin’s terminology) dominates the definition of a strict norm for the goal. Therefore, several solutions can belong to the core. The priority is the reverse with the Shapley value (4/6, 1/6, 1/6), in our example, which is a simple application of a social norm, strictly defined as an allocation of the outcome of the game among the players. This norm allocates to each player his marginal contributions to the worth (‘justice’ in Moulin’s terminology). Different mechanisms, including the service of an agency, can be used to perform this profile. Consequently, the content of the prior agreement is not exactly the same in each case. In the first case, it means to agree on some procedural rules; in the second, to agree on a norm of justice. The last step is to explain how the players can agree on a definite standard of behaviour when alternative standards, corresponding to different solutions, are consistent with their available information on the situation. We have stressed the difficulty due to their comparison. But the introduction of strategic considerations can help us to circumvent this difficulty. It is clear in the example studied that player 1 would prefer the core solution, whilst player 2 and player 3 prefer the Shapley value. Such an opposition results from their expected pay-offs in both cases. At first sight, it reinforces the difficulty for all the three players to agree previously on a same standard of behaviour. More reflection shows, however, that this conflict of interests might also open a way to solve it. Indeed, player 2 and player 3 can agree together to reject the core’s prescriptions. Doing that, neither player 2 nor player 3 adopt an irrational behaviour. They know that whatever their decision, they will obtain 0 in the core system. Player 1, on his side, cannot obtain anything by himself in staying alone. If player 2 and player 3 agree to refuse an agreement with player 1, player 1 has nothing to object. Finally, the three players will be penalized in that issue, but only player 1 can in addition regret it. Following this line of reasoning, player 1 reasonably believes that player 2 and player 3 do not agree on the standard of behaviour which leads to the core solution. Such a conviction is a strong incentive for him to move from the core standard to the Shapley standard. As players 2 and 3 cannot obtain more than in agreeing with player 1 on the Shapley solution, they will also tend to agree with him in its corresponding standard. Therefore, in this example, the Shapley order appears, for the reasons which have been developed, ‘socially more stable’ than the core.13 Moreover, such a result does not emerge here from a prior common acceptance of Shapley’s norm of justice.

Epistemic foundations of organizations 265 What we learn from this example remains modest and cannot be extended without great caution. Anyway, it proves that, at variance with a majority of game theorists, to understand the cognitive considerations for which players can agree on gaming standards belongs to the domain of game theory in a comprehensive definition. Such an investigation is to be included in future research agenda.

Notes 1 The broad term of ‘decision-maker’ is used here in order to make a distinction between the ‘agents’ and the ‘players’, which becomes relevant in every game in extensive forms. 2 In game theory, what is called the selfishness of the players is no more than the assumption that each player only maximizes his own pay-offs, without taking into account the others’ pay-offs. Such an assumption is by no means necessary for the study of a game. One can easily assume, on the contrary, that players have preferences on the other players’ pay-offs, without modifying the structure of game theory. 3 In the orthodox theory of individual decision under uncertainty, all the subjective probability distributions over unknown states of the world are supposed to be equally rational. This is not the case in game theory, because players’ uncertainty does not concern unknown states but rational beliefs to be held by the players about each other. Such a difference entails that the consequent axioms built for the theory of individual decision cannot be applied to multi-person game situations. 4 In a series of models, Volterra makes distinctions between different situations: one and the same species with many individuals, two opposite species, more than two species (Volterra 1931). 5 This judgement can be reversed on the grounds of other arguments: the description of non-cooperative games is much more detailed than the cooperative games (see the extensive form in comparison with the coalitional form), including the mechanism of information transmission among the players. Consequently, the rules of a non-cooperative game are generally more precise, in connection with the corresponding solution concept. Those rules can also be viewed as the starting point of institutions (Schotter 1981). 6 When n  2, the Nash equilibrium must commonly be known by all the players. Incidentally, this clearly points out the rule of the number of the players in the analysis of information and knowledge conditions. 7 With the important exceptions of the Shapley value. 8 Greenberg himself recognizes the point in his comment of a similar example. But he gives a different interpretation of this awkwardness. According to him, a rational recommendation to all the players of the game might conflict with the individual rationality of the players. Thus, standards of behaviour give recommendations and not predictions on players’ behaviour (Greenberg 1990: 166–167). 9 It can be argued in an opposite direction, that according to c3 the players can refuse their strategies C and c with an easy mind, because each of them takes for sure that the other does not choose this strategy, no matter what the other player does. 10 For Moulin, however, the three modes of cooperation cannot be treated as Greenberg’s social situations. Thus, there is no one-to-one correspondence between a given mode of cooperation and a specific solution concept. In that

266

Christian Schmidt

spirit, the core, the Nash equilibrium and the Shapley value are used here as illustrative examples. 11 This is precisely the case with the Nash bargaining model, which gives rise to situations labelled as ‘anti-conflict’ when players’ demands are more than mutually compatible (Nash 1950b). 12 A standard of behaviour is neither an event nor a state of the world, but a code. To accept a code has a meaning for the players if, and only if, each player knows that the other players accept it, and knows that they know, and so on, to infinity. 13 This example provides an intuitive argument against the intrinsic instability of the Shapley value, which has some connection with its so-called irrationality (Roth 1980, 1986).

References Aumann, R.J. (1996) ‘Reply to Binmore’, Games and Economic Behavior, 17: 138–146. —— (1999a) ‘Interactive Epistemology I: Knowledge’, International Journal of Game Theory, 28: 263–300. —— (1999b) ‘Interactive Epistemology II: Rotability’, International Journal of Game Theory, 28: 301–314. Aumann, R.J. and Brandenburger, A. (1995) ‘Epistemic Conditions for Nash Equilibrium’, Econometrica, 63: 1160–1180. Aumann, R.J. and Dreze, J.H. (1974) ‘Cooperative Games with Coalition and Structures’, International Journal of Game Theory, 3: 217–238. Binmore, K. (1990) Essays on the Foundations of Game Theory, Oxford: Basil Blackwell. —— (1993) ‘De-Bayesing Game Theory’, in K. Binmore, A. Kirman and P. Tani (eds), Frontiers of Game Theory, Cambridge, Mass.: The MIT Press, 321–339. Borel, E. ([1921] 1953) ‘Theory of Play and Integral Equations with Skew Symmetric Kernels’, Econometrica, 21: 91–100. (Original 1921 version in French.) —— (1924) ‘Sur les jeux ou interviennent le hasard et l’habileté des joueurs’, Association Française pour l’avancement des sciences (1923), reprinted in Théorie des Probabilités, Paris: Hermann, 204–224. —— (1939) Valeur pratique et philosophie des probabilités, Paris: Gauthier-Villars. Greenberg, J. (1990) The Theory of Social Situation. An Alternative Game Theoretic Approach, Cambridge: Cambridge University Press. Harsanyi, J.C. (1977) Rational Behavior and Bargaining Equilibrium in Games and Social Situations, Cambridge: Cambridge University Press. Harsanyi, J.C. and Selten, R. (1988) A General Theory of Equilibrium Selection in Games, Cambridge, Mass.: The MIT Press. Kohlberg, E. and Mertens, J.F. (1986) ‘On the Strategic Stability of Equilibria’, Econometrica, 54: 1003–1037. Luce, R.D. and Raiffa, H. (1967) Games and Decisions, New York: Wiley. Maynard Smith, J. (1982) Evolution and Game Theory, Cambridge: Cambridge University Press. Moulin, H. (1995) Cooperative Microeconomics: A Game Theory Introduction, Princeton, N.J.: Princeton University Press. Nash, J.F. (1950a) ‘Equilibrium Point in n-Person Games’, Proceedings of the National Academy of Sciences, 36: 48–49.

Epistemic foundations of organizations 267 —— (1950b) ‘The Bargaining Problem’, Econometrica, 18: 155–162. Roth, A.E. (1980) ‘Values for Games without Side-Payments: Some Difficulties with Current Concepts’, Econometrica, 48: 457–465. —— (1986) ‘On the Non-Transferable Utility Value: A Reply to Aumann’, Econometrica, 54: 981–984. Schmidt, C. (2001a) ‘Does Nash Equilibrium Imply the Players’ Rationalilty in Non-Cooperative Games?’, mimeo, Lesod, Paris, Université Paris-Dauphine. —— (2001b) La théorie des jeux: Essai d’interprétation, Paris: P.U.F. Schotter, A. (1981) The Economic Therapy of Social Institutions, Cambridge: Cambridge University Press. Spohn, W. (1982) ‘How to Make Sense of Game Theory’, in W. Stegmuller, W. Balzer and W. Spohn (eds), Philosophy of Economics, Berlin: Springer-Verlag, 239–270. Steinhaus, H. (1948) ‘The Problem of Fair Division’, Econometrica, 16: 101–104. Volterra, V. (1931) Leçons sur la théorie mathématique de la lutte pour la vie, Paris: Gauthiers-Villars. von Neumann, J. ([1928] 1959) ‘Zur Theorie der gesselschaftsspiele’, Mathematische Annalen, 100: 295–320. English trans: ‘On the Theory of Game of Strategy’, in A.W. Tucker and R.D. Luce (eds), Contributions to the Theory of Games, Vol. IV, Annals of Mathematics Studies, Vol. 40, Princeton, N.J.: Princeton University Press, 13–45. von Neumann, J. and Morgenstern, O. (1944) Theory of Games and Economic Behavior, Princeton, N.J.: Princeton University Press.

11 Interactive learning and technological knowledge The localised character of innovation processes1 Cristiano Antonelli, Jean-Luc Gaffard and Michel Quéré

Introduction There is a growing insistence that economic systems are facing a period of structural change, this being mainly characterised by a transition from a commodity exchange manufacturing system towards a weightless one. Advanced economies are facing the ultimate transition towards a service economy with a declining share of the gross national product generated in the manufacturing of physical goods. In such a transition the direct content of knowledge is enhanced and appears more and more important. This new context makes even more evident the limitations of an economic analysis traditionally based upon the biunivocal relationship between prices and quantities in all adjustment processes where changes in knowledge and in its distribution among agents could not take place. During such a transition, it becomes much more fuzzy to identify product characteristics as well as to understand the dominant regulation of economic activity due to the fact that price mechanisms no longer express all the subtle aspects of ‘exchange mechanisms’. Many scholars try to qualify the seemingly new conditions for the working of economic systems by referring to a so-called ‘learning economy’. The development of the ‘learning’ or ‘knowledge-based’ economy is based on the fact that contemporary life is facing high qualitative changes, be they characteristics of production or markets. Both sides of economic systems appear very sensitive to rapid change, and expectations from firms and households as well seem to be much more volatile than they were in the recent past. In this chapter, we want to provide a critical analysis for this hypothesis of a transitory phase for economic systems. More precisely, we want to argue that the ‘learning economy’ has less to do with the usual definition of learning in economics than with co-ordination issues at different stages in the working of economic systems (see Amendola and Gaffard 1988, 1998). If we share the fact that the latter are facing continuous innovation processes, and that this situation seems to require a deeper attention to

Learning and technological knowledge

269

the conditions of learning characterising the economic functioning, we consider that this is neither a new phenomenon nor that learning is central to understand the economic working of social societies. Considering that innovation has always been the source for economic growth, refining economic activity, be it through the emergence of new capabilities or the strengthening of existing ones, has not to be thought of as a new requirement for economic systems. The emphasis on a learning economy mainly expresses changes in the conditions to innovate, and this is essentially what will be under discussion within this contribution. As noticed, the recent emphasis on a ‘learning economy’ results from the increasing volatility of commodities and from deep changes in the conditions to manufacture products, be they the inputs involved, the related distribution networks, or the customers’ influence on design and technical specifications. As a consequence, considering a ‘learning economy’ essentially implies an ability to adapt to this new economic environment and a capability to set up suitable tools and means in order to face this very evolving context. Among the crucial aspects of this economic challenge, the conditions for firms’ innovative behaviours to be economically viable require specific attention. At first glance, it is possible to associate innovation in the production of new knowledge to be embodied in profitable activities. As a consequence, what appears very challenging is that economic systems are facing a context where the generation of knowledge has never been so much in proximity with its economic use. This questions the conditions of emergence, diffusion, and economic use of new knowledge. More precisely, this importance of knowledge in productive activities has to do with the debate on its localised character (Antonelli 1995, 1999). In other words, we are now facing a context where the generation and diffusion of knowledge becomes specific, depending on the local conditions into which this knowledge is embedded. There is no mistake here: the ‘localised’ character of this knowledge does not mean that it centrally refers to a geographical meaning. Rather, it concerns the architecture of firms’ productive networking; that is, the set of inputs and relations required to implement their activities. It can display a geographical localised content, but in the majority of cases the influential area of firms is world-wide and as a result localised character of knowledge is disconnected with geography and physical proximity. The localised character arises mainly from the idiosyncratic features of the bottom-up learning process associated with the techniques in use and the switching costs engendered by the irreversibility of sunk factors. Third, the localised character is determined by the chains of weak and strong complementary effects among firms, and among technologies which closely link the economic system and are appreciated in terms of economies of scope (if internalised) and externalities. In this chapter, we aim to explore the consequences of the localised

270

C. Antonelli, J.-L. Gaffard and M. Quéré

character of knowledge as regards the central focus on co-ordination issues for economic systems.

The need for a better understanding of technological knowledge Actually, economics has already drawn attention to the importance of knowledge for firms’ behaviours. Even if this attention is not new, and one can refer here to the fact that knowledge already was of central importance in Penrose (1959) to cope with the problem of the growth of the firm, recent considerations on knowledge are not completely satisfying because of their difficulty to cope with the very economic problem that firms are facing. Many of these attempts are considering knowledge for itself and focus on the conditions by which knowledge becomes available to firms or diffuse within the economy. As a consequence, those economic attempts have incorporated learning in knowledge theories (Nonaka and Takeuchi 1995); but there is, here, a misleading trend in the representation of the contemporary working of economic systems. One can think about the learning economy as the combination of two complementary blocks: one could be devoted to the production of new knowledge and encapsulate all the necessary infrastructure to provide the rest of the economy with new resources and knowledge, including academic and scientific resources, R&D labs for large firms, ‘high-tech’ SMEs, and the like; the rest of the economy could then be presented as benefiting from this set of competencies and capabilities that orient and organise the potential learning for the economy as a whole. Recent attempts in the economic literature towards the understanding of national systems of innovation can support such a view. But it fails to completely cover the conditions by which innovation occurs in firms by neglecting the other side of the knowledge problem. Knowledge also results from firms’ daily activities in the sense that it emerges from the set of routines, the usual workings of production, and market interactions. On the one hand, no one can reject either conception of knowledge-creating conditions, as noticed by Foray and Lundvall (1996: 13); although on the other, no one has actually provided a satisfactory framework to consider the interactive character of both dimensions. This is essentially due to the fact that the understanding of innovation cannot satisfy itself by a theory of knowledge. It requires more than that, and especially a better understanding of how fundamental imbalances in resources induced by any process of change for firms’ activities (mainly human and financial imbalances) are managed over time. Our analysis elaborates on an approach where agents are able at each point of time to generate new knowledge and to try to make use of it in their specific market and production contexts. A bottom-up approach to the generation of technological knowledge, as distinct from scientific knowledge, finds a clear reference in the analyses put forward by Hayek

Learning and technological knowledge

271

(1937, 1945) and Richardson (1960) on the role of economic knowledge in the market place. Successful technological innovations are the result of discovery, as opposed to invention. The actual discovery of new suitable technologies emerges in the marketplace, in out-of-equilibrium conditions, where agents do more than adjust prices to quantities and vice versa. A variety of new technologies at each point in time and space are tested in the marketplace and sorted out. Further recombination and integration of localised knowledge, together with the creation of appropriate coalitions of users and producers, may eventually generate successful technologies. In this context, the understanding of the specific mechanisms by means of which new knowledge is generated, recombined, experimented and eventually applied, becomes a key issue – an issue which cannot be separated from the specific competitive and productive context into which firms’ conducts and strategies are embedded. Complementary effect among agents in the identification of appropriate technological solutions, as a guiding post for the formation of effective coalitions, is the key to understanding such dynamics, a dynamics where competition often follows co-operation sequentially in selecting and assessing the basic technological requirements and interfaces. This is the reason why we intend to focus on a specific definition of knowledge, which we will refer to as ‘technological knowledge’. Technological knowledge is not used in this chapter according to its usual definition. Technological knowledge is more than technological resources in that it incorporates a specific ability to organise, control and combine technological resources with the aim of making the firm profitable as well as ensuring its ability to change its activities over time. Technological knowledge has more to do with the localised character we previously referred to. Technological knowledge incorporates the interactive character that a company has to manage with its productive environment. This of course refers to other firms being involved as suppliers of resources for production, but it also refers to customers and their influence in the working of production processes as well as to non-productive organisations such as business associations, banking institutions and science and technology institutions. Taking all these components into consideration is a necessary condition in the understanding of the localised character of technological knowledge – that is, to understand the conditions firms are facing in order to ensure the evolution of their activities from an economic viewpoint.

Some empirical evidence about the localised character of technological knowledge: contrasting sectoral examples Because of the interactive character of technological knowledge, there is an obvious need to consider in more detail the huge variety in the mechanisms and designs suited to its production. There are contrasting sectoral patterns that show the relevance of the localised character of technological

272

C. Antonelli, J.-L. Gaffard and M. Quéré

knowledge for economic reality. This is due to the complexity of productive requirements as well as the variety of its implications in terms of interfirm relationships. In that respect, we can learn from empirical contexts and elaborate on sectoral differences in the ways firms are facing technological knowledge. There are important difficulties to overcome in order to express the diversity of sectors with regard to the emergence and diffusion of technological knowledge. The method we have chosen consists of mapping sectors’ characteristics, according to a list of variables we consider central to qualify firms’ formal and informal relationships. The latter include factors purely internal to firms (be they technological or organisational), information related to market characteristics, and environmental variables related to the institutional infrastructure to which firms are subjected. In the telecommunications industry, firms are evolving in a multitechnological context and are facing an explosion of potential uses. On the supply side, this industry is composed of different layered activities: the equipment supply (terminal equipment, network equipment); the network provision (lines, switches, backbones); the service provision (short- and long-distance telecommunications, mobile telecommunications, data transmissions, the Internet); the entertainment provision (edition, multimedia, broadcasting, virtual reality, software and middleware); the commercial and distribution level. On the demand side, some new requirements emerge: the necessity to have simultaneous access to different types of information (texts, audio and video documents); the communication within or between different groups of users; the user-friendliness, mobility, reliability and safety of communication protocols. Technological knowledge here has to do with the capability to reduce mismatches between the supply and demand sides during the transformation process faced by that specific industry. Communication and knowledge channels have the central aim of co-ordinating knowledge flows so as to reconcile technological opportunities with market potentialities in a context which is facing rapid changes, be they the supply or demand sides of the industry. A contrasting example to telecommunications lies in the aircraft industry, where no significant mismatch between the supply and demand sides of the sector is noticeable. On the contrary, an obvious stability characterises producer–user relationships. However, the industry is facing great changes due to the very high innovative character of technological knowledge within it. Even if the products made by the leaders in this industry are stable in the long run, the conditions under which they are made and the technologies they incorporate are rapidly changing. Here, the organisational design among the set of firms involved in aircraft industry (large firms, large SME relationships, or KIBS) is central in the understanding of the technological knowledge concerned. Aircraft industry outputs are system-products that require a complex network of inter-firm relationships from the conception stage to the manufacturing. This complex system of

Learning and technological knowledge

273

productive relationships makes the analysis of how knowledge emerges especially relevant. The analysis can be shared among partners, and affects the way it is organised to conduct the strategy of continuous technological innovation that characterises the evolution of this industry and explains its international competitive success. The development of the so-called ‘life sciences’ industry offers an example of a more informal structuring for an industry and, consequently, for its capability to produce and diffuse technological knowledge. Structural uncertainty in terms of productive potentialities are stronger in that case because of pressure from academics and consumers. The term ‘life sciences’ industry actually refers to a set of industrial activities where knowledge is continuously renewed and reorganised, and where the generation of knowledge from academics flows very rapidly through complex communication channels into industrial applications. As such, this industry shows the diversity of ways by which institutional and productive constraints are set together and help to design potential productive opportunities progressively. In the case of the ‘life sciences’ industry, a special role is played by ‘high-tech small and medium-sized enterprises’. This type of firm appears as the central channel by which technological knowledge is created and progressively designs productive opportunities. They can be thought of as a structural link between academics and large pharmaceutical firms that becomes essential to transform scientific discoveries into economic activities. However, this transformation process is very uncertain and complex because of the various components that need to be co-ordinated towards the same target. The car industry is another interesting example to map the complexity of technological knowledge channels. What dominates the current evolution in this industry is the process of out-sourcing knowledge implemented by large car manufacturers. The systematic internalisation of external technological knowledge, available through the establishing of technological clusters centred upon mechanical engineering, and the complementary valorisation of internal skills, had been the driving force behind the accumulation of internal knowledge and technological capability of firms. Here, the growth of large corporations seems also to be the result of specific competencies and managerial routines, which have been able to keep open a variety of communication channels between the ‘walls’ of the company and the external environment. The corporate organisation of the production of knowledge now is shifting away from the ‘intramuros’ model based upon well-specified and self-contained research and development activities. A variety of tools are nowadays used by corporations to take advantage of external knowledge and minimise the tragedies of intellectual enclosures. Intentional participation into technological districts and technological clusters and business strategies, characterised by flexible and porous borders, appears to be increasingly practised by a growing number of corporations. Consequently, technological

274

C. Antonelli, J.-L. Gaffard and M. Quéré

knowledge requires local practices of interactive learning among partners that are not embedded in the firm alone (large or small). Local institutions (such as academic, educational training, or business and technical associations) are necessary to promote this local embeddedness. Those sketchy considerations about sectoral diversity in the production, diffusion and accumulation of technological knowledge have the sole objective to shed some light on the intrinsic differences among productive contexts within the European industrial reality. It especially expresses the diversity in the mechanisms of generation and diffusion of knowledge at the sector level. It shows the relevance of considering the localised character of technological knowledge as an actual issue to understand industrial dynamics and, consequently, to qualify innovative behaviours of firms through a better understanding of the architecture of intra-firm and interfirm relations; that is, the set of inputs and relations required to implement their activities (Metcalfe 1995). However, those contrasting sectoral patterns raise the difficult question of understanding the reasons for such a diversity, as well as what kinds of analytical tools can be provided to qualify it.

How to cope with the localised character of technological knowledge We have emphasised the diversity in the conditions of learning coming from the variety of technological knowledge – that is, from differences in systemic constraints faced by productive contexts. We have shown the variety of this localised character of technological knowledge through a few sectoral patterns. In order to cope with that variety, two central points have now to be emphasised. One is the variety of organisational designs that enable the emergence of technological knowledge and its transformation into profitable applications. The other is the obvious diversity of communicative channels that allows for the diffusion of technological knowledge within innovation systems. Organisational designs Technological knowledge cannot be firm- or context-specific. As such, technological knowledge is systemic and questions the understanding of firms’ capabilities. This systemic aspect largely depends on productive contexts, and the analysis requires an ability to deal with the peculiarities of those productive contexts. This questions the internal characteristics of a firm; that is, the way by which functional and divisional operations are coordinated. It also questions the way a firm’s organisation interacts with its environment. Recent emphasis in the economic literature has made more explicit how public resources and incentives, academic infrastructures, and

Learning and technological knowledge

275

firms’ innovative behaviours interact in a complex manner and constitute innovation systems that favour the generation and use of technological knowledge. Understanding the conditions required for the emergence of technological knowledge implies the simultaneous analysis of those three components. In other words, knowledge is organised in bundles. Strong complementarity, hence spillovers, hence increasing returns, occurs only within such bundles. The difficulty comes from the fact that no one unique and performing model exists; on the contrary, there is a huge variety of innovation systems. It is this variety that emphasises the difficulty facing the implementation of technological knowledge in contemporary economies. Therefore, there is a need for clarifying models of generating and diffusing technological knowledge. A strong challenge lies in understanding the variety resulting from the diversity of productive constraints faced by innovative firms. Organisational designs used to produce and experiment with technological knowledge can be characterised as a combination of productive and institutional contingencies. Productive contingencies come from the fact that the firms’ organisational designs required to perform the generation of technological knowledge are moulded by productive constraints and, especially, by the nature and/or history of the sectors under scrutiny. Three main requirements have to be taken into account: the fact that manufacturing a product increasingly means incorporating numerous technologies; that products are becoming more complex in the sense that they require increasing skills for their manufacturing to become profitable; that aggregating numerous technologies increase the co-ordination needs (i.e. costs and difficulties) within the firm to engage in innovative potentialities. All those requirements largely question the characteristics of a firm’s organisational design, as well as the understanding of the firm’s innovativeness in accordance to its use of technological knowledge. However, those productive contingencies do not completely explain why technological knowledge can become profitable and transform a ‘body of technological understanding’ into a ‘body of economic practices’ (see Pavitt 1998). They need to be complemented and incorporated in a set of institutional constraints that result from the external environment faced by agents to promote and implement innovative choices. Institutional contingencies not only refer to the institutional structure of production that characterises the productive context (i.e. the complex network into which a firm is embedded, including suppliers, customers, cooperative partners, sub-contractors, etc.) but also to the institutional infrastructure that appears specific to the related industry (business and professional associations, banks, academics, etc.). The combination of productive and institutional contingencies results in a variety of productive contexts, the diversity of innovation systems, and the relative ability to ensure a suitable evolution of economic systems

276

C. Antonelli, J.-L. Gaffard and M. Quéré

(see Quéré 2000). Then, the localised character of technological knowledge has to be analysed through the combination of both (productive and institutional) aspects, and considering the characteristics of this combination largely allows a better understanding of the variety of productive contexts aimed at favouring the viability of firms’ innovative behaviours. The variety of communication and knowledge channels External technological knowledge does not fall from heaven like manna. It cannot be considered as a usual input that can be immediately internalised by firms. It requires specific absorption and ‘listening’ costs which depend upon the variety of codes and the number (and type) of communication channels selected by firms. The costs of the production of knowledge, including such communication costs, are lower for firms able to establish co-operative relations and access to the pool of collective knowledge made available. Appropriability also is affected. The opportunity costs engendered by the uncontrolled leakage of technological knowledge are lower, the higher the mutuality and trust conditions in place within a group of firms. For ‘given innovation costs’, including research, learning and communication activities, a higher collective output can be identified. The latter makes possible the existence of external increasing returns in the production of knowledge: the larger the number of connected firms, the larger the amount of knowledge generated. Therefore, communication plays a central role in such a context. Communication is instrumental in that it allows users and producers to identify, qualify, explore and assess the potential for knowledge externalities. As communication contributes to make knowledge externalities actually relevant from the perspective of potential users, communication channels appear very crucial to render knowledge opportunities efficient from an economic viewpoint. While knowledge holders cannot prevent the dissipation of their knowledge, perspective users may be unable to make good use of it. As a consequence, the role of communication in the production of technological knowledge is emerging as an important area for theoretical and empirical research in the economics of innovation. However, the understanding of the conditions by which such communication takes place is still developing. Though a large consensus has been established about the key role of knowledge externalities in the production of new knowledge, the conditions by which those externalities appear require greater analysis. The variety of knowledge channels is actually puzzling. There is a central problem to be solved – the embedded character of many of the communicative channels. First of all, for communication to take place at least two parties must be involved: communication is inherently a collective activity. Second, the establishment of effective communication links requires a long time to implement, and also the codification of shared

Learning and technological knowledge

277

protocols and communication rules. Third, effective communication relies on material as well as immaterial infrastructures, which can be created over time and with reciprocal consensus. Finally, in the short term, the amount and importance of the actual traffic of signals and information can vary greatly; in the long term, however, communication takes place and effective successful transfers of information between parties can also take place. For knowledge externalities to appear and be effective it is necessary to address this obvious diversity of communication and knowledge channels. In some cases, for knowledge externalities to appear and learning to diffuse, informal relationships are essential; in other cases, sharing common equipment or an infrastructure seems to be the answer; in yet other cases, contractual commitments among firms appear as a necessary condition; still others make the need for co-operative projects or jointcompanies more explicit, etc. This variety confirms the difficulty to find appropriate mechanisms of governance for the generation, use and distribution of localised technological knowledge and the broad array of factors that affect the combination of learning, socialisation, recombination and R&D. However, the understanding of why some innovation systems are more effective than others is still very unclear. This is why we propose to put specific emphasis on this localised character of technological knowledge. Towards an operational approach to cope with the localised character of technological knowledge Emphasising the importance of the localised character of innovation processes is a means of considering the importance of the structural and sectoral diversity to which we previously refer in the creation, diffusion and accumulation of technological knowledge. As a consequence, to face this diversity, the essential difficulty is to figure out the peculiarities of technological knowledge as well as the related communication and knowledge channels, and, consequently, to deduct from this variety the relative effectiveness of those channels in their ability to generate and diffuse technological knowledge. This can be done by organising a series of specific criteria to make more explicit the productive and institutional constraints faced by firms to which we refer in mapping the contrasting characters of sectoral patterns. Those criteria can be grouped in three complementary categories: internal to the firms; external, but depending on inter-firm relationships; external, but depending on institutional characteristics: 1

Internal knowledge essentially refers to organisational criteria and includes factors such as the characteristics of internal R&D and its related use in the production process, the firms’ organisational designs

278

2

3

C. Antonelli, J.-L. Gaffard and M. Quéré and their ability to favour internal learning, and the internal technological infrastructure (intranet capabilities and, more importantly, any other I&T facilities). External knowledge related to the requirements of products essentially includes the formal and informal relationships that are seen to be necessary for an organisation to succeed in the industry. This refers to technological peculiarities such as the multi-technology characterisation of the sector products, the role of capital equipment and physical infrastructures. It also refers to individual mobility as well as the importance of user/producer relationships for the evolution of demand, the importance of knowledge-intensive business services (KIBS) and, more, importantly, the appropriability conditions of sector-specific technological knowledge. A specific aspect of external knowledge lies in the growing importance of the Internet interface in its ability to favour the acquisition and diffusion of technological knowledge. External knowledge related to institutions refers to ‘environmental’ criteria such as the importance of public institutions for the evolution of demand (direct support, regulatory framework), the relative importance of knowledge-sector characteristics (tacit/codified-individual/ collective-generic/specific), the importance of academics and of related institutions in the generation of technological knowledge and, more largely the type and importance of knowledge externalities.

This classification helps to cope with the diversity and the underlying complexity in the mechanisms driving the working of technological knowledge and its implementation into new productive activities. This complexity is basically due to the multi-technological character that a sector’s patterns exhibit, where a variety of coexisting and partly complementary knowledge is identified. Knowledge in fact can be conceived as a single folder of a variety of specific and localised knowledge, each part of which has a specific context of application and relevance. However, strong complementary effects exist among technological knowledge and help make the folder a single container. In a mono-technological context, direct competitors can make rival use of proprietary knowledge and reduce its economic value for original holders. In a multitechnological one, perspective users are not direct competitors and external knowledge is an intermediary input which, after proper recombination and creative use, becomes a component of the localised production process of new knowledge. Local cumulativeness and indivisibility are clearly important attributes of technological knowledge: new knowledge is built upon previous knowledge, and indivisibility is relevant both diachronically between old and new technologies as well as horizontally among the (limited) variety of new technological knowledge being introduced at each point in time.

Learning and technological knowledge

279

Therefore, technological knowledge is crucially embedded in local contexts and it is important to qualify this localised character. Specifying the previous criteria helps to characterise the systemic dimension of this localised embeddedness of technological knowledge in that it offers a sort of ‘structural’ map aimed at ordering the diversity of technological knowledge characteristics and infers the conditions by which innovation occurs and becomes feasible and diffuse within the production system. From such a mapping, it becomes possible to identify the main characteristics of the communication and knowledge channels that appear effective, as regards the productive and institutional constraints encountered by firms’ innovative behaviours. This analysis contributes to the better identification of technological knowledge as a collective activity where potential knowledge externalities, because of the active implementation of communication activities, can be shared and become the source of major increasing returns. Such a collective character, however, is actually workable only when localised within circumscribed regional and/or technological environments. The costs of communication, and the fall in the positive effects of knowledge externalities associated with dissipation, driven by the increase of distance and heterogeneity among users and producers, limit the scope of fruitful interaction.

Policy conclusions Discussion about the localised character of technological knowledge is correlated, but it also largely contrasts with the current debate on the working of the learning economy. Our analysis provides an attempt to link questions related to the learning conditions of an economic system, not directly to a theory of knowledge by itself but to the co-ordination issues that underlie a more suited theory of innovation. This shift in the analysis makes particular sense when considering policy implications and recommendations. The fact that the conditions required for the generation and use of technological knowledge cannot be but systemic and localised in their aim is deducted from the intra- as well as the inter-sectoral diversity of institutional designs favouring the generation and the profitable use of technological knowledge. Complex characteristics of technological knowledge justify the need for considering centrally the systemic character of policy-making. Science and technology policies, industrial policies and competition policies are all components influencing the working of technological knowledge. They need to be articulated effectively in order to perform their capability to support firms’ innovative behaviours. The analysis of the localised character of technological knowledge is of central importance in order to understand interactions among those complementary policy aspects, as well as to discuss the scope of coherent innovation systems that the interplay among policy decision-making and firms’ or sectors’ evolution can design.

280

C. Antonelli, J.-L. Gaffard and M. Quéré

Note 1 The financial support of the European Union Directorate for Research within the context of the Key Action ‘Improving the socio-economic knowledge base’ to the project ‘Technological Knowledge and Localised Learning: What Perspectives for a European Policy?’ carried on under the research contract No. HPSE-CT2001-00051 at the Fondazione Rosselli is acknowledged. The work has benefited from ongoing discussions with all the members of the TELL working group and specifically from the comments of Martin Fransman, Bruno van Pottelsberghe, Pier Paolo Saviotti, Mario Vale and Virginia Acha. The comments of two referees have been most useful in the drafting of the final version.

References Amendola, M. and Gaffard, J.L. (1988) The Innovative Choice. An Economic Analysis of the Dynamics of Technology, Oxford: Blackwell. —— (1998) Out of Equilibrium, Oxford: Clarendon Press. Antonelli, C. (1995) The Economics of Localized Technological Change and Industrial Dynamics, Boston: Kluwer Academic Publishers. —— (1999) The Microdynamics of Technological Change, London: Routledge. Foray, D. and Lundvall, B.A. (1996) Towards a Learning Economy, OECD Report, Paris. Hayek, F.A. (1937) ‘Economics and Knowledge’, in F.A. Hayek (1948) Individualism and Economic Order, Chicago: University of Chicago Press. —— (1945) ‘The Use of Knowledge in Society’, American Economic Review, 35(4): 519–530. Metcalfe, J.S. (1995) ‘Technology Systems and Technology Policy in Historical Perspective’, Cambridge Journal of Economics, 19: 25–47. Nonaka, I. and Takeuchi, H. (1995) The Knowledge-Creating Company, Oxford: Oxford University Press. Pavitt, K. (1998) ‘Technologies, Products and Organisation in the Innovating Firm: What Adam Smith Tells Us and Joseph Schumpeter Doesn’t’, Industrial and Corporate Change, 7(3): 433–452. Penrose, E. (1959) The Theory of the Growth of the Firm, Oxford: Oxford University Press. Quéré, M. (2000) ‘Innovation, Growth, and Coordination Through Institutions: A Discussion About “Innovation Systems” ’, in O. Fabel, F. Farina and L. Punzo (eds), European Economies in Transition, London: Macmillan. Richardson, G.B. (1960) Information and Investment. A Study in the Working of the Competitive Economy, Oxford: Clarendon Press.

12 Cognitive economies and the ‘nature of the district’ Margherita Turvani

Introduction In ‘The Nature of the Firm’, Coase quotes D.H. Robertson who notes that it is possible to find ‘islands of conscious power in this ocean of unconscious co-operation like lumps of butter coagulating in a pail of buttermilk’ (Coase 1937: 386). Coase asks why is it so. As we know, he contends that the balance between the costs of producing in-house and the costs of using the market explains the existence of firms ‘coagulating transactions’ within the market. In this chapter I want to reframe the same argument at a different scale, and ask why is it possible to see firms lumping together in a specific ‘context’ that it is named ‘district’? Theoretical and applied research on the industrial district shows that this ‘context’ is thick; it is made up of economic transactions which evolve together with the social and territorial relations (Becattini 2001). These features of the district attract the interest of many scholars and policy-makers, trying to find out the conditions for existence and for replication of such experiences. In most cases their analysis is guided by two different frameworks: one to look at the working of the firm and another one to look at the network of relations among firms and, more generally, among different agents and institutions operating at the economic and social level within the specific context of the industrial district. I contend that the same reasons explaining the ‘nature of firm’ and its existence should be transferred to understand the ‘nature of the district’ and the emergence of industrial clusters. A cognitive perspective for looking inside the black box ‘of the firm is an appropriate tool to deal with ‘the nature of the district’, describing it in terms of the knowledge-creation process that takes place within and among firms, relying on the development of ‘cognitive economies’ for individuals and for firms. The chapter is organised as follows. The first section offers a brief survey of the theories of the firm with respect to the role of information-processing and knowledge-creating processes within organizations. In the next section I describe how individual cognition develops at the individual level and in a specific context to highlight the social dimension of

282

Margherita Turvani

the knowledge-creation process and the role of organizations in creating the social community which sustain the production of knowledge. The third section discusses the most common understandings of the ‘nature of the district’, showing that they are characterized by the isolation of an entity, the firm, within a ‘context’ which defines its outside or environment. The fourth section applies a cognitive perspective to the functioning of firms, showing how it improves our comprehension of the dynamic boundaries between the firm and its outside ‘context’. Penultimately, the fifth section shows how these dynamics within the context of a process of a collective generation of knowledge give rise to the emergence of distinctive ‘lumps’ of knowledge, which by means of entrepreneurial action gets organized in firms (this being discussed in the sixth section). Conclusions follows.

Information, knowledge and the nature of the firm To clarify how a distinction between information and knowledge is useful to discuss the nature of the firm I refer to the definition – utilized by Nonaka and Takeuchi (1995) – of information as a good capable of producing knowledge.1 In economics the two terms are frequently confused. Knowledge is simply made up of information and the latter is described as a set of data about the states of the world. This set is close-ended and can potentially be acquired. This overtly contrasts with the open-ended nature of knowledge, the processes of discovery and creation of new forms of knowing (Loasby 1991). If information is capable of giving rise to knowledge, this cannot or does not happen in a way that can be determined a priori. In the case of knowledge being effectively built, its relation with a specific closed-ended finite set of data may not necessarily be reconstructed. Interpretative ambiguity and uncertainty become features of the process which gives rise to knowledge. Fransman (1994: 716) remarks, ‘knowledge (belief) will be used to interpret incomplete information’, and the results of this process will always be subject to verification, revision, and comparison. It is precisely for this reason that knowledge is an unfolding open-ended, incomplete and local process (Hayek 1952, 1945). This approach calls into question the interactive dimension in individual and social cognition. Information and knowledge are not easily combined and accumulated, and the media through which knowledge flows become relevant. Information and knowledge play different roles in the theories of the firm. Some theories focus on the information costs encountered in completing transactions – both internal and with the market; some theories see the firm as a collection of forms of knowledge. In the first case individual agents manage a variety of information problems, all implying costly solu-

Cognitive economies and the ‘district’ 283 tions: organizations are seen as information-processing devices and the firm is depicted as a nexus of contracts.2 In the second case rationally bounded agents are unable to use optimizing calculation and they develop local scale problem-solving techniques. Organizations are thus described as a set of local forms of know-how, fostered by the recurrent activity of problem-solving. The representation of these forms of knowledge, and the way in which they are generated, selected, and modified, becomes the field of research. Following Marshall’s contribution, knowledge becomes a strategic factor pointing the way to the development of a different perspective to understand the nature of the firm. The resource-based view of the firm (Penrose [1959] 1995) and the studies on bounded rationality and organizational learning (Cyert and March 1963) offered the tools to study why firms differ in knowledge management and development and why these differences persist through time, building up a firm’s specific strength or weakness.3 Differences are lasting and self-renewing because knowledge and skills are accumulated through learning, and learning ‘happens’ in specific time and space (Nonaka and Takeuchi 1995). Organizational learning (March and Simon 1958; Senge 1990; Argyris and Schön 1996) explains why knowledge and skills are not easy to imitate, transfer and trade: learning is not simply a process of internalization of information but is also an interactive process implying individual participation in communities of practice (Lave and Wenger 1991).

Cognition in context Cognitive processes reside in the mind of the individual, but the interactive process between individual and the community to which s/he belongs moulds his/her cognitive development and the outcome of cognition.4 Economics was opened to a new frontier of research when Simon (1986) elaborated his work by building on a critique of the paradigm of perfect rationality. His model of procedural rationality takes into account the cognitive limits of individuals and Simon describes how the alternatives are formulated, compared, and assessed, while learning plays a key role in improving the cognitive performance of individuals. Cognitive activities rely on the use of symbols, and symbolic structures are images of reality; they are more or less accurate transcriptions of the outside world. These complex structures are decomposable and the mental activity of problem-solving is made up of a set of elementary information-handling processes. Simon’s view develops out of cognitivism (Piaget 1969), in which perception works to connect the individual to his/her environment. The environment is the ‘outside’ of the individual and it is the source of new perceptive stimuli; individual cognition, by means of a process of assimilation, accommodation and feedback, works for the combination of the old and the new mental structures. The study of human cognition offers social sciences hints for moving

284

Margherita Turvani

attention from individual behaviour to the analysis of relations connecting individuals. Collective mental representations and inter-subjective interpretations ask for an explanation: connectionism, studying the mind of individuals using the connections between neurons as the unit of analysis and attributing to the plasticity of the system of connections the emergence of cognition, offers a useful theoretical support. The study of the cognitive structures (frames, mental models) that work as perceptive filters and link the individual to his environment, become the clue because these structures limit and select the mental activities that the individual performs. Cognitive structures evolve because when they are in action they can no longer be the subject of cognitive activity and knowing consists of creating new, more complex structures; the system of relations among individuals shows similar dynamics (Hutchins 1995). Social interaction is a clue in understanding the development of interpretative patterns and common frames used by individuals and it gives rise to shared forms of behaviour in a social environment and in problemsolving activities (Denzau and North 1994). Despite the highly idiosyncratic subjective nature of cognitive processes, the development of intense communication, favoured by proximity (not necessarily only spatial but also cultural) leads to commonality in developing cognition, which takes on the character of dependency on the context in ‘local and idiosyncratic’ forms. As Witt (1998) observes with reference to the studies carried out by Bandura (1986), interestingly these mechanisms produce shared behavioural models: The significance of such models lies in their vicarious nature, i.e. the fact that their rewarding or non rewarding consequences can be grasped by observation rather than own experimentation. Within intensely communicating groups, learning by observation then means that the group members tend to focus on much the same model. Consequently, their individual learning processes produce correlated results which, via the prevailing social models, can intuitively be grasped by newcomers. (Witt 1998: 5) These mechanisms are important because they offer a vicarious method to appraise success and failures of deviant behaviour – that is, behaviour diverging from that commonly established in specific contexts. In such a way, individuals may receive information to evaluate their own behaviour in different possible alternative situations, previously not under consideration. A reframing of choices may take place, allowing smooth transition from older social models to more innovative ones (Witt 1998: 6). Complex institutional and organizational systems (the outcome of more or less intentional efforts) help individuals and groups in their cognitive activity, supplying aids for decision-making and learning. In this vein,

Cognitive economies and the ‘district’ 285 North (1990) connects the organization of the economic process with the individual cognitive activity; the action of competing agents is the driving force in economics but action depends on the agents’ perception of the possible outcomes – yet perception critically depends on the available information and the way such information is processed. Given the cognitive abilities of the agents, information is often incomplete and is elaborated on the basis of past experience, creating the condition for a ‘path-dependent’ evolution (Arthur 1994). Institutional context and inherited rules therefore affect individual cognition because individuals perceive the environment according to the mental structures they use to interpret the world. These structures are rooted in the past and in culture and, more importantly, they are developed in specific contexts, contributing to limit and to model the possibilities of perception and choices for agents (North 1990). In this way differentiation and heterogeneity are continuously recreated, preserving the use of different mental structures and models leading to the elaboration of different interpretations out of a similar set of information. Subjective interpretation plays a major role in the agents’ ability to frame a problem and to attribute meaning, and therefore it affects decision-making. Just as institutions and organizations limit possible alternatives, so, too, mental structures limit perception and the set of decisions at the individual and at the social level (North 1994).

The firm, its context, and the district The context is made up of external economies Theories of the firm offer different perspectives to look ‘inside’ the black box; the ‘inside’ is seen as a nexus of contracts when organizations serve for agent’s information problems while the ‘inside’ is a collection of capabilities when organizations serve for agent’s cognitive activities. In this case, cognition takes place within the boundaries of a firm (i.e., ‘in context’), and firms ‘are social communities specialised in the speed and efficiency in the creation and transfer of knowledge’ (Kogut and Zander 1996: 503). Outside the boundaries of the firm, of course, there is the world of which the firm is a constituent part, made up of markets, firms, nature, society, etc. Each firm relates with its ‘outside’ in many different ways, and it is embedded in a thick web made of material and intangible flows. Most typically, economists describe what may accrue from the relation existing between what is inside the firm and what is outside in terms of external economies.5 The Marshallian industrial district comes out of this relation between the firm, as a unit, and what is its outside. Bellandi (1989) discusses in detail the notion of the industrial district as this may be understood when reading Marshall:

286

Margherita Turvani Marshall does not strictly define the industrial district in relation to small or medium-sized firms . . . However it is these firms which most interest him when discussing industrial district. He points to ‘those very important external economies which can often be secured by the concentration of many small businesses of a similar character in particular localities: or, as is commonly said, by the location of industry.’ They are an agglomeration of local, external economies, that is the economies of production and transaction costs which a firm enjoys and which derive from the size of the system to which it belongs. More precisely, in Marshall’s thought, the economies come from the working of the firm within the thick local texture of interdependence which bind together the small firms and local population. (Bellandi 1989: 38)

As noted by Bellandi, external economies – that is, economies dependent on the general development of the industry – become even more relevant when activated in and associated with co-localization of firms, giving rise to economies of agglomeration which develop in transaction dealing, skill formation, and innovation processes. Bellandi, discussing Marshall, concludes, ‘Agglomeration effect results from the fact that standardization [is] particular to local producers . . . and depend[s] on the possibility of easy adjustment or setting up of networks of transactions among firms’ (Bellandi 1989: 42). Proximity reinforces agglomeration by the accumulation of skills because, according to Marshall, if many men are employed in the same locality and in the same tasks, they will educate each other. They develop a sense of responsibility and carefulness, too, and ‘local industry’ becomes the ‘property of all’. Furthermore, agglomeration produces advantages in terms of the innovation capacity of firms – not only because good ideas may easily circulate but also because there is a positive attitude towards improvements and innovations: Industry’s secrets are ceasing to be secrets: they are, as it were, in the air and children are unwittingly learning many of them. Work well done is rightly appreciated, inventions and improvements in machinery, in process and the general organization of the business have their merits promptly discussed: if one man starts a new idea, it is taken by others and combined with suggestions of their own; and thus it becomes the source of further ideas. (Marshall, 1890: 225; cited in Bellandi 1989: 145) All these factors refer to the existence of external economies which are set in motion by co-operation and competition, and which sustain cooperation and competition within the industrial district (Jong 1994). The

Cognitive economies and the ‘district’ 287 co-existence of co-operation and competition among firms, localized in proximity, favours individual success, and the origin of the district may be found more often in close proximity of firms. It provides ‘opportunities for entrepreneurs to specialize and for the district as a whole to secure economies of scale denied to isolated individual firms because of internal restriction on growth. They can therefore afford to stay small and concentrate their initiative and inventiveness on what they do best’ (Jong 1994: 264). The existence and the nature of the district rest on external economies which fill up the outside of the firm; the nature of the firm does not matter in explaining why firms lump together because the firm is more a black box than a living organization. The context is ‘the institutional setting’ New institutional economics develops a theoretical framework to explain the existence of various discrete organizational forms (Williamson 1996). It focuses on transaction costs and its determinants at the micro and macro level, and it contends that the rules, laws and culture of beliefs and values all affect the cost of transacting and shape the possible governance structures. This perspective may be used to develop an explanation of the existence of the industrial district, one of the many forms of organization for economic transactions. The existence of positive transaction costs threaten not only the economic but also the social and political systems, since when there are transaction costs, individuals can exploit them to the detriment of others. Of course, the state plays a leading role in determining transaction costs: supplying a more or less stable system of property rights and a more or less adequate system to make them effective through laws and law enforcement, the state offers the rules of the game for economy and for society (North 1994).6 Yet, the rules to which we conform are much more articulated: conventions, cultures, beliefs and all sorts of informal rules prevail in determining everyday social and economic life. Within these systems of rules, transactions get organized; markets, hierarchies, and hybrid forms are developed. Having identified how transactions differ, and the features distinguishing alternative modes of governing transactions, transactions take place within the most appropriate governance structure, in order to minimize transaction costs. From the methodological point of view, institutional economics develops its analysis at two different levels, that of institutional environment and that of organization – or institutional arrangements to use Williamson’s words (Williamson 1996). Within this framework the ‘industrial district’, as much as networks of firms, are put into the catalogue of hybrid forms of organization which develop, thanks to the low level of transaction costs, within a favourable institutional environment. Much effort needs to be poured into the analysis of the ‘environment’, made

288

Margherita Turvani

up of both formal and informal institutions which carry the burden of minimizing the transaction costs within the complexity of transactions taking place in the industrial district. A favourable ‘institutional environment’ is required and should be promoted to recreate the frame supporting the replication of successful experiences of the ‘industrial district’. Given the proper institutional environment the industrial district will emerge as the most-suited organizational form for transactions completion. The approach is static in its essence, the feedback from the economic transactions to the social and institutional relations is not taken into account (Turvani 1998). Again the virtuous cycle connecting the firm with its outside, and vice versa (for example, the flow of Marshallian externalities), is overlooked. The context is ‘the district’ The specific feature of the most acknowledged theory of the nature of the district is the embeddedness of the firm; that is, the firm and its outside are subsumed in the ‘district’ and melt into an organic whole. The geographical aspects of agglomeration loses relevance with respect to the territorial dimensions of development (Gambarotto 2002). Since the end of the 1970s and the early 1980s the credibility of this approach has grown, building on the work on ‘third Italy’, and it has been developed along different streams of research on the role of small and medium-sized enterprises in the economic development process (Bagnasco 1977; Becattini 1979; Brusco 1982; Fuà and Zacchia 1983).7 This vein of studies assumes the Marshallian framework describing how the progressive division of labour is accompanied by complementary forms of integration, giving rise to the peculiar features of dense externalities within the ‘district’. Moreover, in the same Marshallian vein, the socioeconomic relations that characterized the territory and the district are extremely relevant (Becattini 2001). These specific conditions shape transactions, lower transaction costs and keep the average dimension of firms small. A great deal of attention is therefore devoted to the comprehension of all those mechanisms that sustain exchange, as community life, trust and general ‘atmosphere’ (Dei Ottati 1995). A distinctive feature of this research approach on ‘district’ is the combination of the ‘economic’ elements, as developed by Marshall, with a strong twist in favour of sociological analysis. It results in a very high level of embeddedness of firms within their environment, which becomes the centre of the analysis. Descriptive studies offer insights for policy recommendations, even though each reality ‘uniqueness’ and its organic, historical process hamper the possibility of replication. In the Marshallian vein, the organic evolution of the district rests on the processes which mobilize and create specific resources at the local level, while the ‘post-fordist’ vein emphasizes the role of the technological

Cognitive economies and the ‘district’ 289 progress in creating the conditions for the permanence of district-like forms of organization, which appear as prototypes of a flexible production system.8 The technological and the socio-economic features are integrated in the study of the milieu innovateur (Camagni 1991) and the innovation and learning processes that take place within a territory are under scrutiny. Networks and interrelations between different entities (individual, firms, institutions) constitute the substance of the milieu, made up of different and overlapping ‘spaces’ for communication. Relational aspects are core features of the geographical approach (Storper 1997) which weld the socio-economic and the technological dimensions. The guiding metaphor is the economy of relations, the economic process as conversation and coordination, the subject of the process not as factors but as reflexive human actors, both individual and collective, and the nature of economic accumulation as not only material assets, but as relational assets. Regional economics in particular, and integrated territorial economies in general, will be redefined here as a stocks of relational assets. (Storper 1997: 28; cited in Gambarotto 2002) Emphasis on the relational aspects in all these approaches leads to a common trait which unifies each of them, in various forms and to a degree: embeddedness of firms within their environment. A major consequence is that each firm loses its character of uniqueness: that of being a repository and creator of unique knowledge. Departing from the archetypal Marshallian district based on the ‘faceless’ concept of external economy, the necessity to give to territory a ‘face’ results in the idea of ‘thickness’ of social and economic relations. A colourful description of what is going on in a district is obtained, but indeed ‘thickness and density’ come to substitute ‘agglomeration’ (Malmberg and Maskell 2002). A second major consequence of blurring the boundaries between the firm and its environment is the definition of a (seemingly) powerful concept: social capital. This notion is able to condense a variety of explanations, enabling the researcher mentally to ‘locate’ those factors that ‘make the difference’, to explain development in a specific location. It is indeed a too parsimonious explanation, used to fill up the gap between explanation offered by theory (especially economic theory) and the real world.9

Knowing in firms Processes of learning and cognition are core activities in firms and organizations. Within each firm, individual cognitive processes are developed by interacting with other minds; the features of such collective production of knowledge have been previously described. The way in which individuals

290

Margherita Turvani

learn and know in specific contexts by interacting – for example, working in a social community like the firm – has implications trespassing on the boundaries of each firm. Understanding how these processes take place within each firm may be used as a guide in explaining why firms learn and develop knowledge by lumping together. A better understanding of what is, perhaps, very incompletely understood among economists about the individual mind functioning in the context of the firm suggests a framing of the understanding of ‘the nature of the district’ by utilizing what we learnt about the ‘nature’ of the firm. We know that firms are structures which govern the production of knowledge and its utilization in novel combinations in production processes; we know that knowledge generation is an unfolding open-ended process going on by means of subsequent structuring; we know that the outcome of such a process assumes specific and local features being developed in a ‘context’. This understanding is usefully transferable and can be applied to the analysis of the functioning of the district. The interactive contextual nature of knowing gives rise to knowledge that is renewed but is never purely self-propagating on a larger scale. The growth of knowledge, like the growth of an organism, does not follow a logic of ‘more of the same’, since its form and contents are transformed (Winter 1982; Langlois and Foss 1999). Knowledge proceeds along a growth pattern with the generations of new forms and structures by which individuals are able to select, interpret and organize the full range of stimuli coming from the environment; only within this framework is information meaningful – that is, it is information and not simply raw data. Sociality and the existence of a community of practice is necessary for cognition, as recent findings in cognitive studies has shown, and there is a process of selective perception and retention progressively building up mental structures. These are used for knowing and they evolve by filtering and selecting signals following the path drawn by past experiences (Hayek 1952; Lave and Wenger 1991; Rizzello and Turvani 2000). These micro processes happen in ‘firms’ as much as in ‘districts’ because they take place in the individual minds and shape organizational behaviour in each and any ‘context’. Theories of the firm, such as the evolutionary, the resources-based, and the knowledge-based, describe the working of firms in terms of individual cognitive activities, in terms of routines, or of core capabilities. These structures sustain persistent patterns of behaviour and evolve through time according to different and past experiences of firms. Firms then remember and know by doing and their knowledge preserves a character of tacitness for the people holding it (Penrose [1959] 1995; Polyani 1962; Nelson and Winter 1982; Nonaka and Takeuchi 1995; Teece et al. 2000). The knowledge that organizations handle and continuously create resides in the complex of routines, and in the pool or capabilities making

Cognitive economies and the ‘district’ 291 up the firm. It is through this lens that organizations approach messages coming from within the organizations themselves and also from outside. Routines serve as much as a repertoire of capabilities – that is, they describe what the firm knows and how it knows it; but they are also the frame by which the outside is understood. They can be interpreted as a firm’s cognitive apparatus and, at any time, they are the specific organization’s cognitive model or frame (or map) (Langlois 1983, 1995). Organizational routines not only filter what may be ‘seen’ in the environment but also shape the firm’s absorptive capacity (Cohen and Levinthal 1993), which is the ability to understand and take advantage of ideas (or knowledge) developed elsewhere. Perceptual abilities of firms play a major role, not only in guiding a firm’s strategy but in determining its success in the initiatives undertaken. Firms are endowed with a system of routines that shape their perceptual abilities in two broad areas: (1) the ability to see ways of gaining and improving competencies and (2) the ability to recognize opportunities for applying those competencies in a way that generates value. At the risk of oversimplifying a bit, let’s call those areas operational perception and market perception. (Langlois 1995: 8) Perceiving is not a trivial activity, given that ambiguity and uncertainty and many forms of misperception may occur. It may result from thinking that an opportunity exists when, by ex post consideration, it does not; or it may result in lacking the perception to recognize an opportunity which, ex post, will turn out to be profitable.10 Langlois (1995) applies this line of reasoning to the effects of misperceiving opportunity within firms operating in the development of the computer industry, and concludes: What determines the ability of an organization to perceive opportunities? It is my contention that perceptual ability is a matter of the ‘fit’ between the environment and the organization as a cognitive apparatus . . . If an organization is to be able to classify information in the right (the most useful) categories, it must posses a structure of categories that is somehow ‘like’ those of the outside world, or at least that small piece of the outside world relevant to it. (Langlois 1995: 10) This conclusion may be both trivial and profoundly complex, yet it fits well with our description of how organizations develop knowledge relying on the mechanisms of individual cognition; it also fits well with evidence about the development of knowledge along scientific paradigms and, furthermore, with the evidence and research on localized technological change (Antonelli 1999; Nelson 1980).

292

Margherita Turvani

Collective knowing Firms do not know in a vacuum; they know in ‘context’, and this is recognized in the economic literature on innovation: technological knowledge cumulates systemically, it is created using pre-existing knowledge, and it is highly indivisible. Producing technological knowledge is a collective process in which each agent has unique yet complementary information (Antonelli 1999; Richardson 1972). This perspective describes knowledge as a collective good, given its features of radical indivisibility and input complementarity in its production: When these conditions hold, the productivity of a given technological knowledge will be much higher if everybody would be willing to put in a common pool all its knowledge that is complementary to many others . . . We may say that such generation of knowledge is an archetype of a network externality effect, where resources productivity is as much higher as many more agents are participating in the net. (Antonelli 2001: 229) The possibility for each agent to take advantage in producing such a collective good, knowledge, is strongly related to the existence of a reliable net of communications. Information flows and the features of the network need to be analysed from this perspective, referring to communication theory, while individual positional advantages with respect to the flows of information influence ability and performances (Antonelli 2001). This framework offers a deeper understanding of what constitutes the ‘atmosphere’ of the district; many networks for communication are overlapping in a specific territorial environment, and a picture of what is a ‘thick’ set of transactions is neatly emerging, both at the economic and the social levels. Networks enabling communication include local labour markets, user–producer relations, technological outsourcing, subcontracting and relations between small and leading firms, co-operative projects between firms, relations with research institutions of various kinds, financial markets, and so on. The overall image is that of an innovative milieu, but firms’ knowledge is the engine of growth; if overlapping networks of this kind exist, positive feedback will result; development at local level will appear as a self-reinforcing virtuous circle. The ‘district’ is the outcome of the previously described processes of the collective production of knowledge. Each firm enjoys a wide range of externalities in the form of economies in technological knowledge production: the overall costs that each firm needs to bear to innovate, both in terms of production and transaction cost, are lower when efficient communication networks exist between the firm and its environment (mainly made up of other firms). To put it another way, with any given level of resources, such economies allow each

Cognitive economies and the ‘district’ 293 firm to reach a higher level of innovative performances. These conditions are very likely to happen in ‘proximity’ and because of ‘proximity’ (Antonelli 2001). Firms know in ‘context’ and, as we saw, the collective dimension of knowing is a source of economies for firms. Furthermore, firms know and remember by doing (Nelson and Winter 1982), and this happens in the ‘context’ provided by the specific set of overlapping networks which feed firms with information. Each of them elaborates on a similar set of information and, according to specific capabilities or cognitive maps, is able to produce new forms of knowledge. Communication among firms is crucial, and the variety of networks overlapping within the district contribute to ‘fine tuning’; yet communication should not be seen as a way to reallocate information and distribute signals. Communication involves deeper processes, such as recognizing differences in cognitive structures, transfer of knowledge from one context to the other, comparison and convergence of cognitive models, and so on. Social learning will occur (Bandura 1986), welding together the interdependence between subjective perceptions and experiences at the individual level with the process of learning and knowledge growth at the macro level. Knowledge is dispersed, local and specific and competition spreads it and helps the formation of ‘opinions’ in markets (Hayek 1946). Social learning then develops on individual cognition and its subjective and idiosyncratic features: this specificity triggers novelty. Experience and insight preserve uniqueness, and each learns by experiencing and getting insight. These features of the knowledge-creation process were analysed a long time ago by Adam Smith, as Loasby noted (1999). Adam Smith observed that any process of division of labour is accompanied by a process of division of knowledge: ‘Not only does the division of labor encourage invention and other improvement; different kinds of division encourage different kinds of invention and improvements’ (Smith [1795] 1980: 20–21). Smith identified three categories: The first is produced by workmen, as their ‘attention comes naturally to be directed towards some one very simple object’; the second by those who have developed the making of machinery into a ‘peculiar trade’; and the third by those ‘who are called philosophers or men of speculation, whose trade it is, not to do anything, but to observe everything; and who upon that account are often capable of combining together the powers of the most distant and dissimilar objects’. (Loasby 1999: 71) Smith is describing a process in which the division of knowledge and the division of labour spur each other on, and new branches of knowing and doing develop out of this process; while developing our knowledge through experience, by specializing in peculiar trade and focusing our

294

Margherita Turvani

attention, we also open our minds to insight, putting together the most distant and dissimilar objects, thus giving rise to novelty. Selective perception, working both as a filter and as a focusing device, shapes the outcome of mental processes. Accordingly, knowledge can’t be directly deduced from a closed set of information; rather, it comes out of cognitive activity pursued by individuals who elaborate information in an increasingly idiosyncratic and local way. The very same processes take place in firms: knowledge supporting decision-making in the firm is characterized by the strategic assessment of the available information, by interpreting it in relation to the set of relations that the firm is able to ‘see’ or ‘perceive’ between its own behaviour, the behaviour of other firms, and of all the other agents involved. The knowledge the firm requires is thus the outcome of a process of elaborating elementary information, organized around an interpretative model, connecting the firm’s objectives and their context. The calculation of the outcome of the decision-making process relying on a complete set of information is a fiction, and, as Knight (1921) pointed out, decisions in a firm are based on judgement, and this judgement is a constituent element in knowledge-production processes. A vast scope for judgement and entrepreneurial activity, using Knight’s words, may be envisioned, even in everyday activities of the firm, to seize opportunities in the market and in the operational field.

Entrepreneurial activity, the firm, and the nature of the district Entrepreneurial activity, the ability to seize opportunity by putting together ‘the most distant and dissimilar objects’ to use Smith’s words, is the ability to drive profit out of novelty – that is, innovating in product, processes and organization. Relying on our description of individual cognition and on Adam Smith’s hints, it is possible to assume Langlois’s conclusion with confidence: ‘when operational and/or market environment is highly focused, those organizations will do best whose cognitive apparatus allows them to perceive the finer details and to solve the smaller but important puzzles’ of everyday activities of problem-solving (Langlois 1995: 11). There are good reasons for firms to lump together creating districts, and we have described some of them; but there must be a good reason for each specific firm to join the district. This reason is to be found in the entrepreneurial judgement of a good opportunity for profit (Kirzner 1979). This trivial observation takes us back to the problem we have started with: why do we see ‘islands of conscious power’? Richardson (1972) explained that transaction cost alone cannot explain the emergence of firms in a ‘world of transactions’, but that a production costs and transaction costs comparison needs to be taken into account. In his view, firms

Cognitive economies and the ‘district’ 295 always exist in a dense network of co-operation and affiliation by which they are interrelated: co-operation is just one way in which activities can be co-ordinated, the other being direction and market transactions (Richardson 1972, 1998). He develops the analysis of capabilities; that is, the appropriate knowledge, experience and skill that sustain efficacious performance of firms. Each firm will tend to specialize in a set of activities requiring similar capabilities because the unified control of very dissimilar activities may hinder productivity, unless firms grow to reach the appropriate level of economies of scale. Economic literature on the district has shown that these conditions do not apply when firms are in close proximity and each new firm is able to take advantage of the tight productive connections in the local market. In this case, as Loasby observes, vertical integration is not necessary to align complementary activities and ‘by avoiding it firms at each stage of the production process are better able to preserve the distinctive outlook of their separate trades, which, as Adam Smith noted, enables them to perceive opportunities that would be overlooked within a single business’ (Loasby 1999: 74). Each kind of knowledge, then, may be deepened; economies may be obtained in innovative activities even if there is no formalized research going on within the firm, and rents may be appropriated more easily. It is important to note that the possibility to rely on a very high and progressive division of labour offers some other cognitive economies: new knowledge in any of the sub-fields and applications has a positive consequence for the firm producing it, but it also has a value for many other firms that, operating in similar activities, may rely on vicarious learning (Bandura 1986) to avoid expensive blind alleys and to concentrate on profitable opportunities. Absorptive capacity is also enhanced when entrepreneurial activity is facilitated by the high division of labour/knowledge, both because the entrepreneur’s ‘attention comes naturally to be directed towards some one very simple object’ and because entrepreneurs operate in a ‘peculiar trade’. But entrepreneurial activity may imply radical novelty ‘combining together the powers of the most distant and dissimilar objects’, opening new lines of business in the district and progressively diversifying its original vocation – in this case, the operating of forms of vicarious learning may spread and foster diversification (Witt 1998). All these kinds of activities implying entrepreneurial judgement may be found in any firm as much as in any district. Judgement and uncertainty absorption performed by the entrepreneur allows the formation of new firms, because by centralizing the assumption of responsibility for uncertain outcomes it is possible for firms to function as administrative and planning units (Penrose [1959] 1995; Richardson 1998; Turvani 1995). Each firm judges opportunities to make profit, and the perception of what is a good opportunity is, of course, developed in ‘context’. Each firm seeks for sources of economies in sizing technological and market opportunities,

296

Margherita Turvani

and these are exploited by entrepreneurs whose cognitive models and judgement capabilities are shaped and focused according to the context in which they are developed (Lorenzen and Foss 2002). Entrepreneurial action is the lifeblood of the industrial district experience, and much of the entrepreneurial action is put forward by individual entrepreneurs: if misperceptions of opportunities in technology and market fields depend on the fit between the organization and its environment, then in the context of ‘district’ entrepreneurial action implies a lower level of uncertainty absorption or, to put it another way, entrepreneurial initiatives are favoured even if, on average, a lower level of profit may accrue to them. Furthermore, the prevalence of small-size enterprises, which is correlated with the high level of the birth rate of firms (entrepreneurial initiatives), shapes the quality of human resources: not only is labour mobility very high but the knowledge that human resources holds is both highly tacit, idiosyncratic and complex, given the low level of division of labour within each firm. The district becomes the place in which subjectivity of individual cognition is welded with the construction of shared mental models: the way of knowing within a firm feeds in and is nurtured by knowing taking place among other firms.

Conclusion The attempt to describe some features of the microfoundations of knowledge dynamics within the firm suggested some implications for understanding the nature of specific forms of organization of firms and territories. If Nonaka et al.’s view (2000: 8) that ‘to understand the true nature of the knowledge creating firm, viewing a firm as a black box, a set of transactions or a collection of resources is not enough’ is accepted, the necessity to understand the features of the creation of knowledge, and the forms in which it takes place within the firm, is not questionable. The creation and utilization of knowledge is a core source of a firm’s capability to search and exploit profit opportunities and indeed, at any given time, the knowledge that firms hold provides the future opportunities to innovate. Yet firms differ in knowledge management and developments and these differences persist through time and build up a firm’s specific capabilities. An understanding of the mechanisms of cognition is useful to describe how new opportunities for action emerge with the generation of new knowledge and how persistence and change co-exist in the knowledge-creation process in the individual mind and in the organization. The reason for this must be sought in the social dimension of individuals’ cognitive activity, which simultaneously opens up opportunities for change and innovation but also preserves a degree of inertia. The mechanisms underlying the knowledge-creation process work in interactive contexts, suggesting a picture of the individual as ‘a dynamic being and the firm as a

Cognitive economies and the ‘district’ 297 dynamic entity that actively interacts with others and the environment’ (Nonaka et al. 2000: 2). In this picture a social dimension of knowledgecreation appears; it grows on communication, imitation and comparison, and it encourages both the development of interpretative patterns and common frames and the emergence of novelty, fostered by the day-by-day knowledge-creating process at the individual and at the organizational level. Adam Smith’s insights help to reframe this picture at the district scale. The progressive division of labour in firms and among firms goes hand in hand with the redefinition and the renewal of the division of knowledge: novelty is the outcome and it opens up new opportunities for the use of entrepreneurial judgement, together with the permanence of shared forms of behaviour in the social environment and in problem-solving activities.

Notes 1 This definition has been developed by the philosopher Dretske (1981). 2 In particular, Alchian and Demsetz’s work of team production (1972), the principal-agent theory by Jensen and Meckling (1976) and the pioneering work of Coase (1937) on which Williamson (1985) develops the economics of transaction costs. In these views hierarchy solves information problems relying on more or less complex contractual arrangements. 3 There is a vast literature on the subject. For a survey, see Foss (1993) and Montgomery (1995). For an examination of the points of contact between the strategic approach and the approach based on resources, see Mahoney and Pandian (1992) and Prahalad and Hamel (1990). 4 This theme was developed by Rizzello and Turvani (2000) in a paper discussing the relevance of the cognitive sciences for institutional economics. 5 Let’s recall the definition of internal and external economies given by Marshall: economies ‘dependent on the resources of the industrial houses of business engaged in it, on their organizational and the efficiency of the management’ are called internal economies; while those ‘dependent on the general development of the industry’ are called external economies (Bellandi 1989: 38). 6 This definition has been given by Nobel Prize winner Douglas North: ‘Institutions are the humanly devised constraints that structure human interaction. They are made up of formal constraints (rules, laws, constitutions), informal constraints (norms of behavior, conventions, and self imposed codes of conduct) and their enforcement characteristics. Together they define the incentive structure of societies and specifically economies’ (North 1994: 360). 7 To offer a complete picture of the literature on the industrial district and on the various forms of localized productive organizations and systems is outside the scope of this work. For an appraisal of such literature see, for example, Maillat (1998) and Gambarotto (2002). 8 This line of research was first introduced in Piore and Sabel’s analysis (1983) and then widely applied and developed (for example in Italy by Rullani 1993). 9 Social capital, made up of fiduciary relations and duties that foster each other, is a novel ‘invisible hand’, that is a substitute for the market in explaining how co-operation and competition trigger development. The working of civil society, after all, becomes the engine of growth: ‘to focus on

298

Margherita Turvani

social capital means, then, to recognize that the propensity and the capabilities to cooperate that members of a society express, may deeply affect the character of economic and political development that the community may achieve’ (Mutti 1998: 12). 10 Langlois treats these possible misperceptions as errors of type one and errors of type two, according to statistical testing of hypothesis.

References Alchian, A. and Demsetz, H. (1972) ‘Production, Information Costs and Economic Organization’, American Economic Review, 62: 777–795. Antonelli, C. (1999) The Microdynamics of Technological Change, London: Routledge. —— (2001) The Microeconomics of Technological Systems, Oxford: Oxford University Press. Argyris, C. and Schön, D. (1996) Organizational Learning II. Theory, Method, Practice, New York: Addison-Wesley. Arthur, B. (1994) Increasing Returns and Path Dependence in the Economy, Ann Arbor: Michigan University Press. Bagnasco, A. (1977) Le tre Italie, Bologna: Il Mulino. Bandura, A. (1986) Social Foundation of Thought and Action. A Social Cognitive Theory, Englewood Cliffs, N.J.: Prentice-Hall. Becattini, G. (1979) ‘Dal settore industriale al distretto industriale’, Rivista di Economia e Politica Industriale, 7(1): 7–21. —— (2001) ‘From the Industrial District to the Districtualization of Production Activity: Some Considerations, Firenze: mimeo. Bellandi, M. (1989) ‘The Industrial District in Marshall’, in E. Goodman and J. Bamford (eds), Small Firms and Industrial Districts in Italy, London: Routledge, 31–62. Brusco, S. (1982) ‘The Emilian Model: Productive Decentralization and Social Integration’, Cambridge Journal of Economics, 6(2): 167–184. Camagni, R. (ed.) (1991) Innovation Networks: Spatial Perspectives, London: Belhaven. Coase, R. (1937) ‘The Nature of the Firm’, Economica, 4: 386–405. —— (1991) ‘The Institutional Structure of Production’, Alfred Nobel Memorial Prize Lecture, Stockholm. Cohen, M. and Levinthal, D. (1993) ‘Absorptive Capacity: A New Perspective on Learning and Innovation’, Administrative Science Quarterly, 35: 128–152. Cyert, R. and March, J. (1963) A Behavioral Theory of the Firm, Englewood Cliffs, N.J.: Prentice-Hall. Dei Ottati, G. (1995) Tra mercato e comunità: aspetti concettuali e ricerche empiriche sul distretto industriale, Milano: Franco Angeli. Denzau, A. and North, D. (1994) ‘Shared Mental Models: Ideologies and Institutions’, Kyklos, 47(1): 3–31. Dretske, F. (1981) Knowledge and the Flow of Information, Cambridge, Mass.: The MIT Press. Foss, N.J. (1993) ‘Theories of the Firm: Contractual and Competence Perspectives’, Journal of Evolutionary Economics, 3: 127–144. Fransman, M. (1994) ‘Information, Knowledge, Vision and Theories of the Firm’, Industrial and Corporate Change, 3: 713–757.

Cognitive economies and the ‘district’ 299 Fuà, G. and Zacchia, C. (1983) Industrializzazione senza fratture, Bologna: Il Mulino. Gambarotto, F. (2002) ‘Le relazioni economiche istituzionali nello spazio delle economie locali: una ipotesi teorica’, in R. Camagni, R. Fiorentini and M. Mistri (eds), Auto-organizzazione ed apprendimento strategico, Padova: Cedam. Hayek, F. (1945) ‘The Use of Knowledge in Society’, American Economic Review, 35: 519–530. —— (1946) ‘The Meaning of Competition’, in Individualism and Economic Order, London: Routledge. —— (1952) The Sensory Order. An Inquiry into the Foundations of Theoretical Psychology, London: Routledge & Kegan Paul. Hutchins, E. (1995) Cognition in the Wild, Cambridge, Mass.: The MIT Press. Jensen, M. and Meckling, J. (1976) ‘Theory of the Firm: Managerial Behaviour, Agency Cost, and Capital Structure’, Journal of Financial Economics, 3: 305–360. Jong, Il You (1994) ‘Competition and Co-operation: Toward Understanding Industrial Districts’, Review of Political Economy, 6(3): 259–278. Kirzner, I. (1979) Perception, Opportunity, Profit, Chicago: Chicago University Press. Knight, F.H. (1921) Risk, Uncertainty and Profit, Chicago: University of Chicago Press. Kogut, B. and Zander, U. (1996) ‘What Firms Do? Coordination, Identity, and Learning’, Organization Science, 7: 502–518. Langlois, R. (1983) ‘System Theory, Knowledge and the Social Science’, in F. Machlup and U. Mansfield (eds), The Study of Information, New York: Wiley. —— (1995) ‘Opportunities Seized and Missed in the History of the Computer Industry’, in G. Raghu, N. Praveen and Z. Shapira (eds), Technological Entrepreneurship: Oversights and Foresights, New York: Cambridge University Press. Langlois, R. and Foss, J. (1999) ‘Capabilities and Governance: The Rebirth of Production in the Theory of Economic Organization’, Kyklos, 52(2): 201–218. (Reprinted in R.N. Langlois, T. Fu-Lai Yu and P.L. Robertson (eds), Alternative Theories of the Firm, Vol. I, Cheltenham: Edward Elgar, 2002.) Lave, J. and Wenger, E. (1991) Situated Learning: Legitimate Peripheral Participation, Cambridge: Cambridge University Press. Loasby, B.J. (1991) Equilibrium and Evolution, Manchester: Manchester University Press. —— (1999) ‘Industrial District as Knowledge Communities’, in M. Bellet and C. L’Harmet (eds), Industry, Space and Competition, Cheltenham: Edward Elgar. Lorenzen, M. and Foss, N. (2002) ‘Cognitive Coordination, Institutions, and Clusters: An Exploratory Discussion’, in T. Brenner and D. Fornahl (eds), Cooperation, Networks and Institutions in Regional Innovation Systems, Cheltenham: Edward Elgar. Mahoney, J. and Pandian, R. (1992) ‘The Resource-based View within the Conversation of Strategic Management’, Strategic Management Journal, 13: 111–126. Maillat, D. (1998) ‘From the Industrial District to the Innovative Milieu: Contribution to an Analysis of Territorialized Productive Organisations’, Recherches Economiques de Louvain, 64(1): 111–129. Malmberg, A. and Maskell, P. (2002) ‘The Elusive Concept of Localization Economies – Towards a Knowledge-Based Theory of Spatial Clustering’, Environment and Planning, 34(3): 429–449.

300

Margherita Turvani

March, J.G. and Simon, H. (1958) Organizations, New York: Wiley. Marshall, A. (1890) Principles of Economics, London: Macmillan. Montgomery, C. (ed.) (1995) Resource-based and Evolutionary Theories of the Firm: Towards a Synthesis, London: Kluwer Academic Publishers. Mutti, A. (1998) Capitale sociale, Bologna: Il Mulino. Nelson, R. (1980) ‘Production Sets, Technological Knowledge and R&D’, American Economic Review, 70(2) (May): 62–67. Nelson R. and Winter, S.G. (1982) An Evolutionary Theory of Economic Change, Cambridge, Mass.: The Belknap Press of Harvard University Press. Nonaka, I. and Takeuchi, H. (1995) The Knowledge Creating Company, Oxford: Oxford University Press. Nonaka, I., Toyama, R. and Nagata, A. (2000) ‘A Firm as a Knowledge-creating Entity: A New Perspective on the Theory of the Firm’, Industrial and Corporate Change, 9(1): 1–21. North, D. (1990) Institutions, Institutional Change and Economic Performance, Cambridge: Cambridge University Press. —— (1994) ‘Economic Performance Through Time’, American Economic Review, 84(3): 359–368. Penrose, E. ([1959] 1995) The Theory of the Growth of the Firm, Oxford: Basil Blackwell. Piaget, J. (1969) The Mechanism of Perception, London: Routledge & Kegan Paul. Piore, M. and Sabel, C. (1983) The Second Industrial Divide, New York: Basic Books. Polanyi, M. (1962) Personal Knowledge, New York: Harper & Row. Prahalad, C. and Hamel, G. (1990) ‘The Core Competence of the Corporation’, Harvard Business Review, 68(3): 79–91. Richardson, G. (1972) ‘The Organization of Industry’, Economic Journal, 82: 883–896. Richardson, G.B. (1998) ‘Production, Planning and Prices’, Oxford, mimeo. Rizzello, S. and Turvani, M. (2000) ‘Institutions Meet Mind: The Way Out of an Impasse’, Constitutional Political Economy, 11(2): 165–180. Rullani, E. (1993) ‘Networks and Internationalization: Managing Complexity through Knowledge’, in L. Zan, S. Zambon and A. Pettigrew (eds), Perspectives on Strategic Change, Dordrecht: Kluwer Academic Publishers, 107–142. Senge, P. (1990) The Fifth Discipline: The Age and Practice of the Learning Organization, London: Century Business. Simon, H.A. (1986) ‘Theories of Bounded Rationality’, in C. McGuire and R. Radner (eds), Decision and Organization: A Volume in Honor of Jacob Marschak, Minneapolis: University of Minnesota Press. Smith, A. ([1795] 1980) Essays in Philosophical Subjects, W. Wightman (ed.), Oxford: Oxford University Press. Storper, M. (1997) The Regional World, New York: Guilford. Teece, D., Pierce, L. and Boerner, C. (2000) ‘Richard Nelson’s Dynamic Capabilities’, Paper presented to the Conference for Honoring Nelson, New York, October. Turvani, M. (1995) ‘The Core of the Firm: The Issue of the Employer–Employee Relationship’, in J. Groenewegen, Transaction Cost and Beyond, Boston: Kluwer Academic Publishers, 189–208. —— (1998) ‘Black Boxes, Grey Boxes: The Scope of Contracts in the Theory of the Firm’, in K. Nielsen and B. Johnson (eds), Evolution of Institutions, Organization and Technology, Cheltenham, UK and Lyme, US: Edward Elgar.

Cognitive economies and the ‘district’ 301 Williamson, O.E. (1985) The Economic Institutions of Capitalism, New York: Free Press. —— (1996) The Mechanisms of Governance, Oxford: Oxford University Press. Winter, S. (1982) ‘An Essay on the Theory of Production’, in S. Hymans (ed.), Economics and the World Around It, Ann Arbor: Michigan University Press. Witt, U. (1998) ‘Cognition, Entreprenurial Conceptions and the Nature of the Firm Reconsidered’, mimeo, Max Planck Institute, Jena.

13 Learning in economics Some Austrian insights1 Pierre Garrouste

Introduction Even it is not a new one, learning, as a field of research, is becoming an increasingly important topic in economics. Indeed, Arrow (1962) introduces the idea that individuals were learning by doing, before Rosenberg (1982) says that the process of learning by using was much more relevant. Thereafter learning by interacting, institutional learning, and organizational learning and learning by learning were introduced in order to explain how, in an evolutionary perspective, technologies, organizations and institutions were changing. However, learning in economics recently becomes an important topic per se. In other words, learning is not just a means to explain how innovations or some other economic phenomena are taking place and evolving but is much more a way to explain how individuals modify their behaviour. In fact, the story seems to begin when the possibility to conceive individuals as perfectly rational is becoming problematic.2 Indeed, it seems that one of the reasons why economists are dealing with learning is due to the fact that they consider that it is necessary to justify either why individuals are becoming rational3 or why they are not.4 As an example, learning was recently introduced in microeconomics5 and particularly in game theory.6 The idea is either to try to explain how individuals’ strategies converge toward a Nash equilibrium solution, or to look at the way individuals modify their behaviour due to the way the others behave even if this process is not converging to an optimal or unique solution. It is difficult not to link these last attempts to one of the main ‘Austrian’ assumptions: Individuals are not perfectly rational. They progressively learn how it is possible for them to modify their behaviour in order to improve their economic wealth. Two questions emerge from this: 1

First, why did Austrians wait until now to build up a theory of learning? Indeed, there are many references to learning processes in Austrian writings, but there is nothing that looks like a unified theory of

Learning in economics: Austrian insights 303

2

learning. It seems very strange, especially because of the main assumption recalled above. Second, what are the possible relations between learning models in economics, and particularly in game theory, and the Austrian tradition? In other words, is it possible to show that the developments of learning theories in economics are integrating some essential elements of the Austrian tradition, and, if it is the case, to what extent? This question is important because, depending on the answer given, it may be possible or not for the Austrian tradition to borrow some of the theoretical results from the ‘economics of learning’.

In fact, it may be possible to assume that if individuals are learning, they have, before learning, imperfect rational behaviour, even if it is also possible to assume that they are learning because they are rational. The significance of the introduction of learning in economics, then, is not so easy to appraise and the relationship between Austrian economics and learning theories is not easy to analyse. In the first section of this chapter I present a selective review of the literature on learning theories in economics. This review is a selective one because I am focusing on the way learning theories enable individuals’ rationality. In other words, I want to analyse the assumptions made by the different theories of learning in terms of the rationality the individuals have before and after the learning process, as well as how and what individuals are learning. In the second section I show that even if learning theories sometimes take into account the fact that individuals are increasing their knowledge, those theories are often either assimilating knowledge to information or considering that knowledge is a structure of information. I also appraise the capability for learning theories to assimilate the fact that individuals are different from the point of view of the ways they use the information they receive and its results in terms of coordination. Finally, I discuss the possibility to consider learning theories as ‘real’ Austrian contributions, concluding in the third section with the presentation of a research agenda.

Learning in economics: a selective review Learning is introduced in economics in very different ways. It is possible, however, to typify the models of learning depending on (1) the kind of conception of rationality they assume, (2) how individuals are learning, and (3) what individuals learn. Learning and rationality Some contributions tell us that individuals are learning, but because they are unable to say why and how they modify and possibly improve the

304

Pierre Garrouste

results of their behaviour it is uncertain that individuals are really learning. Indeed, many scholars dealing with learning in evolutionary economics are assuming that individuals, as well as groups, are learning but they don’t explicitly explain the way individuals are doing so.7 At the opposite in the eductive process (Binmore 1987) individuals have perfect knowledge of the structure of the game they are playing. Because they have perfect instrumental as well as cognitive rationality (Walliser 1998), individuals are learning in a very specific sense. They are only modifying their behaviour due to the information they gather. This kind of zero learning level is then paradoxically close to the one assumed by evolutionary game theory (Weibull 1995), even if some recent works introduce the idea of learning in ‘stochastic games’ (Young 1993, 1998b). In fact, in economics, the individual is often considered as a kind of automata responding to stimuli coming from his/her environment. However, in between those two extreme conceptions – lack of rationality and perfect rationality – there is, in fact, a continuum of conceptions of ‘bounded rationality’. When Walliser (1998) analyses the ways rationality and learning were introduced and are developed in game theory, he shows that between the two extreme conceptions listed above (lack of rationality and perfect eductive one) it is possible to appraise the conceptions game theory holds in terms of the way they deal with the instrumental and cognitive rationality as well. According to Walliser, cognitive rationality ‘deals with consistency between available information and constructed belief, in order to form expectations about the relevant environment’, whereas instrumental rationality ‘deals with consistency between given opportunities and fixed preferences, in order to derive intended strategies from previous expectations’ (Walliser 1998: 68). Those two fixed points are interesting because they are not too far from the usual distinction made in psychology between behavioural versus cognitive approaches of learning. Perfect rationality and learning It is important to take into account the fact that some economists try to assure that perfect rationality, as a goal, is compatible with learning. In this perspective, the problem is to describe how individuals are becoming rational. The fictitious play learning process can be considered as a possibility to justify that, after a kind of ‘pre-play process’, individuals are adjusting their behaviour and are then progressively playing their best responses to the others’ best responses. In this way, it is possible to justify the existence of a process towards some Nash equilibrium. Zwick et al. (1999) give a similar interpretation when they propose an alternative interpretation of equilibrium to the cognitive one. ‘If people face a similar situation numerous times they learn to adopt better strategies by a trial-

Learning in economics: Austrian insights 305 and-error process. A Nash equilibrium is considered a stationary point in this process’ (Zwick et al. 1999: 14). Learning in economics can be linked with the fact that individuals are perfectly rational in the sense that they optimally change their behaviour because information coming from outside justifies such a change. They modify their strategy either because the ‘objective’ state of the world changes or because the others’ strategies do. Technically they revise or update their beliefs depending on the static or evolving character of the world. The main example of this kind of learning process is the famous Bayesian one where individuals revise their beliefs due to the signals they receive. Individuals can then be considered as automata that react mechanically to their environment. If, however, individuals are bounded in terms of rationality, it is possible to assume that they try to modify or improve their behaviour. In other words they can learn. Learning and bounded rationality Bounded rationality is an old notion due to Simon, even though the idea that individuals are limited in their cognitive capabilities had been introduced before.8 However, this notion is now recognized as an essential one in economics (Conlisk 1996; Rubinstein 1998). The problem is that there is no unique and definitive notion of bounded rationality, and also that there is not a unique model of bounded rationality (Rubinstein 1998). Indeed, economists have traditionally avoided explaining behavior as less than rational for fear of developing many fragmented theories of mistakes. Part of the attraction of highly rational models is the idea that there may be many ways to be less than rational, but only one way (or in light of the equilibrium refinement literature perhaps only a few ways) of being highly rational. (Erev and Roth 1998: 848) If different conceptions of bounded rationality exist, different ways to link them to the learning process also take place. In fact, even if it is difficult to organize the ways economists deal with bounded or soft rationality, I think that two main ways of analysing the relations between rationality and learning can be considered. The first one, exemplified by Langlois (1997) and Rubinstein (1998), is to start from the main assumptions of the ‘neoclassical notion of rationality’ or the notion of ‘rational man’ and to explain how it is possible to amend them in order to propose a more useful notion of rationality. The second one is to propose different definitions of the notion of perfect rationality and to look at the ways they can be modified (Walliser 1998).9

306

Pierre Garrouste

Those two main ways of dealing with bounded rationality imply different conceptions of learning. In the first case the models of learning are characterized by the type of assumptions one can modify. As an example, according to Langlois (1997), the neoclassical model of rationality combines the following four elements: • • • •

Self-interest Omniscience (‘complete information’) Conscious deliberation (or an ‘as if’ equivalent) The representative agent (Langlois 1997: 3)

Depending on the element(s) (supposing the hypothesis is accepted that those elements are independent) you modify, you define a determined kind of rationality and then implicitly of the learning process: 1

2

3

4

If individuals are altruists they are supposed to gather information not only from their environment (that is, the state of the world and others’ strategies) but also from the others’ satisfaction levels or utility functions. If individuals have incomplete information they learn by gathering information and then revise or update their beliefs and modify their strategies. If individuals are not consciously deliberating, they can behave as rule-followers; then the learning process will determine how individuals (or somebody or something else) modify the rules they follow. The idea that individuals are rule-followers (because it is rational for them to be so), and are learning by trial and error, is based on the fact that individuals sometimes deliberate and sometimes do not (Hayek 1988; Heiner 1983; Vanberg 1994). In the same way it is possible to demonstrate (Langlois 1986) that it is rational for individuals to follow rules both when the environment is uncertain and when it is certain, and they learn according to the context where they are located. In the same Austrian way, Koppl and Langlois (1994) analyse actions as types of social games. Moreover, it can be assumed that individuals’ behaviours are determined not only by beliefs about their consequences but also by the emotional residues of those actions (Zwick et al. 1999: 17). In other words, if individuals are not always deliberately and consciously acting, and if emotion and other non-deliberated behaviours are important, then learning processes need to take into account the way those non-oriented actions are evolving. The Austrians deny that there can be something like a representative man. Kirman (1999) also develops this kind of conception and supposes that individuals are heterogeneous. In this way, individuals can

Learning in economics: Austrian insights 307 learn by looking at the others’ actions or strategies; learning by imitation is a perfect example of the interest of the ‘heterogeneous assumption’ in terms of individuals’ behaviour. All those conceptions of bounded rationality are closely related to conceptions of the learning process, because of the fact that they are based on one or more hypotheses of perfect rationality. In the second case, individuals learn depending on the degree of instrumental or perfect cognitive rationality they are deemed to have. In the case of perfect instrumental rationality, individuals are able to link without ambiguity the information coming from their environment and the action that allows them to get the best outcome, while perfect cognitive rationality permits them to anticipate others’ behaviour in order to define their strategies. If they have not a perfect instrumental rationality they are learning by modifying their behaviour, taking into account their past outcomes. It is a kind of behavioural learning (Walliser 1998). If they are updating their beliefs concerning the ways others are behaving it is an example of epistemic learning. Indeed, Hayek ([1937] 1948) is only interested in epistemic rationality, while later (1988) he also considers behavioural learning. In both cases, however, individuals are learning because they are not fully rational; the scope of this learning process can be appraised by the distance between the perfect (instrumental or cognitive) rationality and the hypothesis one makes concerning actual individuals’ rationality. How do individuals learn? Young (1998a) distinguishes four models of learning: the first one is based on the natural selection process, the second on imitation, the third assumes that individuals are learning by reinforcement, and the fourth asserts that ‘best reply’ is at the basis of the way individuals learn. Those models are based on different assumptions that concern not only the rationality individuals have at the beginning of the process of learning but also (and more importantly) how those individuals are learning: 1

The first kind of model considers that individuals are assumed not to be deliberately rational at all. ‘Suppose that initially all individuals are genetically or otherwise “programmed” to play some pure or mixed strategy in this game’ (Weibull 1995: 33). In fact, the selection continues in the same way whether we are looking at animals or at human beings. Indeed, individuals survive if they use strategies that allow them to get the best outcome when they are confronted by invaders.10 This kind of learning process is very close, in terms of reasoning, to the ‘as . . . if’ idea developed by Friedman concerning the behaviour of the entrepreneur. At first glance, this kind of model seems not far

308

2

3

Pierre Garrouste from Hayek’s idea that rules of conduct are the result of a selective unintended process. The second type of model considers that individuals imitate others because they are not able to determine the best behaviour. The cornerstone is here to define the way individuals are imitating others and why they are doing so. This kind of learning process is very close to the one Menger implicitly introduces when he says that ‘there is no better way in which men become enlightened about their economic interests than by observation of the economic success of those who employ the correct means to achieving their ends’ (Menger [1871] 1976: 261). Weibull (1995) and Orléan (1998) are perfect examples of such a model. Imitation is also justified by the ‘information cascades’ models where individuals are making a trade-off between their private information and the public information when they calculate the a posteriori probability of a given event (Anderson and Holt 1997). Erev and Roth (1998) exemplify the third type of model. Indeed they are dealing with the fact that individuals change their behaviour on the basis of the following hypothesis: •

• •

• •

4

Respect the law of effect; that is to say, ‘choices that led to good outcomes in the past are more likely to be repeated in the future’ (Erev and Roth 1998: 859). The power law of practice that assumes that the learning curve is ‘normal’. Experimentation (or generalization). The law of effect is completed by the assumption that successful choices are generalized; that is to say, ‘similar choices will be employed more often as well, and players will not (quickly) become locked in to one choice in exclusion to the others’ (Erev and Roth 1998: 963). Recency, which means that recent experiences play a more important role than past ones. Reference point, which is to say something like a reservation or a minimum value (which can be adjusted).

Erev and Roth (1998) show that this model allows them to explain the way people really play games when one compares the theoretical and the actual strategies of individuals. The fourth kind of model is rooted in the process of ‘fictitious play’ that assumes that individuals are behaving as if they think they are confronted to stationary and unknown distribution of others’ strategies. On this basis, Young (1993, 1996, 1998b) introduces the idea that, (a) individuals are able to make mistakes (that is to say, that, with a given probability, they are not choosing their best response strategy), (b) they have a limited memory, and (c) they are looking only at a part of the available set of information.

Learning in economics: Austrian insights 309 All these models are based on assumptions concerning the way individuals learn. The first one deals with individuals learning without knowing that they do. The second one considers that individuals are learning on the basis of a comparison between what they obtain and what the others obtain, or what they believe that the others do. The final two are much more sophisticated because they introduce the idea that individuals learn both by comparing their actions with past experience and by taking into account the ways their opponents behave. The problem is now to look at what individuals do learn. What do individuals learn? In the economics of learning it is often assumed that what is important to analyse is the way individuals modify their behaviour without looking at the cognitive processes that explain those modifications. In others words, it is much more a behavioural than a cognitive learning which is analysed. However, it seems important to look at the process by which individuals build up and modify their ‘representations’ in order to give significance to a complex as well as uncertain reality (Dosi et al. 1996). The Austrian subjectivism is very close to this kind of conception. Such a way of dealing with (behavioural) learning is linked with the fact that learning is fundamentally conceived as an informational problem. Indeed, economists rarely try to analyse learning as a problem of knowledge acquisition (that is a cognitive problem). On this basis, information is defined as a flow and knowledge as a stock. In other words, information is a kind of input that is entering a ‘black box’ that produces an output; that is to say, new information. The way knowledge is modified or not by entering information is not taken into account. However, it is not sufficient to consider information as something that modifies knowledge because this last notion is considered as a passive way of reacting. In fact it seems much more appealing to consider knowledge as a structure. Then, ‘in order for a message to “stick” to the structure – or, more importantly, for a message to modify the structure in a useful way – that message must be meaningful to the receiving system. The message must somehow “fit” ’ (Langlois and Garrouste 1997: 288). It seems the reason why Rubinstein (1998) defines knowledge as a structure of information that allows the defining of a sub-set of the states of the world that individuals believe are the case.

Learning in economics and the Austrian tradition However, these kinds of model do not solve all the problems that learning implies when it is introduced in economics.

310

Pierre Garrouste

Some similarities Without any doubt learning models are improving the economic analysis of individuals’ behaviour, and some of the main assumptions of those models are close to ‘Austrian’ ones: •

•

•

•

First, these models generally assume that individuals are not perfectly rational. Even if it is not always the case, the learning models usually assume that individuals are boundedly rational. Such an assumption is not far from the Austrian idea that ignorance (O’Driscoll and Rizzo 1996) is much more the case than perfect rationality is. Indeed, ignorance is stronger than bounded rationality in the sense that bounded rationality is a ‘positive’ notion; that is to say, that the possibility for bounded rationality to converge to perfect rationality can be assumed (Simon 1980). Conversely, ignorance is much more a ‘negative’ concept. It is impossible to eliminate ignorance because it is a very essential reality. Individuals are necessarily ignorant. Such a negative assumption can, however, be transformed into a positive one inasmuch as individuals search because they are ignorant. In other words, they act because they are not aware of the reality. Second, in some learning models (Erev and Roth 1998; Young 1996) individuals can learn because they are interacting with one another. In fact, it is necessary to take into account the fact that the problem of learning cannot be reduced to a pure logic of choice one but needs to be analysed as a problem of coordination. In other words, people take into account the behaviour of others. Even though this is a basic assumption of game theory and Austrian economics, it is an important aspect of learning models. Third, the idea that individuals can learn without knowing that they do is an important topic. From a Hayekian perspective, individuals learn abstract rules of conduct but they do not know that they learn them. Much more importantly, economic agents can learn something without knowing that they learn it. This kind of conception is not far from the evolutionary game theory conception of learning. In fact, even if Hayek is distinguishing his own conception of evolution and selection and the natural one, such a way of conceiving learning is close to his idea that individuals are using and learning rules of conduct unconsciously. Fourth, imitation as a learning process is common to some learning models as well as to some Austrian ways of analysing the evolution of individual behaviour and the emergence of institutions. Indeed, when he explains how money emerged as an organic institution Menger ([1871] 1976) introduces the idea that some individuals are much more able than the others to exchange economic goods on the market in order to satisfy their own needs: they exchange the goods they have

Learning in economics: Austrian insights 311

•

for more marketable goods. Looking at their success, other individuals imitate them. Progressively, by means of a reinforcement process, an institution is constituted. In a similar way of thinking, Hayek writes that language, morals and such like are ‘the products of selective evolution transmitted by imitative learning’ (Hayek 1988: 24). Young (1998b) explicitly makes a reference to Menger’s conception of the emergence of money and applies his own conception of ‘stochastically stable strategy’ to this topic. Fifth, the ‘updating of propensities’ (Erev and Roth 1998) that assumes that individuals are modifying their propensity to play a strategy by means of a comparison between the payoff of that strategy and the smallest possible payoff, is a kind of trial-and-error learning process. Yet this kind of learning process is perfectly identified by Hayek when he explains how individuals adjust their plans of action on the market. He refers to ‘what, since the physical sciences began to concern themselves with spontaneous orders, or “self-organizing process”, we have learnt to call “negative feed-back” ’ (Hayek 1978: 184).

Important differences Even if some similarities exist between the hypothesis of learning models in economics and some of the Austrian assumptions, it is difficult not to find some important differences between them. It is interesting to specify the main Austrian assumptions. They are as follows: •

•

Subjectivism. Knowledge – inasmuch as it is dispersed – and its growth seem to be basic elements in the Austrian conception of economics. Combined with the Austrian specific conception of time such a phenomenon defines the subjectivism of the Austrian tradition. This makes the analysis of the coordination of individual plans of action the real object of our discipline. Austrians consider that the problem of the coordination of the individual plans of action is solved by a spontaneous process that permits the emergence of institutions (organic ones), rules of action, conventions, traditions and so on. In this way, ‘cognitive commonalties, that is, socially shared tacit knowledge including knowledge about social models of behavior, may emerge spontaneously from intense communication as an unintended collective outcome and may, as such, be difficult to influence’ (Witt 1998: 166). This idea is not Austrian-specific and is now developed by the cognitivist approaches. Entrepreneurship. In fact, ‘Sometimes, however, the institutional setup of the interactions assigns certain individuals a position in which they get a chance to shape the communication processes and thus to exert an influence on the collective outcome. The firm organization is

312

•

Pierre Garrouste a case in point’ (Witt 1998: 166). More generally, the Kirznerian idea of ‘alertness’ is one of the main characteristics of an entrepreneur. Market process. The individuals’ plans of action are adjusted by means of a spontaneous coordination process. As shown above, a selforganized process progressively makes the different plans of action coherent.

The first point seems the more important to deal with in order to compare learning models and the Austrian tradition. Indeed, subjectivism (dynamic one) implies, first, that knowledge is an essential economic variable; second, that time is conceived as real (as opposed to ‘Newtonian’); and third, that individuals are essentially different. On this basis it is possible to suggest the following dissimilarities between learning models and Austrian tradition: •

•

•

First, learning models are essentially behavioural. In other words, they do not look at the cognitive processes of individuals that determine the way their behaviour is defined and modified. Even if some learning models assume that it can be of some interest to look at the individuals’ mental models, or that people can be characterized by some cognitive dissonance (Rabin 1994), learning models are not usually concerned with the cognitive aspect of learning. As we have seen, this lack of interest is due to the fact that they are information-based and not knowledge-based processes. This lack of interest is, however, difficult to assume when some experimental results show that there is a tendency for individuals to maintain a strategy even if the state of the world does not justify this maintenance (Egidi 2000). Second, even if individuals are not considered as identical, they are often so due to certain parameters. In Young’s models, the probability not to play the best-response strategy, the memory of players, and the size of the space of information are exogenous and are given for all players. It is the same thing in the Erev and Roth models. The problem is the uncertainty that those parameters are the same for all individuals, and it seems that they need to be endogenous and it is not sure that they are. Even if the learning models are abandoning the notion of ‘representative agents’, they are obliged to maintain some common characteristics that make their models not really Austrian. Third, time is Newtonian (O’Driscoll and Rizzo 1996). Indeed, by construction, duration is excluded from those models and individuals are supposed to be confronted by the same time. This last difference seems impossible to avoid, inasmuch as those models are formal ones and are not set up in order to take into account such a complex reality.

Learning in economics: Austrian insights 313

Conclusion: towards an agenda Two conclusions are possible. The first takes into account the fact that learning models, even if they incorporate some important elements of the Austrian tradition (bounded rationality, coordination problems, some elements of subjectivism), are not able to integrate some important assumptions that characterize Austrian economics. The second is to take the proximity between learning models and Austrian tradition positively and to define some directions that can be used. As an example, some recent works (Egidi 2000; Garrouste 2000) try to develop cognitive learning models based on the idea that individuals’ minds are characterized by a cognitive inertia and that learning needs to take into account this important characteristic.

Notes 1 Thanks to Nicolai Foss and Roger Koppl for their remarks on an earlier version of this chapter, and to the International Centre for Economic Research (Torino) for its financial support. 2 In fact such a debate is an old one. See, for example, the famous controversy concerning the assumption of perfect rational behaviour of the producer and the natural selection argument used by Alchian (1950) and Friedman (1953). 3 According to Börgers (1996: 1), literature on learning and evolution ‘is concerned with dynamic processes which describe how economic agents adjust their behaviour over time, and how, after agents have gained experience, their behaviour may become rational in the economists’ sense of the word’. 4 Rabin (1998: 32): ‘ “learning” can even sometimes tend to exacerbate errors’. 5 ‘The question of learning, which has received wide attention in other disciplines, notably psychology, has only recently occupied a serious place in the economics and game theoretic literatures’ (Kirman and Salmon 1995: 1). 6 See Fudenberg and Levine (1998), Walliser (1998), Young (1998b). 7 Social learning and, more generally, learning processes based on game evolutionary theory are supposing non-deliberating individuals. 8 Menger ([1871] 1976) was one of the first to introduce the concept of individuals’ limited knowledge. 9 In fact Langlois (1997) is using those two methods, even if he is much more interested in the first one. 10 More precisely, a population with a given profile i and a population of ‘invaders’ in proportion p with a profile j. The strategy i is defined as an evolutionarily stable strategy (ESS) if, u (.,.) being the expected utility of the profiles (.,.), u(i, (1  p)i
pj)  u(j, (1  p)i
pj) for a sufficiently small positive value of p, or equivalently, (1  p)u(i,i)
pu(i,j)  (1  p)u(j,i)
pu(j,j).

References Alchian, A.A. (1950) ‘Uncertainty, Evolution and Economic Theory’, Journal of Political Economy, 58: 211–221. Anderson, L.R. and Holt, C.A. (1997) ‘Information Cascades in the Laboratory’, American Economic Review, 87(5): 847–862. Arrow, K.J. (1962) ‘The Economic Implications of Learning by Doing’, Review of Economic Studies, 29: 155–173.

314

Pierre Garrouste

Binmore, K. (1987) ‘Modeling Rational Players’, Economics and Philosophy, 3: 9–55. Börgers, T. (1996) ‘On the Relevance of Learning and Evolution in Economics’, University College, London, mimeo. Conlisk, J. (1996) ‘Why Bounded Rationality’, Journal of Economic Literature, 34(2): 669–700. Dosi, G., Marengo, L. and Fagiolo, L. (1996) ‘Learning in Evolutionary Environments’, Working paper 1996–05, Computable and Experimental Economics Laboratory. Egidi, M. (2000) ‘Biases in Organizational Behavior’, Trento University, Computable and Experimental Economics Laboratory, mimeo. Erev, I. and Roth, A.E. (1998) ‘Predicting How People Play Games: Reinforcement Learning in Experimental Games with Unique, Mixed Strategy Equilibria’, American Economic Review 4: 848–881. Friedman, M. (1953) ‘The Methodology of Positive Economics’, in Essays in Positive Economics, Chicago: University of Chicago Press. Fudenberg, D. and Levine, D.K. (1998) The Theory of Learning in Games, Cambridge, Mass.: The MIT Press. Garrouste, P. (2000) ‘A Simple Model of Learning with Similarity’, Paris, ATOM, mimeo. Hayek, F.A. ([1937] 1948) ‘Economics and Knowledge’, Economica, reprinted in Individualism and Economic Order, London: Routledge & Kegan Paul. —— (1978) Studies in Philosophy, Politics, Economics and the History of Ideas, London: Routledge & Kegan Paul. —— (1988) The Fatal Conceit, the Errors of Socialism, edited by W.W. Bartley, London: Routledge. Heiner, R. (1983) ‘Origin of Predictable Behavior’, American Economic Review, 73 (September): 560–595. Kirman, A. (1999) ‘Quelques réflexions à propos du point de vue des économistes sur le rôle de la structure organisationnelle dans l’économie’, Revue d’Economie Industrielle, 88: 91–110. Kirman, A. and Salmon, M. (1995) Learning and Rationality in Economics, London: Basil Blackwell. Koppl, R.G. and Langlois, R.N. (1994) ‘When Do Ideas Matter? A Study in the Natural Selection of Social Games’, Advances in Austrian Economics, 1: 81–104. Langlois, R.N. (1986) ‘Coherence and Flexibility: Social Institutions in a World of Radical Uncertainty’, in I. Kirzner (ed.), Subjectivism, Intelligibility, and Economic Understanding: Essays in Honor of the Eightieth Birthday of Ludwig Lachmann, New York: New York University Press. —— (1997) ‘Rule-following, Expertise, and Rationality: A New Behavioral Economics?’, www.lib.uconn.edu/Economics/Working/RATIONL2.HTML Langlois, R.N. and Garrouste, P. (1997) ‘Cognition, Redundancy and Learning in Organizations’, Economics of Innovations and New Technology, 4: 287–299. Menger, C. ([1871] 1976) Principles of Economics, New York and London: New York University Press. O’Driscoll, G.P., Jr. and Rizzo, M. (1996) Economics of Time and Ignorance, London: Routledge. Orléan, A. (1998) ‘The Evolution of Imitation’, in P. Cohendet, P. Llerena, H.

Learning in economics: Austrian insights 315 Stahn and G. Umbhaeur (eds), The Economics of Network: Interaction and Behaviors, Berlin: Springer-Verlag. Rabin, M. (1994) ‘Cognitive Dissonance and Social Change’, Journal of Economic Behavior and Organization, 23: 177–194. —— (1998) ‘Psychology and Economics’, Journal of Economic Literature, 66(1): 11–46. Rosenberg, N. (1982) Inside the Black Box: Technology and Economics, Cambridge: Cambridge University Press. Rubinstein, A. (1998) Modeling Bounded Rationality, Cambridge, Mass.: The MIT Press. Simon, H.A. (1980) ‘From Substantive to Procedural Rationality’, in S. Latsis (ed.), Method and Appraisal in Economics, Cambridge: Cambridge University Press. Vanberg, V. (1994) Rules and Choice in Economics, London: Routledge. Walliser, B. (1998) ‘A Spectrum of Equilibria Processes in Game Theory’, Journal of Evolutionary Economics, 8: 67–87. Weibull, J. (1995) Evolutionary Game Theory, Cambridge: The MIT Press. Witt, Ü. (1998) ‘Cognition, Entrepreneurial Conceptions, and the Nature of the Firm Reconsidered’, Paper presented at the International Society for New Institutional Economics Conference, Paris, September. Young, H.P. (1993) ‘Evolution of Conventions’, Econometrica, 61: 57–84. —— (1996) ‘The Economics of Conventions’, Journal of Economic Perspectives, 10: 105–122. —— (1998a) ‘Individual Learning and Rationality’, European Economic Review, 42: 651–663. —— (1998b) Individual Strategy and Social Structure: An Evolutionary Theory of Institutions, Princeton, N.J.: Princeton University Press. Zwick, R., Erev, I. and Budescu, D. (1999) ‘The Psychological and Economical Perspectives on Human Decisions in Social Interactive Contexts’, in D. Budescu, I. Erev and R. Zwick (eds), Games and Human Behavior: Essays in Honor of Amnon Rapoport, Englewood Cliffs, N.J.: Lawrence Erlbaum Associates Publishers.

14 Beliefs, knowledge and equilibrium A different perspective on Hayek Richard Arena

Introduction A broad consensus prevails today amongst economists about what Caldwell (1988) has labelled ‘Hayek’s transformation’. MacCloughry (1984), Lachmann (1986), Boettke and Prychitko (1994), Fleetwood (1995) or Longuet (1998), for instance, consider that Hayek’s scientific life is characterised by a fundamental break, permitting us to distinguish two different research programmes, the second beginning in 1936 with the presentation of ‘Economics and Knowledge’ (Hayek 1937) to the London Economic Club. More precisely, before 1936 Hayek is broadly described as a Walrasian scholar who thought that resort to the concept of general economic equilibrium is necessary, and who developed a strict formalistic conception of economic theory that left little space for the problem of knowledge. After 1936, Hayek is supposed to have ‘seen the light’, abandoning his Walrasian outlook and building a theory of knowledge, thereby progressively substituting the concept of order for that of equilibrium. This new research programme is supposed to give up the determinism of traditional economic theory for the relativism of a theory of social orders. Boettke, Horowitz and Prychitko express this view when, referring to the notion of social order, they observe that it entails the use of the concept of process instead of that of equilibrium interpreted as a final state. They add: An evolutionary process is open-ended, in that the process does not tend towards any state. Consider what it would mean for human evolution to tend towards a final state. No biologist would even say that we need to have a fully evolved human to understand the process of evolution. (Boettke and Prychitko 1994: 8) The prevailing interpretation therefore implies a kind of opposition between ‘two Hayeks’. The ‘young’ Hayek would have promoted a theory of business cycles ‘superimposed upon the existing theory of equilibrium’, to use Hayek’s own words (Hayek 1939 (1963): 442), ignoring the problem

A different perspective on Hayek 317 of the compatibility of mutual microeconomic beliefs. The ‘older’ Hayek would have discarded the notion of equilibrium, replacing it by a theory of social order and developing a subjective approach to knowledge. The purpose of this chapter is to provide a critical discussion of the perspective just outlined from three different points of view: •

•

•

To begin with, we will provide text-based evidence to show why the prevailing perspective or interpretation, and its main assertions in particular, is highly contentious. This will cast some doubt on the idea that Hayek can be considered as a follower of Walras: our view is that he never actually used the concept of general economic equilibrium in a positive or affirmative sense; that is, in the sense in which it is used in Walrasian or neo-Walrasian economic analysis. We also point out that, even before 1936, Hayek already considered problems related to knowledge and beliefs, and addressed the crucial question of the heterogeneity of individual agents. Moreover, there is also some evidence that, far from having abandoned the concept of equilibrium after 1936, Hayek continued to use it, at least implicitly in his theory of market order, to the end of his life. Second, we will argue that, emphasising the opposition between order and equilibrium, the predominant interpretation of Hayek’s contribution underestimates what appears to us to be a crucial problem in this contribution – namely, the question of the compatibility between Hayek’s permanent reference to the heterogeneity of individual agents and the nature and stability of Hayek’s notion of economic equilibrium. Finally, we will stress that this crucial problem can be solved if Hayek’s overall message is reinterpreted in a way that differs from the prevailing interpretation. Our solution is to suggest that Hayek’s contribution to an economic theory of knowledge and beliefs, far from supporting the traditional notion of economic equilibrium and the existence of a tendency to a unique equilibrium, leads to a different approach favouring disequilibrium dynamics.

Is Hayek’s notion of equilibrium before 1936 Walrasian? A detailed discussion of the problems arising from a Walraso–Paretian interpretation of Hayek’s concept of equilibrium before 1936 can be found elsewhere (Arena 1999). Here, we will briefly recall only the most serious of these problems. One of the main arguments in favour of a Walrasian interpretation of Hayek’s contribution before 1936 refers to the ubiquitous references to ‘equilibrium theory’ in the first chapter of Monetary Theory and the Trade Cycle (Hayek [1933] 1966: 27). These references are not just rhetorical since Hayek agrees on the need to ‘take the logic of economic action as

318

Richard Arena

the starting point’ (p. 30) of the construction of a trade cycle theory. In fact, Hayek describes his method as the operation of ‘widening’ the narrower assumptions of ‘equilibrium theory’ (ibid.). This alone does not, however, imply that Hayek adopts a Walrasian concept of general economic equilibrium. He never actually refers explicitly to the necessity of Walrasian microeconomic foundations or to the view of economic aggregation this entails. As several authors have pointed out, the only straightforward reference Hayek makes to Walras’s theory is to be found in a footnote to Monetary Theory and the Trade Cycle. However, more often than not, reference is made only to the first part of this footnote that we reproduce here in full: By ‘equilibrium theory’, we here primarily understand the modern theory of the general interdependence of all economic quantities, which has been most perfectly expressed by the Lausanne School of Theoretical Economics. The significant concept of this theory was contained in James Mill’s and J.B. Say’s Théorie des débouchés of L. Miksch, Gibt es eine allgemeine Überproduktion – Jena – 1929. (Hayek [1933] 1966: 42, n.2) This quote perfectly demonstrates our point: The ‘significant basic concept’ of equilibrium theory mentioned is market-clearing since this is in fact the only essential assumption shared by J.B. Say, J. Mill, and Walras in this context. From this point of view, Walras’s approach appears to be only one among other possible versions of an economic equilibrium theory, even if Hayek refers to it as reflecting his ‘primary’ understanding of equilibrium theory. We know that J. Mill based his notion of long-run equilibrium on an assisted labour theory of value, while Say interpreted prices as the indirect units of measure of total utilities and Walras linked relative prices to the concept of relative scarcities. All this gives us a clearer idea of what Hayek had in mind: he refers only to the broadly shared notion of equilibrium that economists used in the 1920s; namely, the concept of long-run equilibrium associated to cleared markets and uniform rates of profitability. The second main argument, usually used in support of a Walraso– Paretian interpretation of the ‘young’ Hayek, is based on his 1928 article on ‘intertemporal price equilibrium’. This article is, in fact, frequently assumed to provide a first version of the neo-Walrasian concept of intertemporal general economic equilibrium. Some passages and arguments developed in the text appear to corroborate this interpretation since Hayek indicates, for example, that ‘differences in the supply of some goods between two points in time affect the intertemporal exchange ratios between all goods’ (Hayek [1928] 1984: 79). In other words, Hayek’s article refers explicitly to an intertemporal equilibrium that is assumed ‘to include the possibility of exchange between several individuals of goods

A different perspective on Hayek 319 which are technically equivalent but available at different points of time’ (p. 78). This first impression is, however, misleading. The framework that underlies Hayek’s article is, in fact, based on a temporary equilibrium approach. Time is described as a sequence of ‘flows’ of ‘individual processes’. They form the ‘economic period’ (the ‘year’) that constitutes the horizon within which agents take their decisions (p. 72). Each temporal ‘flow’ corresponds to a sub-section of the ‘economic period’, defined as a ‘day’ or a ‘season’. When a new ‘day’ or a new ‘season’ begins, it brings a flow of new economic ‘processes’ of production. In each sub-period, changes affect production techniques as well as consumer preferences or ‘merely external circumstances’ (p. 73). These circumstances explain why prices are not constant as in a steady state but differ at ‘different points of time’. What Hayek actually defines in this framework is what Hicks later called an equilibrium over time. As Hayek (p. 72), Hicks also remarks that this type of equilibrium does not imply constant prices: Of course, even in a changing economy, people may still expect constant prices, but if they do then expectations are very unlikely to be realised. It will generally be expectations of changing prices which can be realised. In equilibrium, the change in prices which occurs is that which was expected. If tastes and resources also remain what they were expected to remain, then in equilibrium nothing has occurred to disturb the plans laid down on the first Monday. So far as can be, no one has made any mistakes and plans can continue to be executed without any revision. (Hicks 1946: 132) If this situation prevails, it is clear that nobody can gain either by storing goods to try to sell them at a later point of time or at the end of a later ‘flow’, or by buying goods on a future market instead of producing them. Put differently, an equilibrium over time is only logically equivalent to an intertemporal equilibrium in the modern sense of Debreu. In more concrete terms, this means that, at any point in time, it is equivalent for an individual agent to buy goods on a future market or to produce them by himself. Therefore, if the prices that prevail on the different ‘days’ or during the ‘seasons’ are equilibrium prices over time, agents will not try to change these; they will accept these prices at every point in time. As a result, such price fluctuations are eliminated from the picture, so that all acts of exchange are conceptually transferred to a single point in time within the economic period. (Hayek [1928] 1984: 72) This explains why in an equilibrium over time, everything happens as if the individual ‘would make the appropriate decisions at the very outset of

320

Richard Arena

the period’ since he/she is able to foresee all possible economic changes (Hayek [1928] 1984: 75). In this case, any transfer of goods from one subperiod to another must be possible and realised according to intertemporal prices compatible with ‘exchange ratios for goods which are simultaneously available’ (p. 81). The existence of intertemporal prices (in the modern sense) only confirms the absence of any possible speculative gain or loss. Therefore, it would be completely misleading to interpret Hayek as a forerunner of Debreu. If Hayek’s 1928 article anticipates any contribution, it is Value and Capital rather than the Theory of Value. Moreover, Walras himself never constructed any intertemporal version of his general equilibrium theory. It would therefore be confusing to interpret Hayek’s intertemporal equilibrium as Walrasian and, as we have just seen, it would also be mistaken – for different reasons – to regard it as ‘neo-Walrasian’. Finally, the theory of production underlying Hayek’s view of the ‘economic period’ is clearly Austrian and not Walrasian. Hayek’s article begins, in fact, as follows: All economic activity is carried out through time. Every individual economic process occupies a certain time, and all linkages between economic process necessarily involve longer or shorter periods of time. (Hayek [1928] 1984: 71) Thus, it is essential to grasp that Hayek’s theory of production in his 1928 article is based on subsequent ‘flows’ of vertical ‘individual economic processes’ of different lengths in the technically required period of time (p. 72). This clearly excludes the horizontal Walrasian theory of capital. Finally, our view is confirmed by the fundamental distinction Hayek draws between a one-agent economy and a decentralised market economy. The former serves essentially a pedagogic purpose and permits us to understand deviations from the equilibrium over time at the level of the economy as a whole (Hayek [1928] 1984: 77). By contrast, the latter concept allows the introduction of a new and fundamental question; namely, the analysis of the intertemporal compatibility between individual plans decided upon by heterogeneous agents: The obvious precondition for an exchange to take place is that, on this, as on all occasions, those engaged in exchange set relatively different valuations upon the goods to be exchanged. That this precondition can be fulfilled follows from the fact that the temporal ranking of subjective evaluations alluded to above relates wholly to the individual, and so different persons can arrive at two completely opposed sets of valuations. (Hayek [1928] 1984: 78)

A different perspective on Hayek 321 This multi-agent economy also entails the presence of strategic uncertainty that, in turn, can lead to the occurrence of disequilibria. The differences between Hayek’s and neo-Walrasian conceptions of intertemporal equilibrium also appear when actual prices do not correspond to an equilibrium over time or a Debreu-type equilibrium. As we know, in a Walrasian context the solution would be found within the theory of tâtonnement. As in an a-temporal framework, the theory of tâtonnement is assumed to ensure the stability of equilibrium prices and quantities. What happens in the Hayekian framework is, however, entirely different.

Walrasian tâtonnement or Hayekian tendency towards equilibrium? Hayek considers that, in the real world, economic systems are not real but monetary. This implies that agents take decisions based on money prices. This is not important if the economy is in a state of equilibrium over time: The ‘same regularity of behaviours’ ([1928] 1984: 83) will occur. However, what happens if the economy faces a ‘disturbance of this system of prices’ due to ‘influences which are wholly unrelated to the basic impulses of the economy’, namely – to use modern terms – to external shocks (according to Hayek: 83, generally on the supply side)? This problem is less straightforward since, in a monetary economy, the existence of intertemporal prices is no longer necessary: markets can be incomplete provided there is a monetary medium of exchange that ‘permits the individual to store it for the future in a way which is costless (or even yields a positive return)’ (p. 83). This is a general characteristic of temporary equilibrium theory. Now, if shocks occur, the price structure which is [no longer] compatible with equilibrium will have two main consequences; on the one hand, a widening of the differences in supply that are minimised in equilibrium; and on the other hand, whenever equilibrium has not been achieved, the prices actually realised will not represent the greatest returns that could be obtained by the producer. (Hayek [1928] 1984: 91) In this case, expectations can no longer be correct: The assumptions upon which the entrepreneurs have decided to expand their output for the later point in time reveal themselves to be incorrect. Not only will they have to discover that they would have realised better prices by shifting their production to a point nearer in time, in addition, they will have to pay a price higher than that which they expected for the goods necessary for the continuance

322

Richard Arena of production. At the later point and for that reason, for at least some of them, part of the profit they expected would be converted into a loss. (Hayek [1928] 1984: 94)

It is clear, therefore, that any external shock that affects prices and disturbs an initial equilibrium position exerts actual and not virtual effects. Therefore, since economic processes cannot be conceived outside real time,1 disturbances cannot be analysed through the logic time of tâtonnement. Moreover, the presence of money prevents adjustments towards the ‘natural’ position (p. 96): The essential characteristic of every tied currency and of the gold standard in particular is that every change in prices, so long as it is not merely a matter of two such changes accidentally compensating for each other, leads to changes in the quantity of money and hence to further changes in price . . . The answer is that the difference in prices which necessarily arises from the relationship between the supply and demand for goods at two separate points of time, and which serves to bring about equality in provision between them, is partly prevented from being established. (Hayek [1928] 1984: 96) Therefore, the mechanism of supply and demand no longer works and ‘the equilibrium between supply and demand will be disrupted’ (Hayek [1928] 1984: 97). The framework within which Hayek reasons is not therefore the theory of tâtonnement (that does presuppose the perfect functioning of the law of supply and demand) but what Hayek himself calls the ‘theory of economic crises’ (p. 102). Hence, Hayek’s 1928 article already argued that there is a ‘natural’ tendency towards equilibrium that is impeded by ‘the tendency particular to all commodity currencies to stabilise the purchasing power of money even when the general store of supply is changing’ (Hayek [1928] 1984: 97). The ‘tendency towards equilibrium’ is present also in Monetary Theory and the Trade Cycle: To show how the interplay of these prices keeps supply and demand, production and consumption, in equilibrium, is the main object of pure economics . . . It is, however, the task of Trade cycle theory to show under what conditions a break may occur in that tendency towards equilibrium which is described in pure analysis – i.e. why prices, in contradiction to the conclusions of static theory, do not bring about such changes in the quantities produced as would correspond to an equilibrium situation. (Hayek [1933] 1966: 70–71)

A different perspective on Hayek 323 We know the reasons for this break. In a modern monetary economy, the existence of a banking system permits the occurrence of a difference between the monetary and the natural rate of interest. In a Wicksellian perspective, Hayek notes that ‘money being a commodity which, unlike all others, is incapable of finally satisfying demand, its introduction does away with the rigid interdependence and self-sufficiency of the “closed” system of equilibrium, and makes possible movements which would be excluded from the latter’ (Hayek [1933] 1966: 44–45). However, the reason for the occurrence of this ‘break’ is not only related to the existence of money and credit. It has also cognitive origins: In the exchange economy, production is governed by prices, independently of any knowledge of the whole process on the part of individual producers, so that it is only when the pricing process is itself disturbed that a misdirection of production can occur. (Hayek [1928] 1984: 84–85) In Monetary Theory and the Trade Cycle, Hayek went even further. He assumed that the combination of the existence of money and credit with the fact that entrepreneurial knowledge was only based on the pricing process – in this case a false one – entailed that [p]rices become possible which do not operate according to the selfregulating principles of the economic system described by static theory. (Hayek [1933] 1966: 94) Hayek drew the logical conclusion that the economist had to face ‘tendencies leading away from the equilibrium position’ (p. 108). This did not imply that the tendency towards equilibrium ought to be discarded. This tendency prevails in a monetary economy based on money and credit, but it is constantly hampered. Hayek analysed this countertendency. This is why, in Prices and Production (1931) the starting-point is always an equilibrium. In 1937, this notion of a tendency towards equilibrium did not disappear. Quite the contrary, Hayek further specified its role with a view to operationalise the ‘fictive’ concept of equilibrium: Whatever may occasionally have been said by over pure economists, there seems to be no possible doubt that the only justification for this is the supposed existence of a tendency toward equilibrium. It is only by this assertion that such a tendency exists that economics ceases to be an exercise in pure logic and becomes an empirical science; and it is to economics as an empirical science that we must now turn. In the light of our analysis of the meaning of a state of equilibrium it should

324

Richard Arena be easy to say what is the real content of the assertion that a tendency toward equilibrium exists. It can hardly mean anything but that, under certain conditions, the knowledge and intentions of the different members of society are supposed to come more and more into agreement or, to put the same thing in less general and less exact but more concrete terms, that the expectations of the people and particularly, of the entrepreneurs will become more and more correct. In this form the assertion of the existence of a tendency toward equilibrium is clearly an empirical proposition, that is, an assertion about what happens in the real world which ought, at least in principle, to be capable of verification. (Hayek [1937] 1990: 44–45)

Hayek’s characterisation of the ‘tendency toward equilibrium’ in 1937 is particularly interesting because it draws on what are actually two different interpretations of this tendency. The first is ‘empirical’. This relates to Hayek’s remark that, in the real world, this tendency prevails and that this assertion ‘ought, at least in principle, to be capable of verification’. This first interpretation of the tendency towards equilibrium has been made more explicit in The Pure Theory of Capital, when Hayek notes that Its justification is not that it allows us to explain why real conditions should even in any degree approximate towards a state of equilibrium, but that observation shows that they do to some extent so approximate and that the functioning of the existing economic system will depend on the degree to which it approaches such a condition. (Hayek, 1941: 27–28) In a footnote on the same page, Hayek further specifies what he means: A given phenomenon may tend to (approximate towards) a certain magnitude if in a great number of cases it may be expected to be fairly near that magnitude even if there is no reason to expect that it will even actually reach it, however long the time allowed for the adjustment. In this sense, “tendency” does not mean, as it is usually understood to mean, a movement towards a certain magnitude but merely the probability that the variable under consideration will be near this magnitude. (Hayek 1941: 27, n.2) The second meaning is analytical. Hayek correctly observed that in the traditional theory of equilibrium, the assumption that generally underlies the existence of the stability of equilibrium was ‘the assumption of a perfect market’ (Hayek [1937] 1990: 40). This theoretical view corresponds to the Walrasian framework which Hayek explicitly rejects since it entails

A different perspective on Hayek 325 the view that agents ‘even if they are not supposed to be strictly omniscient, are at least supposed to know automatically all that is relevant for their decisions’ (Hayek [1937] 1990: 40). Now, an economist who does not accept this assumption, will have to study: (a) The conditions under which [the] tendency [towards equilibrium] is supposed to exist and (b) the nature of the process by which individual knowledge is changed. (Hayek [1937] 1990: 39) In ‘Economics and Knowledge’ as in The Pure Theory of Capital, Hayek considered that ‘we are still pretty much in the dark’ about both points (Hayek [1937] 1990: 39), and, therefore, ‘that it is probably impossible to formulate any conditions under which such a state would even be fully realised’ (Hayek 1941: 28). Consequently, he preferred to abandon the project of specifying these ‘conditions’ and limited his investigation to the search for empirical probabilities ‘as a guide to the analysis of concrete situations’ (Hayek 1941: 28).

Agents’ heterogeneity and knowledge in Hayek before 1936 Contrary to what has been argued too often, before 1936, the majority of Hayek’s contributions refer to the role of knowledge and to the problem of the heterogeneity of individual agents. These preoccupations already appear in his 1928 article. As has already been noted, Hayek carefully distinguished the economy of ‘an isolated individual’ (Hayek [1928] 1984: 77) from a decentralised market economy, emphasising that in the latter the existence of heterogeneous agents with different preferences was the precondition for exchange. Moreover, he considered the problem of knowledge in relation to the self-regulating price mechanism. Thus, he noted that, if there are ‘periodically recurring changes in the conditions of production’, agents can acquire the knowledge required to ensure the persistence of equilibrium. Hayek distinguishes between three cases of such changes. The first refers to changes ‘which recur with precise periodicity’ (p. 85). Hayek mentions, for example, night and day services, observing that agents know and take into account differences of prices for services supplied regularly at different hours of daily life (e.g. tramway fares). Next come changes ‘which are of uniform tendency in both direction and extent’ (p. 85). Here one could think of economic activities characteristic of a steady state: agents very quickly learn that in this case, every year, employment and output grow regularly. They are, therefore, able to predict future changes and, thus, to contribute to the existence of a dynamic equilibrium. Finally, Hayek refers to changes ‘whose unique occurrence can be confidently expected for a definite point in time, as the result of developments which are currently observable or of known human

326

Richard Arena

decisions’ (p, 85). This corresponds to the case of empirical and tacit knowledge acquired through the observation of regularities. Within the same vein, and as early as 1928, Hayek introduced an analysis of monetary economies in which agents base their decisions on money prices that do not reflect the natural position of intertemporal relative prices. In Monetary Theory and the Trade Cycle, Hayek discusses the reasons that give rise to this discrepancy in some more detail. On the one hand, Hayek observes that, in a market economy, entrepreneurs do not base their decisions on their ‘knowledge of the quantitative increase in the total demand for [their] product’. Rather, they scrutinise ‘the price to be obtained in the market’ (Hayek [1928] 1984: 68): the entrepreneurs’ ‘knowledge of the conditions of production and the market’ that is largely based on tacit local knowledge of future developments permits them to calculate a ‘prospective price’ and expected costs. For the entrepreneurs, these will prevail in the future, and for Hayek they coincide ‘on the average’ with the equilibrium price and costs (p. 70). On the other hand, this self-regulating mechanism of prices is disturbed as soon as we abandon a real market economy for a monetary economy. Hayek argues, in fact, that prevailing money prices are ‘wrong’ prices leading to ‘wrong’ decisions (p. 85). Jack Birner has recently shown how, between the 1928 article to The Pure Theory of Capital, Hayek developed a theory of business cycles that relies ‘on the distinction between real relative intertemporal scarcities, and hence intertemporal relative prices in terms of goods, on the one hand, and the perception that individual decision makers have of these relative prices, on the other hand’ (Birner 1999: 46). He concludes that ‘locally perceived’ relative prices calculated by individual decision-makers from the observation of money prices – that is ‘in their specific circumstances’ – are the relevant magnitudes that lead entrepreneurs to make mistakes (p. 47). This is consistent with Hayek’s critique of the relevance of monetary aggregates or aggregate magnitudes as general price index. Criticising the quantity theory of money (see Arena 2002), Hayek thus writes: We are doing nothing less than this if we try to establish direct causal connections between the total quantity of money, the general level of all prices and perhaps, also the total amount of production. For none of these magnitudes as such even exerts an influence on the decisions of individuals; yet it is on the assumption of a knowledge of the decisions of individuals that the main propositions of non-monetary economic theory are based. It is to this ‘individualistic’ method that we owe whatever understanding of economic phenomena we possess; that the modern ‘subjective’ theory has advanced beyond the classical school in its consistent use is probably its main advantage over their teaching.

A different perspective on Hayek 327 If, therefore, monetary theory still attempts to establish causal relations between aggregates or general averages, this means that monetary theory lags behind the development of economics in general. In fact, neither aggregates nor averages do act upon one another, and it will never be possible to establish necessary connections of cause and effect between them as we can between individual phenomena, individual prices, etc. I would even go so far as to assert that, from the very nature of economic theory, averages can never form a link in its reasoning. (Hayek 1931: 3–5) This quote, taken from Prices and Production, shows that already before 1936 Hayek was perfectly aware that the analysis of agents’ perceptions of their economic environment is essential for economy theorising. This emphasis on the question of individual perception and knowledge is not surprising. It directly derives from Hayek’s subjectivism (Arena 1999) and complements the notion of individual heterogeneity already referred to. This heterogeneity is not only related to differences in agents’ preferences but also to differences in their perceptions. The debate on socialist planning also reinforces the conclusion that Hayek was already sensitive to the problems of individual knowledge and heterogeneity in economic analysis before 1936. In questioning the possibility of using a general equilibrium framework as a guide to rational decision-making in a socialist planned economy, Hayek cast doubts on the ability of central decision-makers to have explicit and codified knowledge of the parameters of the required calculations. He argued that some of the information about a blueprint of productive techniques in the economy was available only in the form of tacit knowledge related to ‘circumstances’ (Hayek [1935] 1948: 155). In 1940, Hayek provided a convincing example of this problem, showing that it was hardly possible to possess codified and explicit knowledge of real markets and activities. He pointed out that it was difficult for a central planner to define a sufficiently standardised commodity and a sufficiently precise list of supplies and demands to define a given market and, therefore, to investigate changes in the mechanisms of supply and demand in real time (Hayek [1940] 1948: 188–189). Hayek also made it very clear before 1936 that the attention he paid to knowledge and individual heterogeneity did not imply some form of psychologism. This view is developed, in particular, in Monetary Theory and the Trade Cycle, where he explicitly addresses the problem of ‘errors of forecast’ in the context of business cycle theory (Hayek [1933] 1966: ch. II, para. VII). He discusses what he calls ‘psychological theories’; that is, theories for which ‘an overestimate of future demand can occasion a development of the productive apparatus so excessive as automatically to lead to a reaction, unprecipitated by other psychological changes’ (p. 83). Hayek’s

328

Richard Arena

own position is clearly anti-psychological, excluding the possibility of systematic errors: [N]o one would deny, of course, that errors can arise as regards the future movements of particular prices. But it is not permissible to assume without further proof that the equilibrating mechanism of the economic system will begin its work only when the excessively increased product due to these mistaken forecasts actually comes on the market, the disproportional development continuing undisturbed up to that time. At one point or another, all theories which try to explain cyclical fluctuations by miscalculations or ignorance as regards the economic situation fall into the same error as those naïve explanations which base themselves on the ‘planlessness’ of the economic system. They overlook the fact that, in the exchange economy, production is governed by prices, independently of any knowledge of the whole process on the part of individual producers, so that it is only when the pricing process is itself disturbed that a misdirection of production can occur. The ‘wrong’ prices, on the other hand, which lead to ‘wrong’ dispositions, cannot in turn be explained by a mistake. (Hayek [1933] 1966: 84–85) In an article written in the same year, Hayek once more stressed these points, reiterating his scepticism regarding the notion of ‘psychological infection’ (Hayek 1939: 141) and arguing that errors cannot be permanent if they are explained only by contagion and that persistent disequilibria could arise only from ‘wrong’ signals or ‘wrong’ rules. There is nothing surprising about this stance, given Hayek’s overall theoretical outlook. According to Hayek, social regularities are indeed essential and characteristic of individual behaviour. Thus he writes in 1942: If social phenomena showed no order except insofar as they were consciously designed, there would . . . be no room for theoretical sciences of society and there would be, as is often argued, only problems of psychology. (Hayek [1942] 1948: 69) This interpretation is further corroborated by comments Hayek made in 1930 on expectations, refuting the idea that entrepreneurs’ errors could be regarded as a sufficient explanation of crises, unless these could be regarded as what he called ‘justified errors’ (Hayek [1939] 1963: 443). While ‘avoidable errors’ are caused by external shocks, ‘justified errors’ arise from ‘guides or symptoms’ that prove to be systematically misleading, as is the case with disequilibrium monetary prices. It does not, therefore, come as a surprise when, in the introduction to ‘Economics and Knowledge’, Hayek comments on the continuity between

A different perspective on Hayek 329 his present preoccupation and those that underlay his theories of capital and of business cycles: It has become more and more obvious that in the treatment of the more ‘dynamic’ questions of money and industrial fluctuations the assumptions to be made about foresight and ‘anticipations’ play an equally central role, and that in particular the concepts which were taken over into these fields from pure equilibrium analysis, like those of an equilibrium rate of interest, would be properly defined only in terms of assumptions concerning foresight. The situation seems here to be that before we can explain why people commit mistakes, we must first explain why they should even be right. (Hayek [1937] 1990: 29) This continuity should not be underestimated, and this is why we will now consider this question in more detail.

Is Hayek’s interest in knowledge after 1936 really new? Hayek’s contributions after 1936 strongly contributed to the construction of a genuine theory of economic knowledge. However, where Hayek considered the problem of knowledge in economic analysis more thoroughly, he would focus on issues and analytical questions that were already present in his contributions prior to 1936. The first theme one might consider in this context is the problem of individual heterogeneity. Already before 1936, Hayek was perfectly aware of the importance of this issue, his interest deriving directly from his methodological choice in favour of subjectivism. However, as seen, he based his approach to heterogeneity on the differences between individual preferences and those related to the agents’ perceptions of their environment. In 1937, Hayek came back to the distinction between a one-agent economy and a decentralised market economy he had introduced in his 1928 article: I have long felt that the concept of equilibrium itself and the methods which we employ in pure analysis have a clear meaning only when confined to the analysis of the action of a single person, and that we are really passing into a different sphere and silently introducing a new element of altogether different character when we apply it to the explanation of the interactions of a number of different individuals. (Hayek [1937] 1990: 35) In a decentralised market economy, the heterogeneity of agents must not run counter to the workings of mechanisms of co-ordination. According to Hayek, this implies at least two requirements. On the one hand, the

330

Richard Arena

‘external events’ on which individual agents found their perceptions, expectations and decisions must belong to ‘the same set’ (Hayek [1937] 1990: 37). Thus, interpretations of the real world might differ between agents, but they still have to refer to a unique and common real world. On the other hand, agents cannot base their plans on purely external facts or information. Rather, some expectations about the future behaviour of other agents must also be included in the decision parameters. Put differently, this is a situation that is characterised by strategic uncertainty but where individual plans are not incompatible a priori. If both these conditions are fulfilled, the data on which agents base their decisions are no longer limited to the so-called ‘fundamentals’ of general equilibrium theory. Rather, they also include ‘subjective’ data that are related to agents’ own specific position in the social interactions that take place in the market economy. However, Hayek also notes that from a subjectivist perspective, things are even more complex than this since all parameters of individual decision-making – including the usual ‘fundamentals’ as well as agents’ expectations of other agents’ behaviour – are being interpreted subjectively. Therefore, they differ not only from one another but also from ‘the objective real facts, as the observing economist is supposed to know them’ (Hayek [1937]1990: 39). This is why in an interview, recorded in 1945 and published in 1994, Hayek referred to the notion of ‘given data’, used by economists to excess, as a ‘ridiculous concept’: The economist assumes (the data) are given to him, but that’s a fiction. In fact, there’s no one who knows all the data or the whole process, and that’s what led me, in the thirties, to the idea that the whole problem was the utilisation of information dispensed among thousands of people and not processed by anyone. (Hayek 1994: 147) It is thus clear why, in the 1930s, Hayek investigated the concept of individual heterogeneity more thoroughly, relating it to agents’ specific beliefs and knowledge. This enquiry led him to abandon the conventional conception of economic man since it became impossible to conceptualise economic individuals within a formalistic framework that encapsulates decisions, made by homogeneous agents, in utility or production functions. It led him furthermore to reject the economic concept of aggregation and the notion of representative agent, inspired by nineteenth-century mechanics, in favour of an analysis of social interaction that is far removed from physical analogies: We know that people will react in the same way to external stimuli which according to all objective tests are different and perhaps also that they will react in a completely different manner to a physically

A different perspective on Hayek 331 identical stimulus, if it affects their bodies in different circumstances or at a different point. We know, in other words, that in his conscious decisions man classifies external stimuli in a way which we know solely from our own subjective experience of this kind of classification. We take it for granted that other men treat various things as alike or unlike just as we do, although no objective test, no knowledge of the relations of these things to other parts of the external world justifies this. Our procedure is based on the experience that other people as a rule (though not always – for example, not if they are colour blind or mad) classify their first impressions as we do. (Hayek 1952: 43) What this clarifies is what Hayek had already stated before 1936 – namely, that psychologism and subjectivism are different animals. It is not totally impossible to understand the behaviour of other agents. In any given society or group, ‘traditions and conventions’ exist that make it possible to establish ‘normally observed rules that make the behaviour of other people predictable to a high degree’ (Hayek 1952: 23). The observation of such rules allows individual agents to discern patterns of purposive action (p. 66) that they find ‘ready in (his) own mind’ and that serve them as tools for the interpretation of other agents’ behaviour. Obviously, the use of patterns to understand other agents’ behaviour is the more efficient, the more familiar individuals are with the group in which they operate. Put differently, rules and their effectiveness are relative to a certain social and cultural environment. The more (less) abstract they are, the more (less) permanent they are and the more (less) they are universally applicable. Thus, what is clear is that while Hayek further developed his notion of subjectivism after 1936, his later writings do not contract his earlier contributions. Hayek also developed his view of the market in relation to problems of the ‘dispersion and imperfection of knowledge’. As we know, he distinguished between two types of knowledge: explicit knowledge that he saw as the ‘knowledge of general rules’, and tacit knowledge that corresponds to the ‘knowledge of the particular circumstances related to time and space’. This second type of knowledge explains why agents act on the market according to rules that they do not understand completely and why they can understand other agents’ behaviour by observing the rules these agents follow. Rules are, therefore, mostly tacit – agents know them but their knowledge is not codified and explicit. Furthermore, these rules tend to be negative rather than positive. It is this feature that also explains why rules are often abstract, defining a range of possible behaviour rather than specific actions. Agents are thus rule-followers. When imitating others, they simply follow established rules or help to generalise their use. When adopting a trial-and-error approach, they either introduce new rules or help to

332

Richard Arena

eliminate some rule, whether these be new or long established. In this respect, too, Hayek only developed – though to a considerable extent – thoughts that can already be found in some of his contributions prior to 1936. It is the continuity of his preoccupation with this topic, too, that characterises his writings before and after the supposed ‘break’. Essentially, he rendered more explicit some of the factors that can explain changes in agents’ knowledge, and that remained somewhat obscured in his earlier writings of the inter-war period. However, some commentators have emphasised that this development implied the abandonment of the concept of equilibrium and the introduction of the new concept of order.

Did the notion of equilibrium disappear in Hayek’s contributions after 1936? There can be little doubt that Hayek continued to use the concept of equilibrium and the notion of tendency towards equilibrium until up to 1968. He merely specified different meanings of both these notions in order to incorporate the role of knowledge into his analytical framework. We have already provided an example of this evolution when we referred to Hayek’s interpretation of the tendency towards equilibrium in his 1937 article. However, a change of mind appears to have taken place towards the end of the 1960s. In a lecture on ‘Competition as a Discovery Procedure’, delivered in 1968, Hayek remarks that [e]conomists usually describe the order which competition produces as an equilibrium – a somewhat unfortunate term, because such an equilibrium presupposes that the facts have already all been discovered and competition therefore has ceased. The concept of an ‘order’ which, at least for the discussion of problems of economic policy, I prefer to that of equilibrium, has the advantage that we can meaningfully speak about an order being approached to various degrees, and that order can be preserved throughout a process of change. While an economic equilibrium never really exists, there is some justification for asserting that the kind of order which our theory describes as an ideal type, is approached in a high degree. (Hayek [1968] 1990: 190) This apparent abandonment of the concept of ‘equilibrium’ and the introduction of the notion of ‘order’ are often regarded as the definite proof of what Caldwell (1988) calls ‘Hayek’s transformation’. According to Caldwell, Hayek embarked on ‘a broader investigation of the formation of institutions to answer the central question of economics, how a spontaneous order might emerge to solve the co-ordination problem’ (Caldwell

A different perspective on Hayek 333 1988: 533). It is, however, worth while to carefully reconsider this transition from equilibrium to order. According to Hayek, an order is defined as [a] state of affairs in which a multiplicity of different constituent parts are related to one another in such a way that, given some knowledge of certain time- and space-related elements of the whole, we could learn how to make correct predictions as to the rest, or at least, predictions that stand a reasonable chance of being correct. (Hayek [1973] 1980: I, 42) Clearly, if it was true that Hayek had substituted a concept of order for the notion of equilibrium, it would be much easier to follow those who claim that the evolution of Hayek’s writing is best perceived in terms of a break or a transformation of his research programme. Tagliagambe has recently pointed out that, within an order, agents are not ‘assumed’ to be rational, but tend to ‘become’ rational for the following reasons: • •

•

Competition, which is fundamentally a discovery procedure permits them to take over data which were previously unknown. The co-ordination of agents’ plans deriving from market, the rules of which facilitate the agreement and the coincidence of expectations which the successes of individual action plans depend on the price system which works in market economies, which offers the signals through which agents find opportunities and elaborate their plans of activity and is the basis of the only relevant information for them, without implying that this activity of elaboration requires a complete information on all the states of the world, present or future. (Tagliagambe 2000: 12)

In this framework, the tendency towards equilibrium is replaced by the tendency towards order to generate and select behavioural norms and rules that permit a progressive convergence of individual expectations and, therefore, a more efficient economic co-ordination. Fleetwood sums up this evolution, stressing that the abandonment of the first tendency and its replacement by the second shows that Hayek definitely renounced to economic determinism and developed a relativist explanation in which a permanent and endless process replaced the process of convergence towards an end state (Fleetwood 1995). There are, however, good reasons to think that this first impression is inaccurate. First, the notion of order is, admittedly, heuristic; but it is also imprecise. A good example of this lack of precision is provided by Hayek’s depiction of the criterion (or the criteria) for concluding that the market order is the superior one. On the one hand, this criterion appears to refer

334

Richard Arena

to an economy’s ability to endow different economic agents with the maximum chance or probability to realise their individual objectives (Hayek 1973: I, 133, 159). On the other hand, Hayek also seems to equate to a Pareto optimum (pp. 137, 143). Strictly speaking, the first interpretation of this criterion is vague, and it is rather difficult to see what it means in economic terms. If, on the contrary, one goes along with the second interpretation, this refers us back to one of the most basic concepts of modern microeconomics. Perhaps, and contrary to what is often claimed, this is, at least in part, due to the fact that the concepts of order and equilibrium are not mutually exclusive. Seen from this perspective, it is worth while recalling a recent contribution by Lawrence Moss in which he points to ‘the lingering residue of equilibrium constructs in Hayek’s later writings’ (Moss 1999: 8). More specifically, he notes two examples of this ‘residue’. The first is Hayek’s article dealing with ‘three elucidations of the Ricardo effect’ (Hayek 1969). Moss is perfectly correct when he notes that in that piece, there is no evidence whatsoever that Hayek turned away from equilibrium economics. Indeed, he used a variety of comparative static or equilibrium diagrams to illuminate his central thesis. (Moss 1999: 10) The second example provided by Moss is more interesting for us since it concerns chapter 10 of The Mirage of Social Justice (Vol. II of Law, Legislation and Liberty, 1976). Here, Hayek explains how the market order can lead to a convergence of individual expectations and, therefore, to the generalisation of individual plans. The process described gives pride of place to the price mechanism, as already noted by Tagliagambe. Moreover, as Moss points out: [Hayek’s] conclusion is that the market order is heavily populated with producers who compare against each other and (therefore) try to minimise the costs of producing the goods and services they bring to market. They end up supplying levels of output where the marginal cost of production just about equals the selling price of the product or services established in the market. Now, it is one of the shining proofs of the general equilibrium school that when all producers balance their hiring of resources against a common (market-determined) set of relative resource prices, they all equate the ‘marginal rate of factor substitution’ in each and every field of production . . . The equating of all marginal rates of factor substitution to a Common relative factor price ratio may be termed the ‘First Equilibrium Condition’ of the achievement of a ‘Pareto optimal’ allocation of resources. Hayek adopts this ‘first equilibrium condition’ – the equality of the marginal rates of factor substitution – and insists that it carries with it

A different perspective on Hayek 335 great explanatory power when studying the market process. His exact words are as follows: ‘the marginal rates of substitution between any two factors have become the same in all their uses, the market has reached the horizon of catallactic possibilities at which the greater possible quantity of the particular combination of goods is being produced which in the circumstances can be produced’ (Hayek 1976: 118). (Moss 1999: 11–12) Moss’s remarks are of particular interest in the present context. They show that, far from having abandoned the concept of equilibrium, Hayek implicitly referred to it when he developed his views on economic order. The examples provided are by no means isolated ones. Late contributions by Hayek to unemployment theory or to competing monetary currencies are difficult to understand if some reference is not made to a theory of market equilibrium. Finally, it is striking how close Hayek’s Nobel Memorial Lecture of 1974 is to his articles on equilibrium and knowledge written before, or in, the 1940s. As we know, these articles continue to make use of the concept of equilibrium even if they give this a more precise cognitive meaning than is present in Hayek’s contributions prior to 1937 (e.g. Hayek [1973] 1990: 140). We are thus led to conclude that in spite of his emphasis on the empirical aspect of the tendency towards equilibrium and his move towards the notion of order, Hayek never completely abandoned the concept of intertemporal equilibrium he adopted and developed in his early writings. In short, in our view, there is more continuity than change in Hayek’s contribution to economics as it evolved with time.

Some critical concluding remarks What, above all, the evolution of Hayek’s thought, as discussed above, highlights are the difficulties he encountered in the implementation of his research programme. The reasons are rather obvious. On the one hand, Hayek was a social scientist who gradually developed a theory of interindividual co-ordination within a framework of the dynamics of market economies. This led him to address problems related to imperfections and failures characterising this process of co-ordination. What is rather striking is that Hayek persistently emphasised difficulties, generally overlooked or ignored by mainstream neo-classical economics, and that he insisted on exploring their causes: the existence of money and credit, the subjectivity and heterogeneity of economic agents, the problems related to the acquisition and diffusion of knowledge. On the other hand, Hayek was also a liberal thinker convinced of the superiority of free market economies over planned or even regulated economies. This conviction is perhaps mainly responsible for his unremitting belief in a tendency towards equilibrium in the real world, whatever the specific analytical version into which he clad

336

Richard Arena

this belief. In our view, both these aspects of Hayek’s thought – theoretical and ideological – can be dealt with independently. But this also means that there is an inevitable choice to be made. Either one focuses on Hayek’s contribution to ‘the liberal principles of justice and political economy’ (the subtitle to Hayek 1976) and delegates his contribution to an innovative economic theory of knowledge and beliefs to the more obscure corners of the history of economic thought. Or else one concentrates on Hayek’s contribution to dealing with the central problem of inter-individual coordination in knowledge-based market economies. In this latter case, it is obvious that we must take seriously Hayek’s Nobel lecture with its emphasis on ‘the great advantage of the mathematical technique’ that ‘allows us to describe, by means of algebraic equations, the general character of a pattern even where we are ignorant of the numerical values which will determine its particular manifestation’ (Hayek [1974] 1990: 143). Seen thus, Hayek’s contribution can find its place in a line of economic thought that leads to contemporary models of multiple agents, social interaction and/or knowledge diffusion. It is, however, equally obvious that, in this case, the ‘tendency towards equilibrium’ will have to give way to a more encompassing theory of complex economic dynamics. For those who have interpreted Hayek’s message as an incentive to construct a new and innovative economic theory that can take into account individual beliefs and knowledge, this then is the choice to be made.

Note 1 The title of the first section in Hayek’s 1928 article is ‘The consequences of economic theory’s characteristic abstraction from time’ (p. 70).

References Arena, R. (1999) ‘Hayek et l’équilibre économique: une autre interpretation’, Revue d’économie politique, no. 6 (November/December). —— (2002) ‘Monetary Policy and Business Cycles: Hayek as an Opponent of the Quantity Theory Tradition’, in J. Birner, P. Garrouste and T. Aimar (eds), F.A. Hayek as a Political Economist, London: Routledge. Birner, J. (1999) ‘The Surprising Place of Cognitive Psychology in the Work of F.A. Hayek’, History of Economic Ideas, Vol. VII, nos 1–2. Boettke, P. and Prychitko, D. (1994) The Market Process: Essays in Contemporary Austrian Economics, Aldershot: Edward Elgar. Caldwell, B. (1988) ‘Hayek’s Transformation’, History of Political Economy, 20(4). Fleetwood, S. (1995) Hayek’s Political Economy: The Socio-economics of Order, London: Routledge. Hayek, F. ([1928] 1984) ‘Intertemporal Price Equilibrium and Movements in the Value of Money’, in: F. Von Hayek, Money, Capital and Fluctuations – Early Essays, London: Routledge & Kegan Paul. —— (1931) Prices and Production, London: Routledge.

A different perspective on Hayek 337 —— ([1933] 1966) Monetary Theory and the Trade Cycle, Reprints of Economic Classics, New York: A.M. Kelley. —— ([1935] 1948) ‘Socialist Calculation II: The State of the Debate’, in Collectivist Economic Planning, London: Routledge. Reprinted in F. Von Hayek, Individualism and Economic Order, Chicago: University of Chicago Press. —— ([1937] 1990) ‘Economics and Knowledge’, Economica, Vol. IV. Reprinted in S. Littlechild (ed.), Austrian Economics, Vol. III, London: Edward Elgar. —— ([1940] 1948) ‘Socialist Calculation III: The Competitive Solution’, Economica, Vol. VII, no. 16. Reprinted in F. Von Hayek, Individualism and Economic Order, Chicago: University of Chicago Press. —— (1941) The Pure Theory of Capital, London: Macmillan. —— ([1942] 1948) ‘The Facts of the Social Sciences’, in F. Von Hayek, Individualism and Economic Order, Chicago: University of Chicago Press. —— (1952) The Counter-Revolution of Science: Studies on the Abuse of Reason, Indianapolis: Liberty Press. —— ([1968] 1990) ‘Competition as a Discovery Procedure’, in S. Littlechild (ed.), Austrian Economics, Vol. III, London: Edward Elgar. —— (1969) ‘Three Elucidations of the “Ricardo Effect” ’, Journal of Political Economy. —— (1973, 1976, 1979) Law, Legislation and Liberty, Vols I, II and III, London: Routledge & Kegan Paul. —— ([1974] 1990) ‘The Presence of Knowledge’, Nobel Lecture, in S. Littlechild (ed.), Austrian Economics, Vol. I, London: Edward Elgar. —— (1994) Hayek on Hayek, ed. by S. Kresge and L. Wenar, Chicago: University of Chicago Press. —— ([1939] 1963) ‘Price expectations, monetary disturbances and maladjustments’, in J. Clark and N. Cohen (eds), Business Fluctuations, Growth and Economic Stabilisation: A Reader, New York: Random House. Extract from Profits, Interest and Investment (F. Von Hayek, ed.), London: Routledge & Kegan Paul. Hicks, J. (1946) Value and Capital, Oxford: Clarendon Press. Lachmann, L. (1986) The Market as an Economic Process, Oxford: Blackwell. Longuet, S. (1998) Hayek et l’école autrichienne, Paris: Nathan CIRCA. MacCloughry, R. (1984) ‘Editor’s Introduction’, in: F. Von Hayek, Early Essays – Money, Capital and Fluctuations, London: Routledge & Kegan Paul. Moss, L. (1999) ‘The Coherence of Hayek’s Thought and the Type of Economic Reasoning Toward Which it Pointed’, Babson College, Working paper. Tagliagambe, S. (2000) ‘Von Hayek e l’evoluzionismo: un appoccio metodologico’, in G. Clerico and S. Rizzello (eds), Il Pensiero di Friedrich Von Hayek – organizzazione, informazione e conoscenza, Torino: UTET Liberia.

15 Guesswork and knowledge in evolutionary economics Veblen revisited Francesca Lidia Viano

Introduction Thorstein Veblen has always been treated as little more than a backstage actor in the repertoire of contemporary orthodox economics. For the best part of a century, from John Cummings to Thomas Sowell, the critics pooh-poohed his theories of the firm and capital as a simple variation on the anticapitalist themes composed by Marx or the socialist and radical currents of thought (Cummings, 1899; Knight 1921, 1969; Dobriansky 1956; Harris 1958; Sowell 1987), while his interpretation of the processes of social evolution was dismissed as a confused and approximate system, the fruit of ‘excessive theorising upon scanty biological and psychological evidence’ (Douglas 1924: 155). One can readily admit that the system which emerges from Veblen’s evolutionary theories is indeed a complicated medley of anthropological, psychological and biological explanations. Yet it was by means of this welding together of dissimilar components drawn from several disciplines that Veblen shaped theories whose essentially innovative nature cannot be denied. In the first place, he forged a new notion of the mind as an alternative to its current interpretations by the associationists, on the one hand, and by the pragmatists headed by Peirce on the other. Of greater importance, however, were his unconventional explanation of the processes of institutional transformation and his demonstration that application of Darwin’s and Spencer’s categories to the study of social evolution was much more complicated than was commonly supposed. In days gone by, when economics cast but a condescending eye on institutional development and psychology, Veblen’s theories of the mind and social transformation were naturally regarded as also-ran subjects and left in peaceful slumber on the pages of his oeuvre for many years. Today, however, when economists are becoming increasingly conscious of the importance of psychological and sociological variables within the economic systems, his evolutionary doctrines are proving valuable sources for interpretation of the life of firms and economic behaviour patterns. Some writers have traced the notion of ‘transaction costs’ back to a definition in

Veblen revisited 339 Veblen’s doctrines (Hodgson 1998a: 397), while others consider that his idea of economics as a continuous and sequential process is re-emerging on the surface of modern theories of ‘cumulative causation’ (Argyrous and Sethi 1996), or trace back to his idea of knowledge as an undivided social heritage the theories of the firm founded on the so-called ‘competence approach’, in which knowledge ‘tacitly and socially produced’ within the confines of a firm is a factor of prime importance in understanding its progress and its place within an economic system (Foss 1998: 480). Yet the attention of contemporary economists has been more firmly held by Veblen’s theory of institutional transformation. In a special issue of the Cambridge Journal of Economics (Vol. 22, No. 4) published in July 1998, Geoffrey M. Hodgson, Paul Twomey and Anne Mayhew reconstruct the links between Veblen’s theories and the evolutionary doctrines elaborated by American psychologists, biologists and anthropologists around the turn of the twentieth century, while Nicolai Foss and (in part) Paul Twomey compare his standpoints with those that have recently emerged in contemporary evolutionistic thinking.1 The idea of examining the contents and sources of Veblen’s evolutionary doctrines is not new in itself. It was launched by the historians of economic thought in the middle of the 1970s, when Edgell discerned the ‘integrating dimension’ of Veblen’s theoretical system at the whole in his evolutionary theories (Edgell 2001: ix; 1975: 267; republished in Wood 1993: 1: 195). Hodgson, Twomey and Foss, however, demonstrate that Veblen’s teachings are not simply a valuable historiographic opera omnia through which one or more stages of his intellectual cursus can be mapped out. In their opinion, his evolutionism embodies the germ of a doctrinal system now well to the fore and used in widely differing fields, ranging from cognitive science to psychology and from economic evolutionism to political economy. When seen from this angle, the historiographic reintroduction of Veblen provides the setting for a more complex operation – namely, investigation of the current status of scientific methods in diverse disciplines, examination of their reciprocal relationships and sounding out the prospects of their possible convergence. Indeed, it is this complicated interweaving of historiographic analysis and methodological investigation that makes the Vebleniana comprised in the Cambridge Journal of Economics an enterprise unique in some respects within the more recent literature on Veblen and his work. Three aims will be pursued in this chapter. The first is to show, through an overview of the early stages of Veblen’s intellectual career, that he moulded his evolutionary doctrines by drawing not only on psychology and avant-garde sociology, as asserted by Hodgson and Twomey, but also on traditional philosophical systems such as those of Kant and Spencer (second section). The second objective is an analysis of Veblen’s relationships with the evolutionism of the American pragmatists (third section), on the one hand, and with some of the emerging evolutionistic disciplines (fourth section), on the other, so as to show that they were more complex

340

Francesca Lidia Viano

and controversial than Hodgson, Twomey and Foss would have us believe. In the last two sections (fifth and sixth), we shall demonstrate that the evolutionistic interpretation of Veblen’s doctrines, insofar as its historiographic basis is somewhat strained, provides the foundation for a culturally significant operation designed to form a sound and lasting alliance between the ‘new’ institutionalism, evolutionary economics and the cognitive sciences.

Veblen’s guessing theory According to Hodgson, Veblen worked out the substance of his doctrines as the result of an ‘intellectual revolution’ (1998b: 416, 422) in the second half of the 1890s. After devoting a good number of years to combing the works of Spencer and Marx in search of an explanation for major social transformations, he turned his back on their theories in favour of the sophisticated models then going the rounds in the pacesetting disciplines of the time: biology, anthropology and the social sciences. From the biology of Jacques Loeb and George Romanes, he had learned to appreciate the importance of causes and principles in scientific research.2 Anthropology had furnished the assumption ‘that sociocultural variation could be arrayed in a unilinear fashion’ (Mayhew 1998: 452). The theories of pragmatists such as James and Peirce had offered a conception of the mind as an active, composite entity (Hodgson 1998b: 417; Twomey 1998: 433–434), while the teachings of C. Lloyd Morgan had shown him that as civilisation progresses there is a transference of evolution ‘from the organism to the environment’ (Morgan, 1896: 340; cited in Hodgson 1998b: 421).3 And so it was, therefore, that Veblen, according to Hodgson, gathered together the strands offered by these new sciences to form a thread along which to escape from the labyrinth of stereotyped nineteenth-century thought and adopt an unconventional attitude towards economic and social processes. Yet the ‘post-Darwinian’ tools thus passed on by Morgan, Romanes and others were rough-hewn, unwieldy and quite unsuitable for the systematic sculpting of new and complex theories. Veblen set about the task all the same without waiting for the scientists to forge a finer and sharper version of their armamentarium. This, however, led him into a trap of his own making. In the 1920s and 1930s, in fact, sociologists and psychologists went on to work out a more sophisticated version of the relations between biology and socio-economic systems and anthropologists put aside the by then cumbersome evolutionistic schemes, whereas Veblen was left to make a last-ditch stand in the narrow trenches of an incomplete paradigm.4 The reconstruction of Veblen’s intellectual path put forward by Hodgson, Twomey and Mayhew certainly introduces an important new perspective into Veblenian studies. From Dorfman to Edgell, in fact, the critics have always viewed his theories as the fruit of a more or less con-

Veblen revisited 341 tinuous and harmonious intellectual development dominated by constant influences. Some have cited Spencer and Darwin as his dominant sources (Edgell 1975; Edgell and Tilman 1989; Eff 1989; Tilman 1996), whereas others have attributed greater weight to the pragmatist matrix (Dorfman 1935; Ayres 1951; Watkins 1958; Dowd 1967; Dugger 1979; Dyer 1986; Jensen 1987). Hodgson, however, has unearthed a new chapter in the Veblen story: a radical and tragic intellectual revolution that led him to abandon subjects close to his heart, such as Spencer’s evolutionism and Marx’s materialism, so as to let in others both new and hard to tame. Even the dramatis personae of Hodgson’s piece are cast in roles unfamiliar to the scriptwriters of Veblen biographies. Walk-on players such as Peirce and William McDougall now top the bill as those who re-wove the intellectual fabric within which Veblen was wont to dwell, whereas the stars of yesteryear, Dewey for example, retire before the likes of Lloyd Morgan, once regarded as no better than other ranks.5 Those whom Veblen himself took pleasure in presenting as the founders of a new and heterodox understanding (the evolutionary anthropologists) turn out to be the epigoni of a culture whose day has come and gone.6 This historiographic paradigm is certainly both attractive and original. Even so, it has its obscure sides. In attributing to Veblen an intellectual revolution that may be supposed to have induced him to renege his Spencerian past and embrace the views of avant-garde psychologists and social scientists, Hodgson overlooks the fact that Veblen had expressed evolutionary ideas cast in the pragmatistic and ‘post-Darwinian’ mould well before the middle of the 1890s when he had not yet come into direct contact with Romanes, James and Morgan. In an article published in the Journal of Speculative Philosophy in 1884, for example, he had maintained that the mind is an active entity able to ‘effect events in the course of nature’ and ‘exert a causality on things’ (1994c: 176). To some extent, this proposition was similar to that being thrashed out by the avant-garde psychologists and philosophers. Veblen, however, makes no reference to their theories, but refers his ideas on the subject to Kant’s by now distant teachings, particularly those contained in his Critique of Judgment. In this work, in fact, Kant inserted between the faculty of the intellect, which supplies a priori the objective laws of nature, and the faculty of reason, which guides a person towards the attainment of a higher moral end, an intermediate faculty, namely judgement, ‘which arranges natural experiences in a systematic fashion’ by applying to nature the finalistic principle that governs a subject’s moral conduct, in other words by representing natural objects as though they had been ordained according to an intention by a superior intellect (Kant 1914: 192).7 Here, indeed, in a subject’s ability to bring external reality home to his own interior ‘constitution’ (p. 487), Veblen had espied the foundation of a new conception of the human mind, no longer viewed as a passive, malleable entity confined to gathering images imprinted on it by the external environment but as a complex system

342

Francesca Lidia Viano

capable of imposing its own rules on nature.8 He had also presented the judgements formulated by a subject under the guidance of the interior ‘finalistic’ rules as a completely new form of inference, an alternative not only to deduction but also to induction itself, which Kant had restricted to the sphere of the intellect and in which the philosophical tradition had discovered the deep roots of natural knowledge. According to Veblen, in fact, in his application of the finalistic principle to nature the Kantian subject did not detect uniformity between given empirical cases, in keeping with the traditional process of induction, but, if anything, made guesses with regard to the elements that were beyond the bounds of his experience and as such were imbued with ‘an element of probability’ (Veblen 1994c: 188). Only in this way, he continued, through probabilistic guesswork, was the subject able to direct his own actions and effectively sway the course of nature (Veblen 1994c: 176): There is always an element of probability, however slight, in our knowledge of particular causes; but simple experience – cognition – never has anything to say about probability; it only says what is, and leaves no room for doubt or probability. In order to find what is the cause of a given effect, and still more, what will be the effect of a given cause, we need a guiding principle beyond anything that experience gives. We have to go beyond what is given us, and so we need a principle of search. That is what is afforded by this principle of adaptation [teleological or finalistic principle] . . . What the principle of adaptation does for us is, therefore, in the first place, that it makes us guess, and that it guides our guessing. (Veblen 1994c: 188–189) Veblen thus maintained that he had drawn from Kant’s Critique of Judgment both a new paradigm of the mind designed to underscore the active nature of the subject and the independence of his cognitive schemata from external conditioning, and a new cognitive method capable of deriving an understanding of the great laws of nature – not from simple empirical generalisations but from probabilistic hypotheses. Closer examination, however, shows that Kant’s theory was much less ambitious. His ‘judgement’, in its original version, was a simple form of induction through which experiences could be reduced to order and not a tool for the elaboration of probable hypotheses about the future course of events. Furthermore, Kant had never admitted that inferences drawn from judgement might allow the subject to ‘sway the course of nature’ to fit his own rules, but had several times insisted that the ‘domain of the concept of nature’ and ‘that of the concept of human freedom’ are ‘totally separate’, since ‘the influence of the concept of nature on legislation dependent on the concept of freedom is as tiny as the disturbance wrought by the latter on nature’s laws’ (Kant 1914: 264, 243).

Veblen revisited 343 The discovery of a form of probabilistic understanding capable of guiding a person’s everyday life was a more recent event. A few years before Veblen’s paper in the Journal of Speculative Philosophy, Peirce had maintained that understanding is dictated by a person’s need to construct his own ‘rules of action’, not merely ‘under such circumstances as are likely to arise, but under such as might possibly occur, no matter how improbable they may be’ (Peirce 1957: 39–41).9 And he also showed that attainment of this form of understanding requires complicated guesses that do not simply explain facts similar to those already observed, as demanded by inductive reasoning, but facts of a completely different nature.10 It was Peirce, then, and not Kant, who invented a new and probabilistic type inference as an alternative to both analytical and inductive inference. Veblen, who had attended Peirce’s ‘elementary logic’ lectures at the Johns Hopkins University, believed that he had found in Kant’s finalistic or teleological principle an effective ‘guide’ for his elaboration of the probabilistic inferences discussed by Peirce (Veblen 1994c: 189).11 It is clear, therefore, that Veblen had imbibed the theories and methods of the pragmatists long before Loeb, Morgan and Romanes imbued him with the principles of the new psychological and biological sciences, and this at a time when his stoa was still haunted by the spirit of Kant. In this connection, far from discerning in pragmatist thinking, as Hodgson would have us believe, an alternative to the traditional patterns on which he had been reared, Veblen shuffled all his cards with a certain nonchalance, so much so that the faculty of judgement, which Kant saw as evidence of man’s freedom and moral independence, was dealt by Veblen as a powerful cognitive tool providing rules for guessing the course of future events. All that remains to be determined, then, is whether at this early stage, when Kant was still in the ascendant and the ‘post-Darwinian’ theories of Loeb and Morgan were below the horizon, Veblen was already drawing from Peirce’s views the teachings whose inculcation Hodgson postpones to a still distant future. According to Hodgson, in fact, both Peirce and James passed on to Veblen ‘an activist and reconstructive conception of human agency’ (1998b: 417) and established ‘Darwinism not merely as a biological but also as a philosophical and methodological creed’ (p. 418). While James ‘saw the Darwinian principle of variation as support for a belief in the indeterminacy of the universe and the reality of human initiative’ (p. 417), both he and Peirce ‘made use of Darwin to uncover spontaneity in nature’ (Russett 1976: 77; cited in Hodgson 1998b: 417). There can be no doubt that the influence of Darwin is everywhere apparent in the theory of knowledge Veblen drew from Peirce’s lectures and writings. Peirce himself averred that the ‘immortal work’ of Darwin was one of the main sources of his own logic. And it was from Darwin, as Hodgson rightly acknowledges, that Veblen derived the idea that nature was a spontaneous and tendentially non-uniform order. This idea, however, was already something of a commonplace for scientists. Eight

344

Francesca Lidia Viano

years before the Origin of Species was published, in fact, Clausius and Maxwell had used it to elaborate a probabilistic model for the study of gas molecules. They showed that while it is not possible ‘to say what the movements of any particular molecule of gas would be on a certain hypothesis regarding the constitution of this class of bodies’, one can none the less ‘predict that in the long run such and such a proportion of the molecules would, under given circumstances, acquire such and such velocities’ (Peirce 1957: 6). Peirce had turned to this very theory to find out how to derive general laws from the observation of non-deterministic systems, and to Darwin’s theories because he saw in them a biological application of the statistical armamentarium that Maxwell had invented to solve problems in physics: Darwin, while unable to say what the operation of variation and natural selection in any individual case will be, demonstrates that in the long run they will, or would, adapt animals to their circumstances. (Peirce 1957: 6) What Peirce acquired from Darwin, therefore, was closer to a set of probabilistic and statistical tools with which to carve general laws and properties from a non-deterministic reality rather than a generic thesis about the ‘indeterminacy of the universe’ (Hodgson 1998b: 417). These, then, were the features of Darwinism that Veblen had acquired through Peirce since the beginning of the 1880s.12 It is certain, of course, as Hodgson rightly points out, that Veblen’s thinking underwent a significant development after the publication of his first articles. He had both read James’s Principles of Psychology (1890) and made the acquaintance of Romanes and other biologists who had drawn his attention, according to Hodgson, to unexplored features of Darwin’s thinking – for example, its strictly causal and mechanistic structure.13 Hodgson cites as the most telling evidence of these influences Veblen’s ‘Why is Economics not an Evolutionary Science?’ (1898b) and ‘The Economics of Karl Marx: I’ (1906), in which he maintained that economics must be studied in terms of cumulative causation and that only ‘an interpretation in terms of opaque causes and effects . . . might have led to a conception of evolution similar to the Darwinian conception of natural selection’ (1994b: 416; cited in Hodgson 1998: 423). One cannot, therefore, rule out the possibility that, as time passed, Veblen moved away from the probabilistic approach he had come to adopt under the influence of Peirce’s logic and began to look for something other than statistical calculations, such as the principle of cumulative causation, for example, or the sequential analysis of natural phenomena, in Darwin’s theories. To be sure, causal analysis had always held a central place in Veblen’s thoughts on the subject of methods. Since his article on Kant in 1884, indeed, he had asserted that knowledge of the natural world cannot dispense with the

Veblen revisited 345 principle of cause and effect, which oversees the organisation and systematic arrangement of data. As far as he was concerned, however, this was no more than a simple organisational function that had nothing to do with the concrete investigation of natural phenomena. In his opinion, the principle of causation alone cannot produce new knowledge: It is true, we have the general law of cause and effect given, and it looks as though we ought to be able to establish some system by the aid of it, when the experience gives us the data to which the law applies; but further thought will show that we should be as helpless with that law as without it if no further principles come in to guide us in the application of it. (Veblen 1994c: 188) That which determines the contents of scientific research, therefore, is not, to Veblen’s mind, the principle of causation, itself no more than a simple tool for the organisation and governance of what has already been learned, but the practical and subjective principle whereby the subject looks into individual causes and individual effects. For conventional scientists accustomed to thinking that the world is an essentially harmonious system dominated by immutable ‘natural laws’, causes and effects represented the points of a static equilibrium and were situated between intervals ‘of transient activity’ (Veblen 1994b: 37): The questions of science were directed to the problem, essentially classificatory, of how things had been in the presumed primordial stable equilibrium out of which they, putatively, had come, and how they should be in the definitive state of settlement into which things were to fall as the outcome of the play of forces which intervened between this primordial and the definitive stable equilibrium . . . They formulated the immutable relations in which things ‘naturally’ stood to one another before causal disturbance took place between them, the orderly unfolding of the complement of causes involved in the transition over this interval of transient activity, and the settled relations that would supervene when the disturbance had passed and the transition from cause to effect had been consummated – the emphasis falling on the consummation. (Veblen 1994b: 36–37) Modern scientists, on the other hand, who had begun to discern in nature a colourless, cumulative sequence of correlated phenomena, had transferred their attention away from ‘initial’ causes and ‘definitive’ effects to the intervals ‘of instability and transition’ between them (Veblen 1994b: 37). In this respect, according to Veblen, the true scientific U-turn had not been determined by the emergence of a scientific paradigm founded on

346

Francesca Lidia Viano

causal analysis, but by the application of by now consolidated forms of analysis to continuous, unstable entities instead of discrete, stable entities.14 Here, too, the source is Peirce, who since 1878 had shown that ‘the conception of continuous quantity’ is ‘the direct instrument of the finest generalisations’, since it allows ‘measurement’ of ‘the passage from one form to another by imperceptible degrees’ (Peirce 1957: 58–59). And he had credited Darwinian naturalists with replacing the study of traditional scientific objects, discrete and countable, by that of continuous entities, such as processes and series. Indeed, it was the attempt to study the continuous entities introduced by the naturalists that induced Peirce to take an interest in the statistical and probabilistic armamentarium. Probabilities, in fact, made provision for the inclusion between zero and one of infinite intermediate fractions indicating ‘the degrees in which the evidence leans toward one or the other’ (Peirce 1957: 60). This was the background on which Veblen had planted the propositions on evolutionisms discussed by Hodgson. When, in 1898 and 1906, he identified the objective of the new evolutionary economics in the study of a colourless sequence of cause and effect, his intention was to bring out the importance of the sequential nature of process to which the principle of causation was being employed, not the principle itself, an old invention.15 According to Veblen, therefore, the sequentiality and opacity of causal series constituted Darwin’s main contribution to scientific development.16 When viewed in this perspective, the post-Darwinian methodological programme worked out by Veblen at the end of the 1890s does not mark the commencement of a new intellectual path, as suggested by Hodgson, but represents, if anything, the fruit of consolidated intellectual experiences whose roots lie in Peirce’s theories from which Veblen had, since the start of his career, drawn the elements for illustrating the active nature of the subject and applying the statistical armamentarium in Darwin’s teachings to the human and social sciences. The statistical principles that Peirce drew from Darwin were a constant source of interest to Veblen from 1884, when he presented them logically and formally, to his articles on economic evolutionism, when he defined them more discursively and translated probabilities in terms of continuous series and processes. In this respect, indeed, even his interest in Darwin’s tenets had maintained the same countenance over the course of time. He had continued to view them not as an exemplary model of ‘causal analysis’, as Romanes suggested, but as the first attempt to apply to natural science the continuous and statistical quantities used by the physicists and the mathematicians.

Idle agents and active subjects The previous section has shown that the methodological schemata worked by Veblen at the end of the 1890s were not the outcome of a sudden and radical intellectual change of course, as Hodgson maintains, but stemmed

Veblen revisited 347 from the time when his thoughts were influenced by Peirce and nascent American pragmatism. An assessment will now be made of his evolutionary ideas to determine how far they were the practical application of what he had thus absorbed. According to Twomey, in ‘attempting to forge a new style of economics, Veblen took for granted many of the ideas’ of the pragmatists (1998: 439) and constructed an avant-garde theory of knowledge around them.17 As we shall see, however, his relationship with pragmatism is much less simple and linear than it may seem at first sight. Veblen had drawn from Peirce clear illustrations of how Darwin’s statistics should be applied in the investigation of human behaviour and social dynamics through the interpretation of habits or ‘rules of action’ as aspects of behaviour that appear more frequently ‘throughout an endless series of occasions’ (Peirce 1957: 83) and are assumed to adapt to the environment in the long run (pp. 119, 220–222).18 Veblen, however, did not fully share Peirce’s confidence in the virtually limitless applicability of Darwin’s biology to the study of the human mind and social organisations.19 In his Some Neglected Points in the Theory of Socialism (1892), he showed that the interaction between habits and the environment is a complicated affair that does not necessarily result in adaptation, not even in the long term. The opportunity to illustrate the gap between material and social evolution lay to hand in Spencer’s essay From Freedom to Bondage (1891), with its assertion that social protests and economic dissatisfaction lack a true foundation, since the industrial society has raised living standards, rendered the means of subsistence and comfort more accessible, and extended personal freedom. Spencer included this comment within a more general criticism of European socialism and a condemnation of movements for popular rebellion, which he accused of wishing to overthrow the freedom generated by economic and industrial development and replace it by a tyrannical system. In Veblen’s opinion, however, the contrast between material well-being and dissatisfaction provided the most evident proof of the friction that normally slows and hinders the adaptation of human institutions to their environment. The fact that industrial development offers a more abundant supply of material goods of better quality to an ever-increasing number of persons does not automatically mean that they must be more fully satisfied because, unlike animals, human beings are consumers impelled by reasons that are more complex than the mere satiation of material wants, such as the need to point up their social status and openly display their wealth.20 In an opulent society, therefore, where a wider accessibility to the means of subsistence and comfort allows people to invest more heavily in the proclamation of their economic success, comparison between one person and another is accentuated and extended to ever-broader sections of the population (Veblen 1994b: 392–394). In underscoring the weight accorded to ostentation and pecuniary emulation in evolved societies, Veblen placed his finger on one of the more

348

Francesca Lidia Viano

serious obstacles to the application of Darwin’s schemata to the study of human and social evolution. Peirce and James felt that in the end human behaviour patterns were destined to adapt to their environment. Veblen, on the other hand, showed that they tend to keep their distance and in so doing go into involutionary reverse. As societies become richer their members spend more and more on eye-catching consumer goods and social ostentation. By thus diverting their income away from productive expenditure and investments, they end up by slowing down technological progress and institutional evolution: it may be said in a general way, that this emulation in expenditure stands ever ready to absorb any margin of income that remains after ordinary physical wants have been provided for, and, further, that it presently becomes as hard to give up that part of one’s habitual ‘standard of living’ which is due to the struggle of respectability, as it is to give up many physical comforts. In a general way, the need of expenditure in this direction grows as fast as the means of satisfying it, and, in the long run, a large expenditure come no nearer satisfying the desire than a smaller one. (Veblen 1994b: 395) The result is that the requirements of pecuniary respectability tend 1) to leave but a scanty subsistence minimum available for other than conspicuous consumption, and 2) to absorb every surplus energy which may be available after the bare physical necessities of life have been provided for. The outcome of the whole is a strengthening of the general conservative attitude of the community. (Veblen 1994a: I, 205) Veblen certainly did not set out to deny that competition and selection constitute the engine of human evolution. On the contrary, he believed that struggle is a constant trait of modern societies and shapes all aspects of their institutions. Even so, he maintained that competition was not determined by the need to adapt to natural conditions, as the Darwinians claimed, but by the desire to conform to the canons set by one’s fellows, to established ‘habits of life’ and institutional traditions.21 In itself, the simple transfer of the selective function from the biological setting to social institutions did not seem to involve major modifications of the Darwinian pattern. All in all, Peirce and James had maintained that habits and all social institutions in general are subject to adaptation to the environment and as such constitute a significant link between man and nature. Veblen, however, was not of the same mind. He felt that institutions were inert and static, little inclined to change in response to pressure from outside.22 They were established ‘mental habits’ and traditional ways of thinking, in other words ‘products of the past process . . . never in full accord with the

Veblen revisited 349 requirements of the present’ (1994a: I, 191). The environment had only moulded them in the early stages of human evolution before the species ‘distanced all competitors’ (Veblen 1994c: 82). In ancient times, when survival of the human-like animals depended on the ability to procure the material requirements of life, people directed all their efforts to productive labour and co-operated to boost their efficiency.23 Later, however, when technological progress increased affluence and the pressure of the environment eased, people put aside habits and ways of thinking best suited for shaping their environment and gathering its most abundant fruits and began to compete with each other for the surpluses that were now being produced. This led to the emergence of new institutions no longer tied to the environment, such as wastage and exhibitionism, which themselves became selection agents in their turn and eliminated ‘those individuals and lines of descent that do not conform to the required canon of knowledge and conduct’ (Veblen 1994c: 88). As seen through Veblen’s eyes, therefore, modern institutions were not the links between man and his environment, as was generally supposed, but even a means of keeping them apart, a sort of cavity wall that shielded people from the environment and its selective function, yet at the same time became itself the selector and by perpetuating their obsolete habits and states of mind removed them even further from the material world.24 Nor, indeed, was natural selection the only tool used by institutions to mould social structures. Habits, Veblen believed, did not stop at discouraging new social and psychological trends, but were transmitted to future generations through the mechanisms described by Lamarck and taken over by Spencer.25 Due to the pressure exerted by the new institutions, the ‘acquired proclivity’ to struggle and waste was turned into a ‘transmissible trait, and action under its guidance becomes right and good’ (Veblen 1994c: 88). It was these features, in fact – namely, gradual emancipation of institutions from the environment, on the one hand, and the interaction between mechanisms of selection and those of transmission of the characters acquired, on the other – that steered social evolution along a course rather different from that which Darwin had encountered in biology. Organic evolution of the species was a story of continuous transformation and relentless amelioration, whereas the story of society, as it appeared to Veblen, was a complicated affair marked by delays, moments of inertia and survivals: men’s present habits of thought tend to persist indefinitely, except as circumstances enforce a change. These institutions which have so been handed down, these habits of thought, points of view, mental attitudes and aptitudes, or what not, are therefore themselves a conservative factor. This is the factor of social inertia, psychological inertia, conservatism. (Veblen 1994a: I, 191)

350

Francesca Lidia Viano [the] institutions of today . . . are the result of a more or less inadequate adjustment of the methods of living to a situation which prevailed at some point in the past development; and they are therefore wrong by something more than the interval which separates the present situation from that of the past. (ibid.: 207)

more in particular, the institution of a leisure class, by force of class interest and instinct, and by precept and prescriptive example, makes for the perpetuation of the existing maladjustment of institutions, and even favours a reversion to a somewhat more archaic scheme of life; a scheme which would be still farther out of adjustment with the exigencies of life under the existing situation even than the accredited, obsolescent scheme that has come down from the immediate past. (ibid.: 213)26 Paradoxically, therefore, the selection and hereditary transmission mechanisms, introduced by Darwin and Lamarck respectively to explain the transformation and gradual perfection of species, became in Veblen’s eyes agents of inertia and social conservation. This original thesis distinguished the arrangement of his evolutionary system from that of contemporary psychologists and sociologists. Even those such as Morgan, who had noted the progressive transfer of selection from the environment to institutions, in fact, continued to presume that selection was still the vector of progress and transformation as opposed to stasis, and excluded its interaction with genetic inheritance. Indeed, that which induced Morgan to shift the accent from the biological to the social variables was his attempt to explain technological and institutional progress in a stage when ‘in biotic and genetic terms humankind had evolved only to a very slight degree’ (Hodgson 1998b: 421). From this standpoint, therefore, one cannot agree with Hodgson’s suggestion that Veblen constructed his own view of evolution around the arguments put forward by Morgan in 1896. Not only had Veblen detected in institutional selection the origins of social stasis as opposed to progress, he had also begun to give shape to his theory of evolution at the beginning of the 1890s in his paper on Spencer, and had since simply drawn from an on the whole rudimentary pot-pourri of Spencerian, Darwinian and Peircian theories the conclusion that institutions were ‘emergent entities in the socio-economic sphere’.27 We must, however, still ask ourselves whether Veblen’s rejection of the Darwinian schemata constructed by Peirce, James and Morgan did not undermine the views of the human mind and knowledge he had himself formulated some time earlier under the influence of some of Peirce’s ideas. He had, in fact, observed that the pragmatists’ theories of the

Veblen revisited 351 subject rested on the notion that man had ‘by selective necessity’ developed a natural tendency towards ‘purposeful action’ that induced him to turn ‘the forces of the environment to account’ (Veblen 1994c: 80).28 It was Veblen, too, who showed that technological and social progress had freed man from the influence of the environment, leaving him with archaic, useless habits, such as waste. Peirce’s and James’s theories of the subject were thus likely to sink to the level of obsolete and poorly effective tools whose maximum utility would solely lie in the study of industrious savages in primitive times. Nor did Veblen make any move to dissipate this impression. After the appearance of his The Theory of the Leisure Class, indeed, he drew a clear distinction between his views and those of the pragmatists. They included among pragmatic actions both the knowledge of facts and their ‘conversion to use’, whereas he considered that cognitive processes were not subject to the pragmatic or teleological law (Veblen 1994b: 8–9n). In his opinion, a subject formulated two ‘chains of responses’ to every external stimulus: one ‘pragmatic’, i.e. devoted to the attainment of a useful result, the second ‘unintended’ and ‘irrelevant’ in terms of immediate utility (ibid.: 6–7). The irrelevant responses were always associated with the pragmatic responses, though the bond between them was stronger ‘where a higher range of intelligence was present’; in other words, on the passage from purely executive to cognitive activities (ibid.: 7). They are no more than slightly suggested by the need to produce some appreciable result and furthermore rest on a simple instinct of ‘idle curiosity’, a sort of ‘aptitude for play’ that induced people to investigate the world and assign it a purpose that did not depend on their own practical requirements (ibid.). So that associated with the pragmatic attention there is found more or less of an irrelevant attention, or idle curiosity. This is more particularly the case where a higher range of intelligence is present. This idle curiosity is, perhaps, closely related to the aptitude for play, observed both in man and in the lower animals. The aptitude for play, as well as the functioning of idle curiosity, seems particularly lively in the young, whose aptitude for sustained pragmatism is at the same time relatively vague and unreliable. (Veblen 1994b: 6–7) This did not amount to a denial by Veblen that the individual is first and foremost an ‘agent that acts in response to stimuli afforded by the environment in which he lives’ (Veblen 1994c: 80). Even so, he noted that the subject rarely formulates his response ‘in terms of an expedient line of conduct, nor even necessarily in a chain of motor activity’ (Veblen 1994b: 7), because on the one hand his natural propensity for efficiency is often distorted and obfuscated by human institutions directed to exhibition or waste, while on the other hand his intellectual curiosity tends to obfuscate

352

Francesca Lidia Viano

his practical sense. Under the guidance of his idle curiosity, the individual transfers ‘the pragmatic animus’ from his own action to that of the objects observed: The facts are conceived in an animistic way, and a pragmatic animus is imputed to them. Their behaviour is constructed as a reasoned procedure on their part looking to an advantage of these animistically conceived objects, or looking to the achievement of some end which these objects are conceived to have at heart for reasons of their own . . . The dramatisation of the sequence of phenomena may then fall into somewhat less personal, less anthropomorphic formulations of the processes observed; but at no stage of its growth – at least at no stage hitherto reached – does the output of this work of the idle curiosity lose its dramatic character. Comprehensive generalisations are made and cosmologies are built up, but always in dramatic form. (Veblen 1994b: 7, 8)29 Veblen had thus identified, alongside the pragmatic intentionality discovered by Peirce and by James, a different, cognitive intentionality through which the vocation to act is transferred from the knowing subject to the objects he observes.30 And it is indeed this intellectual, indirect intentionality that survives even when man is freed from the pressure of the environment. While the pragmatic instinct, in fact, is often inhibited by modern institutions, which encourage leisure and waste, reason is free to attribute ends to the natural things it sees around it. It thus becomes clear why Veblen looked for the ordaining principle of empirical knowledge in Kant and not in Peirce. Peirce, in fact, maintained that probabilistic guesses are formulated under the guidance of rules of practical action, whereas Kant discerned the guiding principles of the investigation of nature in an individual’s moral universe and particularly in his will, which renders him master of his desires and leads him towards higher ends. By choosing this path, of course, Veblen was compelled to conclude that man can only acquire knowledge through finalistic and anthropomorphic patterns and that even the sequential analysis of causes and effects retained ‘animistic’ and teleological traits. According to Veblen, that which makes mechanistic and sequential analysis a more appropriate tool of knowledge is the simple fact that it is a schema that ‘can be turned to practical account’, though it is in itself still the fruit of an ‘imputation of activity to the observed objects’ that never ceases, not ‘even in the latest and maturest formulations of scientific research’ (1994b: 15, 16).

Veblen revisited 353

A ‘sabotage-based approach’ The previous sections have shown that Veblen moulded his conception of evolution and cognition in an original manner with respect to pragmatism. He did not share Peirce’s and James’s faith in the adaptability of human habits to the environment and emphasised the delays and incongruities that dominate the development of social institutions. He also worked out a theory of knowledge that linked intentionality not so much to a person’s propensity to act with a view to an end as to his tendency to attribute an end to natural external objects. The following section will examine the relationships between Veblen’s theories and the evolutionistic contents that have taken shape in some emerging disciplines. Twomey and Foss have provided significant indications of the ways in which Veblen’s tenets can be brought within contemporary evolutionism. Twomey has shown that the conception of the active, composite mind, which, in his opinion, Veblen borrowed from the pragmatists, has now re-emerged in the latest forms of psychology and artificial intelligence, while Foss has demonstrated that even without engaging ‘in much rational reconstruction of Veblen’s ideas’ one can see a number of ‘similarities’ between them and evolutionary studies of the firm tied to the so-called ‘competence-based approach’ (1998: 480). By contrast, the dissimilarities between Veblen’s arrangement and that of the pragmatists that were brought out in the third section suggest that the placement of his theories should be at least partly different from that proposed by Twomey. When reinterpreting the pragmatistic paradigm of the mind in the light of Kant’s finalism, Veblen stripped it of the ‘active’ and ‘reactive’ traits that have now re-emerged in psychology, and confined it within the meanderings of a conventional and somewhat obsolete philosophy. We shall not enter, therefore, into a detailed discussion of recent psychological evolutionism in the search for contents bearing Veblen’s stamp. Instead we shall set out to determine whether, as Foss suggests, recent evolutionary theories of economics embody any of the original traits of Veblen’s teachings. The evolutionary theories of the firm examined by Foss rest upon the idea that ‘largely tacit and socially produced and reproduced firm-level knowledge’ (1998: 480) is the main key to understanding inter-firm diversity (Nelson 1994), the sources of competitive advantage (Wernerfelt 1984) and firm growth and development (Penrose 1959). It is indeed true that Veblen maintained that knowledge is produced collectively by the accumulation of past experiences and hence that the history of industrial techniques embraces the history of man and his thinking habits. Yet were we to go beyond Foss and engage in ‘much rational reconstruction of Veblen’s ideas’ we would find that his idea of the collective production of knowledge springs from a theoretical background foreign to the models currently applied in the study of firms. In the first place, he considered that no portion of the economic and social system can be singled

354

Francesca Lidia Viano

out and examined as a mechanism in its own right with no reference to its relations with the other portions: No member of the system, no single article of doctrine, is fairly to be understood, or defended except as an articulate member of the whole and in the light of the preconceptions and postulates which afford the point of departure and the controlling norm of the whole. (Veblen 1994b: 411)31 In the second place, Veblen reckoned that a firm was totally incapable of excogitating independent knowledge to be applied in its manufacturing operations. Firms, to his mind, were no more than bureaucratic and commercial structures that exploited their formal ownership of raw materials and the means of production to unduly appropriate the stock of knowledge and competencies that these embody: It is the ownership of materials and equipment that enables the capitalisation to be made; but ownership does not of itself create a net product, and so it does not give rise to earnings, but only to the legal claim by force of which the earnings go to the owners of the capitalised wealth. Production is a matter of workmanship, whereas earnings are a matter of business. And so the question returns: What are the circumstances by force of which industry yields a net product, which can be turned to account as earnings? . . . It appears, therefore, that the prime creative factor in human industry is the state of the industrial arts; that is to say, the determining fact which enables human work to turn out a useful product is the accumulated knowledge, skill, and judgment that goes into the work . . . For the transient time being, therefore, any person who has a legal right to withhold any part of the necessary industrial apparatus or materials from current use will be in a position to impose terms and exact obedience, on pain of rendering the community’s joint stock of technology inoperative to that extent. (Veblen 1994a: IX, 61, 62, 65) The conclusion to be drawn, therefore, is that the only knowledge produced within a firm consists of stealing or sabotage techniques whereby technological skills are withdrawn from the community: By virtue of this legal right of sabotage which inheres as a natural right in the ownership of industrially useful things, the owners are able to dictate satisfactory terms; so that they come in for the usufruct of the community’s industrial knowledge and practice, with such deductions as are necessary to enforce their terms and such concessions as will induce the underlying population to go on with the work. (Veblen 1994a: IX, 67)

Veblen revisited 355 Two conditions, therefore, must be met before a set of collective knowledge can be regarded as a factor of production: (1) it must spring from industrial experience only and be free from any form of commercial experience; (2) it must be an indivisible whole rooted in the entire population and generated by the labours of all members of the community.

New evolutionary paradigms It is clear, therefore, that the evidence in favour of Veblen as the fons et origo of the new evolutionary paradigm now taking shape in economics and sociology is rather weak. This final section, therefore, will leave aside the historical references put forward in the special issue of the Cambridge Journal of Economics and show that, by crediting Veblen with the original formulation of the notions of evolution now advanced by very dissimilar cultural currents, Hodgson, Twomey and Foss are making him the symbol of a new cultural synthesis in which a broadly understood evolutionary schema can be used to study individuals and institutions, their reciprocal relations and the general pattern of economics. In their presentation of Veblen as a founder of modern evolutionism, the papers in the special number hark back to a topic mooted for the first time by Abraham Harris in the Journal of Political Economy in 1932 and revived by Stephen Edgell, for whom Veblen’s ‘theory of evolutionary change’ was the ‘most misrepresented and undervalued [topic] in the history of the subject’ (1975: 267 = 1993: 195). Edgell’s interpratation of Veblen’s ideas about evolution had mainly provided an opportunity to free current interpretations of his work from the stereotyped opinions mostly put forward by the institutionalists who, in seeking to express these ideas in familiar clothing, confined them within the pragmatistic schemata around which they had constructed their own notions of psychology (Eff 1989: 711). This operation was also appreciated outside the institutionalistic panorama and gave rise to a long historiographic tradition. Edgell, however, put Veblen’s theories of knowledge aside and used his evolutionary conceptions to reveal new sources of this thought, such as Bellamy’s populism, Darwin’s doctrines, and Spencer’s theories of evolution (Edgell 1975, 2001; Edgell and Tilman 1989; Eff 1989; Tilman 1996). In drawing attention to Veblen’s evolutionary manifesto, therefore, Hodgson, Twomey and the other writers in the special issue would appear to be following in spirit the path on which Edgell embarked in 1975, though in effect their purpose is more exacting and involves a medley of the historiographical schemes moulded both by the literature about Veblen’s evolutionism and by the institutionalistic presentations of Veblenian thought. On the one hand, they follow the suggestions of Edgell and Tilman and seek the origin of Veblen’s ideas in late-nineteenth-century Anglo-American culture as a whole, including Spencer’s ‘biological determinism’, Peirce’s pragmatism, Loeb’s materialism, McDougall’s

356

Francesca Lidia Viano

spiritualism and the anthropology of Tylor and Boas; on the other, they scale down the influence of Spencer urged by Edgell, Tilman, and above all Eff, and place greater emphasis on that of pragmatism. In this respect, they dredge up early historiographic hypotheses bearing the institutionalistic stamp and designed to assimilate Veblen’s ideas to those of Peirce and James, but by examining them in the light of the material provided by Edgell and Tilman they construct a more complex and finely divided picture of American pragmatism and its surroundings. Through these complicated historiographic alchemies Hodgson transforms the Veblen story into a single, exemplary vicissitude in which two mainstreams of the heterodox economic tradition – namely, institutionalism with its pragmatistic hinterland and an evolutionism now Darwinian, now Lamarckian – flow as one. Veblen is thus presented as the standardbearer of an avant-garde economics, around which currents of thought that are even poles apart, but none the less united by their distrust of the conventional devices of economic analysis, could be gathered. Yet those who visit the Veblenian sanctuary are now well aware that such an ambitious amalgam is not to be had. Richard Nelson, another contributor to the special issue of the Cambridge Journal of Economics, holds that economics must develop within its own four walls a theory of economic growth and a theory of institutions, but by no means regards his programme as an offshoot from the Veblenian stock. In the work An Economic Theory of Evolutionary Change, published in conjunction with S.G. Winter in 1982, Nelson draws on both Simon and Schumpeter in his elaboration of an evolutionary theory and pays no heed to the Veblenian tradition nor the institutionalistic tradition in general. Simon is his source for the concept of limited as opposed to the conventional maximising rationality, whereas the notion of competition as a process with winners and losers rather than a simple facet of markets whose myriad traders are faced with predetermined prices is taken from Schumpeter. At the beginning of the 1990s, Nelson, as a representative of the evolutionary movement that began in the 1950s,32 had, along with Winter, been one of the interlocutors of Hodgson and his institutionalism. Nelson and Winter constitute the ideal continuation of the tradition inaugurated by Abramowitz, Solow and Swan, who set out in their very different ways to establish an economic growth theory ‘to guide empirical research’ (Nelson 1998: 502). The evolutionary vogue, however, was not limited to the ocean of economics. A few years after Nelson and Winter had launched their ‘evolutionary change’ model, Leda Cosmides and John Tooby struck out on a somewhat parallel course along the new seas of psychology and provided evolutionary explanations of the composite and modular nature of the human mind (Twomey 1998: 443). The stage was thus set for the formation of a powerful coalition to develop the opposition to the neoclassicists. Nelson and Winter had succeeded in incorporating the conception of evolution into a formal system without

Veblen revisited 357 needing to resort to the neoclassical armamentarium. They put aside the fundamental postulates of maximisation of profit and perfect information and assumed that firms are heterogeneous sets of habits and routines in reciprocal competition to comply with the conditions of the market. This hypothesis was used to distinguish both a microeconomic theory of change founded on competition between the routine stocks of individual firms and its selection in keeping with the economic terms offered by the market, and a macroeconomic theory in which the generation and distribution of new technologies or the diffusion of routines within a given industry are described by application of evolutionary models of populations (Markov processes, Leontief coefficients and probabilistic models of in-house research). Psychologists, too, had used different tools to propose a model for the formalisation of Darwin’s evolutionary theory. They had stressed both the importance of mental processes that take place outside the usual channels (Twomey 1998: 441) and the heterogeneous nature of the mind with its functionally specialised circuits (p. 443). By inviting Nelson and the psychologists to contribute to the special issue (in which, however, Nelson makes no mention of Veblen), Hodgson embarked upon a significant cultural endeavour – namely, to create a place for Veblen in the background of evolutionary theories that do not refer back to him and substantiate this operation within the dispute on institutionalism.

Evolutionistic interpretations of Veblen’s evolutionism Hodgson’s Economics and Institutions: A Manifesto for a Modern Institutional Economics was published in 1988. A neo-institutionalist current, of course, had been around for some time. A house divided against itself, it had none the less found a substantial reason for aggregation in its efforts to relaunch institutionalism by proposing it in a version more readily reconcilable with the positivist views that had begun to dominate the culture of the English-speaking world since the 1940s. The most representative institutionalist theoreticians, such as Williamson, Coase and Posner, had aimed at radical emanicipation from the by now impaired inheritance of the old institutionalism, universally regarded as descriptive and antitheoretical. Their project consisted of explaining the birth and development of institutions, such as a firm or the state, by reference to a rational, individual behaviour model. Individuals and their preference functions were thus taken as data according to the example of the neoclassicists. In his Manifesto, Hodgson criticised the approach of Williamson and Coase (1988: 154–156, 180–181, 199–201) and put himself forward as the initiator of an institutionalism capable of reconciling the need to formalise modern economics with a return to the theories of Veblen. He viewed the evolutionary theories as valuable allies with which to counter the ascent of

358

Francesca Lidia Viano

the anti-Veblenians within the institutionalist school (pp. 131, 141–144, 208). In his eyes, evolutionism resembled a hidden passageway through which Veblen’s ideas had continued to exert their influence on contemporary economic thought, even when the old institutionalism had sunk below the horizon (p. 125).33 Certain it is that the evolutionists of the 1950s did not proclaim themselves as continuers of Veblen. They were either economists engaged in adapting economic theories to the empirical results that had emerged in statistical analyses of the American national product, or psychologists who, by rejecting the conventional interpretation of the human mind, interpreted it as a multi-tiered entity determined by habits and instincts. Anyway, in its more recent expression (economic evolutionism of Nelson and Winter, cognitive psychology), evolutionism seemed to revalue Veblen’s interpretations of habits and instincts and the idea of a substantial interdependence of individuals and institutions (pp. 101, 131, 208). Paradoxically, therefore, the original features of Veblen’s system – rejected by the institutionalists – seemed to have survived in currents foreign to the institutionalistic environment, such as Hodgson’s evolutionism or the new psychologies of the mind. On this basis, Hodgson looked for an agreement with Nelson and Winter, on the one hand, and with the psychologists on the other. Even so, Nelson and Winter did not seem at ease in Hodgson’s ad hoc interpretation of their theories. They had always described themselves as the spokesmen of contemporary Schumpeterism, but now all of a sudden someone was telling them that the systems they had constructed so far not only went back to Veblen, but were even more Veblenian than those of the institutionalists themselves. As to the psychologists, they had tapped the tradition of James and McDougall and were probably unaware of the fact that in the meantime someone had discovered that Veblen, too, was one of the pragmatists. Today, however, it would seem that the wind has veered. Nelson and Hodgson are partners in the celebration of the centenary of Veblen’s essay, while Twomey places Veblen’s theories of instincts and habits among the sources of the cognitive scientists. Yet is this indeed a major peace treaty signed in the name of Veblen? Were it so it would not simply lay the foundations of a new transverse alliance erected around a common Veblenian identity, but would also launch an interpretation of Veblenian evolutionism at odds with that put about by the revisionists. Where Tilman, in fact, had described Veblen’s model as a heterogeneous blend of materialism and spiritualism, and of Lamarckian and Darwinian evolutionism, Hodgson and his co-authors credit Veblen with a linear model of evolution that may be supposed to embody the seed of modern theories of the firm founded on the so-called competence-based approach, as well as the nucleus of the current paradigm of the mind worked out by the cognitive scientists. Yet one cannot help thinking that the entente cordiale from which the special issue of the Cambridge Journal appears to spring still

Veblen revisited 359 lacks its finishing touches. If, instead of attributing to Veblen the project of an evolutionary economics both discursive and wholly hostile to formalisation, Hodgson, Twomey and Foss had recognised his interest in empirical research and statistical investigation, the Veblenian programme would have acquired a guise more familiar to some of the contemporary evolutionists who are still, like Nelson, deeply influenced by the post-Second World War tradition of empirical and statistical studies and in search of a formal theory within which the ‘appreciative’ results of such studies can be duly incorporated.34

Conclusions As we have seen, the interpretation of Veblen’s thinking proposed by the special issue of the Cambridge Journal of Economics can be viewed in two lights: as a historiographic contribution to the literature on Veblenian thought, and as a theoretical contribution to the contemporary discussion of the evolutionary method in economic theories. From the historiographic standpoint, it offers a significant means of reassessing Veblen’s teachings and appreciating their influence on the latest trends in economics and psychology. Assimilation of Veblenian evolutionism to that of contemporary economists and psychologists, however, is a risky undertaking. The psychological evolutionism now in vogue can be at least partly traced back to the American pragmatistic tradition. Hodgson and his coauthors, too, point to the underlying analogies between the Veblenian system and that of the pragmatists as a means of assigning Veblen’s theories a place among the archetypes of the new cognitive science. Two fundamental distinctions, however, can be drawn between Veblen’s positions and those of the pragmatists: (1) Veblen denies the complete convertibility of Darwin’s evolutionary schemata from the biological to the social sphere and shows that when selective processes are transferred into the heart of the institutional system they switch from agents of transformation into vectors of stasis and inertia; (2) they attribute intentionality to man’s tendency to assign ends to the action of external objects rather than to his own. The roots of these differences derive from the fact that when Veblen elaborated his theory of knowledge and evolution he did not, pace Hodgson, solely draw upon pragmatism and post-Darwinian science, but mixed these sources with suggestions of a different kind plucked from the earlier philosophical systems of Kant and Spencer. Equally hazardous would seem the attempts of those who set out to make Veblen the forefather of the new tendency in firm economics to solve the main questions concerning the diversity of firms or the sources of their competitive advantages by observing the evolution of the social and group knowledge elaborated within their bounds. Veblen, in fact, had always refused to agree that a firm could generate knowledge of its own to be applied in its manufacturing operations. Productive knowledge was

360

Francesca Lidia Viano

elaborated by society as a whole. Firms were fictitious structures of a formal, bureaucratic kind through which businessmen appropriated common competencies for use to their own advantage and to the detriment of the community. Of greater interest, instead, is the contribution of the Veblenian issue of the Cambridge Journal of Economics to the current debate on the method of the economic disciplines. In this respect, its discussion of Veblen’s theories can be viewed as a simple pretext for embarking on an indirect examination of more general topics of a theoretical kind. Underlying the papers of Hodgson and the other contributors is an attempt to build an image of Veblen in which economists such as Nelson and Winter, hitherto totally extraneous to the Veblenian and institutionalistic traditions and linked, if anything, to the Schumpeterian inheritance, could recognise themselves. In this respect, therefore, the reinterpretation of Veblen’s theories sought by Hodgson and the other authors of the Cambridge Journal is not just an addition to the myriad expressions of the revisionism emerging within the new Veblenian criticism, but a programmatic text within which a sound and lasting alliance can be forged between neo-institutionalism, cognitive science and economic evolutionism. The special issue of the Cambridge Journal thus becomes an opportunity to celebrate the birth of a new institutionalism which, freed from the inheritance of the by now discredited ‘old institutionalism’, restores the original contents of Veblen’s thinking and makes them the foundations for a wider understanding with the new avant-gardes of economics and psychology.

Notes 1 In this way, by gathering the contributions of Hodgson, Twomey, Mayhew, Foss and others, the Cambridge Journal of Economics has celebrated the centenary of the publication of the first article written by Veblen on evolutionism in economics, ‘Why is Economics not an Evolutionary Science’ (1898b). The integral text of this article is also offered in the special issue of the Cambridge Journal besides the articles of Hodgson, Twomey and others. 2 ‘At the same time, the influential biologist George Romanes insisted that Darwinism above all meant causal analysis’ (Hodgson 1998b: 417). See also Dorfman: ‘While Caldwell was analysing the traditional classifying procedure in economics, George Romanes was pointing out that Darwinism meant causal analysis, not classification’ (1972: 95). 3 Hodgson’s thesis, in fact, is that ‘the idea of an evolutionary process of selection of institutions began to develop in Veblen’s work’ after 1896 when Morgan had presented his ideas for the first time during a lecture delivered at the University of Chicago and published his Habit and Instinct (1998b: 420–423). 4 ‘Two final points concerning subsequent developments should be noted. First, Veblen never abandoned the idea that biotic factors may have some significant influence on human behaviour, unlike the many social scientists who attempted in the first half of the twentieth century, and with increasing vigour, to sever all links between biology and social science’ (Degler 1991). Second, Veblen never

Veblen revisited 361 developed the philosophical concept of emergence sufficiently. This task was left to a group of philosophers writing mainly in the 1920s and 1930s (Alexander 1920; McDougall 1929; Morgan 1927, 1933; Whitehead 1926). Not only was this a tragedy for Veblen – who died in 1929 and was unable fully to address these philosophical developments – but it was also a disaster for American institutionalism as a whole. Despite the theoretical efforts of John Commons (1934) and others, it failed to develop an adequate understanding of the relationship between the biotic and the socio-economic worlds and was thus impelled by the 1930s to join the majority of social scientists who were for severing all links between biology and social science’ (Hodgson 1998b: 427). In Mayhew’s view, ‘even as Veblen issued his call for an evolutionary economics – a call to contribute to a grand structure that seemed to be on the verge of creation – the evolutionary vision in the social sciences had began to fade’ (1998: 452). In particular, in ‘anthropology . . . a variety of forces led to the disappearance of the goals and methods of what became the discarded and disavowed evolutionary school’ (ibid.). 5 Dorfman includes Morgan among the ‘apperceptionists’, i.e. Dewey, James, Caldwell and Stuart, who influenced Veblen from the start of the 1890s (1972: 119, 120, 125, 139, 152). The influence of McDougall was stressed in 1948 by Louis Schneider in his The Freudian Philosophy and Veblen’s Social Theory, but his views were not widely accepted. As Tilman explained many years later, the similarity between McDougall’s and Veblen’s classifications of instincts is no more than apparent, since the former rests on a spiritualistic vision of the mind, whereas the latter is permeated by materialistic principles (1996: 79). Tilman is more inclined to accept the links between Morgan’s psychology and that of Veblen, though without according them any greater weight than those which united Veblen to the philosophy of James. In more general terms, he holds that neither James nor Morgan succeeded in ‘renouncing the premises of idealism’, an -ism that Veblen had always regarded as suspect (Tilman 1996). According to Tilman, the incongruities of Veblen’s evolutionary system stemmed from the very fact that he had sought to gather together a variety of dissimilar sources without establishing their ranks, to yoke Lamarckians and Darwinians, on the one hand, and materialists such as Loeb with spiritualists such as McDougall, Morgan and James on the other. ‘It does not appear’, he declares, ‘that many of the scholars who trace most of Veblen’s psychology to these four authors have read them carefully. They have simply taken Veblen’s citations and brief discussion of the sources at face value or have uncritically accepted the claims made by others’ (p. 74). 6 In a letter to one of his pupils, Sarah McLean, Veblen acknowledged that the tools most appropriate for renewing the armamentarium of economics came from the anthropology with which he was most familiar, namely that of Thomas, Starr, Ammon and Lapouge, Tylor and Spencer himself (‘As for the anthropological reading, which I have inveigled you into, I do not know that it will be of much direct use, but it should be of some use in the sense of an acquaintance with mankind’, Dorfman 1972: 132–133). 7 ‘Now, because the conception of an object, to the extent that it also contains the principle of the reality of such object, is known as purpose, whereas the term finality of the form of a thing is applied to the agreement of such thing with that constitution of things which is solely possible by means of ends, the principle of Judgement, with respect to the form of natural things subjected to empirical laws in general, is the finality of nature in the variety of its forms. In other words, nature is represented, by means of this concept, as if an intellect were to contain the single basis of the multiplicity of its empirical laws’ (Kant 1914: 249).

362

Francesca Lidia Viano

8 ‘the more nearly the activity of the mind in thinking a given thought coincides with what would be the mind’s activity if that activity were guided by its own intrinsic laws alone and were not influenced or hampered by the environment – the more fully will the requirements of the mind’s activity be realised’ (Veblen 1994c: 181). 9 ‘[T]he whole function of thought is to produce habits of action; . . . What the habit is depends on when and how it causes us to act. As for when, every stimulus to action is derived from perception; as for the how, every purpose of action is to produce some sensible result. Thus we come down to what is tangible and conceivably practical, as the root of every real distinction of thought, no matter how suitable it may be; and there is no distinction of meaning so fine as to consist in anything but a possible difference of practice’ (Peirce 1957: 40–41). 10 In Peirce’s view, the idea that induction was an appropriate tool for getting to know the truth of reality was the fruit of firmly established preconceptions, such as Mill’s idea that the world is substantially uniform, or that the conditions provided by experience comprise ‘all that is universally true’, as postulated by Kant (Peirce 1957: 102). According to Peirce, therefore, reality is not a deterministic system. To understand it, one needs to subject hypotheses to ‘infinite series of empirical tests’ (1957: 79), on the one hand, and to derive the probability of one’s initial hypothesis, in other words ‘the proportion of cases in which it is true’ (1957: 63), from the frequency of true inferences with respect to false inferences. 11 Peirce was a temporary lecturer at Johns Hopkins from 1879 to 1884. Veblen was a student there for a trimester in 1881 and attended Peirce’s ‘Elementary Logic’ lectures (Dorfman 1972: 41; Dyer 1986: 30). 12 This means that one cannot accept Mitchell’s assertion that Darwinian biology was totally extraneous to statistical and exact sciences, and that Veblen had in common with Darwinism a deep-rooted dislike for mathematics and statistics. Introducing a collection of Veblen’s writings, Mitchell wrote: ‘Representatives of the “exact” sciences stress the importance of measurement. There are those, indeed, who go so far as to claim that the outstanding characteristic of thought is its quantitative precision. Now Darwinian biology was not an exact science; it made but slight use of measurements in any form; it confined itself mainly to “qualitative analysis” . . . Veblen was a good Darwinian in this respect also. His native bent was toward speculation of a philosophical sort . . . It was easy for one who had little liking for mathematical procedures to overlook the promise of statistics. Finally and most important, problems of cumulative change in “life-history” are exceedingly difficult to treat by any method of measurement . . .’ (Mitchell 1936: xxx–xxxi). Mitchell seems to overlook the fact that – as Peirce stressed in 1878 – the advantage of mathematical treatment comes not ‘so much from the conception of number as from that of continuous quantity’, and that the naturalists, not the mathematicians, were the first who took ‘advantage of the idea of continuity, or the passage from one form to another by insensible degrees’ (Peirce 1957: 58, 59). In the same way, Veblen underscored the centrality of continuous quantities in mathematical and statistical procedures: although mathematicians and physicians ‘deny the substantial continuity of the sequence of changes that excite their scientific attention’, all their efforts are directed towards the demonstration ‘that apparent action at a distance must be explained by effective contact, through a continuum’ (Veblen 1994b: 33n, 35n). On the other hand, Veblen never concealed his appreciation for experimental implications of scientific method. In his article on Kant’s Critique of Judgment, he clearly stated that the teleological principle ‘proceeds on the basis of a feeling, and so it can decide whether the hypothesis suits the

Veblen revisited 363 mind, but not at all whether it applies to reality. Experience alone can say whether the hypothesis fits the things it is intended for . . . As soon as a result is obtained by the process of induction, that result becomes, for the purposes of the question in hand, a fact of empirical knowledge, and therefore acquires the character, not of a completed whole, but of an isolated and disconnected datum. As fast as one step of induction is completed it becomes a means to another step, which must inevitably follow it . . . The kantian, reflective, judgement, therefore, is the tool of a “subjective” knowledge, though the evidence by which it seeks to establish the results aimed at is of empirical origin’ (Veblen 1994c: 190, 192). 13 ‘At the same time, the influential biologist George Romanes insisted that Darwinism above all meant causal analysis. Instead of taxonomy and the accumulation of facts, “causes or principles are the ultimate objects of scientific quest” ’; Veblen ‘interpreted Darwinism as essentially a causal analysis of process’ (Hodgson 1998b: 417, 423). 14 In the 1898 article to which Hodgson refers, indeed, Veblen showed that causal analysis had become a fixed feature of scientific analysis: ‘It will not even hold true that our elders overlooked the presence of cause and effect in formulating their theories and reducing their data to a body of knowledge’; ‘With the earlier as with the later generation the basis of valuation of the facts handled is, in matters of detail, the causal relation which is apprehended to subsist between them’ (Veblen, [1898b] 1998: 405). The relationship between the causal approach and statistical science, seen as the science of the continuum, is complicated. Hermann Weyl considered it impossible to ‘derive statistical regularities from causal laws’ (Weyl 1920: 117), because causal processes are reversible in time, whereas statistical processes are irreversible. Maxwell, too, had put the matter very clearly many years earlier. In a lecture given in 1873, he stated that the determinists (those who accept the causal law ‘from the same antecedents follow the same consequents’) base their doctrines on ‘the stability of evolution: a small variation in initial state is supposed to produce only a small variation in the future’ (Von Plato 1994: 86), whereas those who accept statistical laws believe in a ‘world in which the same antecedents never again concur’, and where processes are unstable: ‘an infinitely small variation may bring about a finite difference in the state of the system in a finite time’. This is the main reason why Maxwell contrasted ‘singularities and instabilities’ with ‘continuities and stabilities’, and stressed how the first tended to prevail in ‘phenomena of higher complexity’ (Maxwell 1873: 442, 440, 444). When he explained the difference between the old and the new science in 1908, Veblen may be supposed to have followed Maxwell’s paradigm: ‘The characteristic feature by which post-Darwinian science is contrasted with what went before is a new distribution of emphasis, whereby the process of causation, the interval of instability and transition between initial cause and definitive effect, has come to take the first place in the inquiry; instead of that consummation in which causal effect was once presumed to come to rest. This change of point of view was, of course, not abrupt or catastrophic. But it has latterly gone so far that modern science is becoming substantially a theory of the process of consecutive change, realized to be self-continuing or self-propagating and to have no final term . . . Modern science is ceasing to occupy itself with the natural laws – the codified rules of the game of causation – and is concerning itself wholly with what has taken place and what is taking place’ (Veblen 1994b: 37, 38). 15 ‘[I]t is only question of time when that [substantially animistic] habit of mind which proceeds on the notion of a definitive normality shall be displaced in the field of economic enquiry by that [substantially materialistic] habit of mind

364

16

17

18

19

20

Francesca Lidia Viano

which seek a comprehension of facts in terms of cumulative sequence’ (Veblen, [1898b] 1998: 414); ‘modern science is becoming substantially a theory of the process of consecutive change, which is taken as a sequence of cumulative change, realised to be self-continuing or self-propagating and to have no final term’ (Veblen 1994b: 37). This is the meaning of Veblen’s sentence (quoted by Hodgson): ‘an interpretation in terms of opaque causes and effects could have led to a conception of evolution similar to the Darwinian conception of natural selection’ (1994b: 416). In a previous article, Hodgson himself acknowledged that continuity plays a central role in the processual analysis theorised by Veblen, but he failed to recognise its probabilistic derivation (Hodgson 1993b). ‘[T]he economic theory of Veblen and other early institutionalist thinkers developed under the influence of a more sophisticated view of the mind. This new perspective stressed the active and multi-tiered nature of the mind in which instincts, habits, and conscious reasoning are all significant for understanding human behaviour. The philosophical background for this new perspective of the mind was provided by the American pragmatists – Charles Sanders Peirce and William James’ (Twomey 1998: 437). ‘It seems incontestable, therefore, that the mind of man is strongly adapted to the comprehension of the world . . . How are we to explain this adaptation? The great utility and indispensableness of the conceptions of time, space, and force, even to the lowest intelligence, are such as to suggest that they are the results of natural selection. Without something like geometrical, kinetical, and mechanical conceptions, no animal could seize his food or do anything which might be necessary for the preservation of the species. He might, it is true, be provided with an instinct which would generally have the same effect; that is to say, he might have conceptions different from those of time, space, and force, but which coincided with them in regard to the ordinary cases of the animal’s experience. But, as the animal would have an immense advantage in the struggle for life whose mechanical conception did not break down in a novel situation . . . there would be a constant selection in favour of more and more correct ideas of these matters’ (Peirce 1957: 119–120); ‘In the evolution of science, a Darwinian mode of evolution might, for example, consist in this, that at every recall of a judgement to the mind . . . a slight fortuitous modification of the judgement might take place; the modified judgement would cause a corresponding modification of the belief-habit, so that the next recall would be influenced by this fortuitous modification, though it would depart more or less from it by a new fortuitous modification’ (pp. 222–223). It is also true that, on some occasions, Peirce seemed to raise doubts about Darwin’s theories and their total ability to explain the evolutionary process. In his essay ‘Evolutionary Love’ ([1893] 1968), he identified three evolutionary paradigms (‘evolution by fortuitous variation’, or Darwinian; ‘evolution by creative love’, or Lamarckian; and ‘evolution by mechanical necessity’) and asserted they worked together towards the development of species. Nevertheless, all this does not contradict his faith in the progressive trend of human development and the successfulness of adaptation mechanisms, whatever their origin. This very conviction – as we shall see – is the element that most distinguishes his thought from Veblen’s ideas. It was Veblen, in fact, who not only stressed the relevance of regressive trends, exogenous disturbances and delays in the evolutionary process, but made them the most important elements in the course of human and social development. ‘As we are all aware, the chief element of value in many articles of apparel is not their efficiency for protecting the body, but for protecting the wearer’s respectability; and that not only in the eyes of one’s neighbours but even in

Veblen revisited 365

21

22

23

24

25

26

27

one’s own eyes . . . And the chief motive of dress is emulation – “economic emulation”. The like is true, though perhaps in a less degree, of what goes to food and shelter’ (Veblen 1994b: 395). ‘The cause of discontent must be sought elsewhere than in any increased difficulty in obtaining the means of subsistence or of comfort’ (Veblen 1994b: 391); ‘the cause [of social discontent] is very often that what might be the means of comfort is diverted to the purpose of maintaining a decent appearance, or even a show of luxury’ (p. 392); ‘In our fundamentally industrial society a person should be economically successful, if he would enjoy the esteem of his fellowmen’ (p. 393); ‘All this will hold with still greater force of a generation which is born into a world already encrusted with this habit of mind’ (p. 394). In Hodgson’s opinion, Veblen attributes to institutions the role genes normally play in Darwinian biological theories, i.e. the transmission of selected traits: ‘by recognising the durable character of institutions, Veblen thus discovered an equivalent to the gene in the socioeconomic world . . . institutions and settled habits do have a sufficient degree of durability to regard them as having quasigenetic qualities’ (Hodgson 1993a: 132). It is interesting to note, however, that replacement of genes by institutions or habits as vehicles of hereditary transmission inevitably leads one to embrace Lamarckian and not Darwinian conclusions, and assert the hereditary transmission of acquired characters, since the institutions are the fruit of human efforts. It was Peirce himself who first declared that ‘Lamarckian evolution is thus evolution by force of habit . . . Habit is mere inertia, a resting on one’s oars, not a propulsion’ (Peirce, [1893] 1968: 281). ‘Self-interest, as an accepted guide of action, is possible only as concomitant of a predatory life, and a predatory life is possible only after the use of tools has developed so far as to leave a large surplus of product over what is required for the sustenance of the producers’ (Veblen 1994c: 87). On the contrary, in Hodgson’s view, Veblen credited the institutions with the capability of adapting themselves to the environment: ‘Socioeconomic evolution is regarded as a selection process, working on institutions as units of selection, combined with the simultaneous processes of adaptation of both individuals and institutions to their mutual environment’ (1993b: 132). Darwin himself had recognised some form, although limited, of heredity of acquired characters, but relegated it at the margin of his own system: ‘On the whole, we may conclude that habit, or use and disuse, have, in some cases, played a considerable part in the modification of the constitution and structure; but that the effects have often been largely combined with, and sometimes overmastered by, the natural selection of innate variations’ (Darwin 1884: 127). In his most recent book, Edgell maintains that Veblen’s insistence on delays in evolutionary development is drawn directly from Darwin, since it was a confirmation of his great law ‘we are all the descended from barbarians’ (Darwin, 1874: 613; cited in Edgell 2001: 114). This is certainly true. Yet Darwin set out to understand the mechanisms that could have led man to distinguish himself from his simpler forebears, whereas Veblen sought the reasons for the survival in him of archaic traits. Darwin worked out a theory of transformation, Veblen a theory of conservation. ‘With Morgan’s intervention, the scene was set for Veblen’s intellectual revolution: the concept of the evolution and selection of institutions as emergent entities in the socio-economic sphere. It is perhaps no accident that at about the time of Morgan’s visit to Chicago the idea of an evolutionary process of selection of institutions began to develop in Veblen’s work’ (Hodgson 1998b: 422). This somewhat shaky conclusion is surely belied by the fact that Veblen’s

366

Francesca Lidia Viano

paper on Spencer in 1892 was written for the very purpose of reformulating Darwin’s theory of selection by shifting the accent from the biological to the social variables. 28 ‘Like other animals, man is an agent that acts in response to stimuli afforded by the environment in which he lives. Like other species, he is a creature of habit and propensity . . . He is in an eminent sense an intelligent agent. By selective necessity he is endowed with a proclivity for purposeful action’ (Veblen 1994c: 80); ‘Man’s great advantage over other species in the struggle for survival has been his superior facility in turning the forces of environment to account’ (ibid.); ‘[All men] like to see others spend their life to some purpose, and they like to reflect that their own life is of some use. All men have this quasi aesthetic sense of economic and industrial merit, and to this sense of economic merit futility and inefficiency are distasteful. In its positive expression it is an impulse or instinct of workmanship; negatively expresses itself in a deprecation of waste’ (ibid.: 81). According to Veblen, the instinct of workmanship ‘is a human trait necessary to the survival of the species’, whereas its negation, in other words, ‘antipathy to useful effort’, is ‘a habit of thought possible only in a species which has distanced all competitors’ (ibid.: 82). 29 ‘Pragmatic’ is here used in a more restricted sense than the distinctively pragmatic school of modern psychologists would commonly assign the term. ‘Pragmatic’, ‘teleological’, and like terms have been extended to cover imputation of purpose as well as conversion to use. It is not intended to criticise this ambiguous use of the term, nor to correct it; but the terms are here used only in the latter sense, which alone belongs to them by force of early usage and etymology. ‘Pragmatic’ knowledge, therefore, is such as is designed to serve an expedient end for the knower, and is here contrasted with the imputation of expedient conduct to the facts observed (Veblen 1994b: 8–9n). 30 It is, of course, true, as Dyer says, that Peirce himself attached great importance to ‘disinterested’ forms of knowledge: ‘It begins passively enough with drinking in the impression of some nook in one of the three Universes. But impression soon passes into attentive observation, observation into musing, musing into a lively give and take of communion between self and self. If one’s observations and reflections are allowed to specialise themselves too much, the play will be converted into scientific study; and that cannot be pursued in odd half hours’ (Peirce 1966: 361). But Dyer does not seem to appreciate the fact that Peirce inserted ‘musement’ in the same process of adaptation to the environment within which ‘active science’ and practical actions lay. As Dyer acknowledges, musement for Peirce ‘describes that esthetic state of mind where a person, confronted with some mystery, takes to wondering how the mystery might be solved and expresses the tentative solution as a hypothesis’ (Dyer 1986: 35. See also Peirce 1966: 367: ‘The inquiry begins with pondering these phenomena in all there aspects, in the search of some point of view whence the wonder shall be solved’). Peirce’s musement therefore derives from the necessity to solve problems which emerge in a state of doubt. And, being the doubt a condition characterised by ‘privation of habit’, it induces the inquirer to an ‘erratic activity that in some way must get superseded by a habit’ (1966: 189). This is the reason why Peirce could say that the ‘essence of belief is the establishment of a habit’ (1957: 39) and that ‘the most striking feature of the new theory [of pragmatism] was its recognition of an inseparable connection between rational cognition and rational purpose’ (Peirce 1966: 202, 184). On the contrary, Veblen stressed the difference between the two (see note 29) and showed that cognition founds itself upon the attribution of rational purpose to the object of inquiry.

Veblen revisited 367 31 Veblen is referring here to Marx’s economic theories but, as Edgell stated in 1975, Veblen himself ‘consistently investigated particular phenomena such as production, consumption and competition, and developed theories related to them, always in context of his general evolutionary theory of change. In other words, in theory and practice Veblen regarded the discussion of particular features of social and economic life in isolation as unprofitable and essentially “unscientific” ’ (Edgell 1975, republished in Wood 1993: I, 198). 32 It is Nelson, too, who identifies the deep roots of the ‘growth theory’ in Moses Abramowitz’s Economics of Growth (1952): ‘I believe that such a broad theory of growth already exists, at the appreciative level, and in fact has existed for a long time. Abramowitz’s 1952 essay contains most of the key elements and traces their connections in a plausible way’ (Nelson 1998: 514). 33 One of the most telling examples is Katona’s concept of habit, which resembles the one constructed by Veblen (Hodgson 1988: 125). 34 ‘[E]mpirical studies conducted in the 1950s, 1960s and early 1970s enormously increased our understanding of economic growth. Most of their findings have not been overturned by subsequent studies’ (Nelson 1998: 506). ‘The problem is in formal theory. And here . . . the developments that fall under the rubric of the new neoclassical growth theory are of very limited help’ (p. 514).

References Abramowitz, M. (1952) ‘Economics of Growth’, in A Survey of Contemporary Economics, Vol. II, Homewood: Richard D. Irwin, Inc., 132–178 (published for the American Economic Association). Alexander, S. (1920) Space, Time and Deity (2 vols), London: Macmillan. Argyrous, G. and Sethi, R. (1996) ‘The Theory of Evolution and the Evolution of Theory: Veblen’s Methodology in Contemporary Perpsective’, Cambridge Journal of Economics, 20: 475–495. Ayres, C. (1951) ‘The Co-ordinates of Institutionalism’, American Economic Review, 41: 47–55. Campbell, L. and Garnett, W. (1882) The Life of James Clerk Maxwell, London: Macmillan. Commons, J.R. (1934) Institutional Economics – Its Place in Political Economy, New York: Macmillan. Cummings, J. (1899) ‘Review of The Theory of the Leisure Class’, in Journal of Political Economy, 7: 425–455. Darwin, C. (1874) The Descent of Man and Selection in Relation to Sex (rev. edn), New York: Rand McNally. —— (1884) The Origin of Species, New York: Humboldt Publishing Company. Degler, C.N. (1991) In Search of Human Nature: The Decline and Revival of Darwinism in American Social Thought, Oxford and New York: Oxford University Press. Dobriansky, L. (1956) Veblenism: A New Critique, Washington: Public Affairs Press. Dorfman, J. (1972) Thorstein Veblen and His America (7th edn), Clifton: McKelley (1st edn, New York: Viking Press, 1935). Douglas, P. (1924) ‘The Reality of Non-Commercial Incentives in Economic Life’, in G. Rexford Tugwell (ed.), The Trend of Economics, New York: Alfred A. Knopf.

368

Francesca Lidia Viano

Dowd, D.F. (1967) ‘Veblen, Thorstein Bunde’, in Encyclopedia of Philosophy, Vol. 8, New York: Macmillan, 237–238. Dugger, W. (1979) ‘The Origins of Thorstein Veblen’s Thought’, Social Science Quarterly, 60: 424–431. Dyer, A.W. (1986) ‘Veblen on Scientific Creativity: The Influence of Charles S. Peirce’, Journal of Economic Issues, 20: 21–41; republished in Wood (1993): Vol. I, 339–357. Edgell, S. (1975) ‘Thorstein Veblen’s Theory of Evolutionary Change’, American Journal of Economics and Sociology, 34: 267–280; republished in Wood (1993): Vol. I, 195–208. —— (2001) Veblen in Perspective: His Life and Thought, Armonk, N.Y.: M.E. Sharpe. Edgell, S. and Tilman, R. (1989) ‘The Intellectual Antecedents of Thorstein Veblen: a Reappraisal’, Journal of Economics Issues, 34: 1003–1026; republished in Wood (1993): Vol. I, 439–460. Eff, A. (1989) ‘History of Thought as Ceremonial Genealogy: The Neglected Influence of Herbert Spencer on Thorstein Veblen’, Journal of Economic Issues, 23: 689–716; republished in Wood (1993): Vol. I, 413–438. Foss, N.J. (1998) ‘The Competence-Based Approach: Veblenian Ideas in the Modern Theory of the Firm’, Cambridge Journal of Economics, 22: 479–495. Harris, A.L. (1932) ‘Types of Institutionalism’, Journal of Political Economy, 40: 721–749. —— (1958) Economics and Social Reform, New York: Harper & Brothers. Hodgson, G.M. (1988) Economics and Institutions. A Manifesto for a Modern Institutional Economics, Oxford: Polity Press. —— (1993a) Economics and Evolution: Bringing Life Back Into Economics, Oxford: Polity Press. —— (1993b) ‘Thorstein Veblen and Post-Darwinian Economics’, in Hodgson (1993a). —— (1998a) ‘Introduction’, Cambridge Journal of Economics, 22: 397–401. —— (1998b) ‘On the Evolution of Thorstein Veblen’s Evolutionary Economics’, Cambridge Journal of Economics, 22: 415–431. Jensen, H. (1987) ‘The Theory of Human Nature’, Journal of Economic Issues, 21: 1039–1073. Kant, I. (1914) Kritik der Praktischen Vernunft, Berlin: Bruno Cassirer. Knight, F. (1921) Risk, Uncertainty and Profit, New York: Houghton. —— (1969) Freedom and Reform, Port Washington, N.Y.: Kennikat Press. McDougall, W. (1929) Modern Materialism and Emergent Evolution, London: Methuen. Mayhew, A. (1998) ‘On the Difficulty of Evolutionary Analysis’, Cambridge Journal of Economics, 22: 449–461. Maxwell, J.C. (1873) ‘Does the Progress of Physical Science Tend to Give Any Advantage to the Opinion of Necessity (or Determinism) Over That of the Contingency of Events and the Freedom of the Will?’, Published in Campbell and Garnett (1882): 434–444. Mitchell, W. (1936) What Veblen Taught – Selected Writings of T. Veblen, New York: Viking Press. Morgan, C.L. (1927) Emergent Evolution (2nd edn; 1st edn, 1923), London: Williams & Norgate.

Veblen revisited 369 —— (1933) The Emergence of Novelty, London: Williams & Norgate. Nelson, R.R. (1994) ‘Why Do Firms Differ; and How Does It Matter?’, in R.P. Rumelt, D.E. Schendel and D.J. Teece (eds), Fundamental Issues in Strategy, Boston: Harvard Business School Press. —— (1998) ‘The Agenda for Growth Theory: A Different Point of View’, Cambridge Journal of Economics, 22: 497–520. Nelson, R.R. and Winter, S.G. (1982) An Evolutionary Theory of Economic Change, Cambridge, Mass.: Belknap Press. Peirce, C.S. (1957) Essays in the Philosophy of Science, New York: Liberal Arts Press. —— (1966) Charles S. Peirce: Selected Writings, ed. Philip Weiner, New York: Dover. —— ([1893] 1968) ‘Evolutionary Love’, in Chance, Love, and Logic. Philosophical Essays, New York: Barnes & Noble. Penrose, E.T. (1959) The Theory of the Growth of the Firm, Oxford: Oxford University Press. Russett, C.E. (1976) Darwinism in America: The Intellectual Response 1865–1912, San Francisco: W.H. Freeman. Schneider, L. (1948) The Freudian Psychology and Veblen’s Social Theory, New York: King’s Crown Press. Sowell, T. (1987) A Conflict of Visions, New York: William Morrow. Spencer, H. (1891) ‘From Freedom to Bondage’, in A Plea for Liberty: An Argument Against Socialism and Socialistic Legislation, New York, Appleton; republished in 1981, Indianapolis: Liberty Classics, 3–32. Tilman, R. (1996) The Intellectual Legacy of Thorstein Veblen: Unresolved Issues. Westport, Conn.: Greenwood Press. Twomey, P. (1998) ‘Reviving Veblenian Economic Psychology’, Cambridge Journal of Economics, 22: 433–448. Veblen, T.B. (1884) ‘Kant’s Critique of Judgment’, Journal of Speculative Philosophy, 18; republished in Veblen (1994c). —— (1892) ‘Some Neglected Points in the Theory of Socialism’, Annals of the American Academy of Political and Social Sciences, 2: 45–62; republished in Veblen (1994b). —— (1898a) ‘The Instinct of Workmanship and the Irksomeness of Labor’, The American Journal of Sociology, IV; republished in Veblen (1994c). —— (1898b) ‘Why is Economics not an Evolutionary Science?’, Quarterly Journal of Economics, 12: 373–397; republished in Veblen (1994b) and in (1998) The Cambridge Journal of Economics, 22: 403–414. —— (1994a) The Collected Works of Thorstein Veblen, London: Routledge. —— (1994b) The Place of Science in Modern Civilisation and Other Essays, in Veblen (1994a), Vol. VIII. —— (1994c) Essays in Our Changing Order, in Veblen (1994a), Vol. X. Von Plato, J. (1994) Creating Modern Probability, Cambridge, New York and Melbourne: Cambridge University Press. Watkins, M.W. (1958) ‘Veblen’s View of Cultural Evolution’, in Dougals F. Dowd (ed.), Thorstein Veblen: A Critical Reappraisal, Ithaca: Cornell University Press. Wernerfelt, B. (1984) ‘A Resource-Based View of the Firm’, Strategic Management Journal, 5: 171–180.

370

Francesca Lidia Viano

Weyl, H. (1920) ‘Das Verhältnis der kansalen zur statistischen Betrachtungsweise in der Physik’, in K. von Chandrasekharan (ed.), Gesammelte Abhandlungen, Vol. II, Berlin: Springer. Whitehead, A.N. (1926) Science and the Modern World, Cambridge: Cambridge University Press. Wood, J.C. (1993) Thorstein Veblen. Critical Assessments (3 vols), London: Routledge.

16 Problem complexity and problem representation Some implications for the theory of economic institutions Luigi Marengo1 Abstract Bounded rationality has been embodied in models of individual and organizational behaviour mainly – if not exclusively – in terms of limitations in the (algorithmic) computational capabilities of agents. Much less attention has been given to the bounds agents encounter in their capabilities of representing the complex problems they face. Most of economic theory, even when allowing for informational and computational limitations, assumes that agents correctly represent the problems they face. In this chapter I present a simple abstract model of a class of problems (puzzles) which require the coordination of a large number of interdependent elements and discuss how different representations of the problem induce different levels of difficulty. In particular, it is shown that by acting on the representation of the problem agents can radically change its degree of decomposability into independent or quasi-independent subproblems and thus show that changes in the representation can be more powerful a problem-solving heuristic than devising more sophisticated search algorithms. Moreover, I argue that there exists a trade-off between the level of detail of the representation and the level of difficulty it implies. Finally, I discuss implications of the model for individual and collective problem-solving and, in particular, for the theory of markets and organizations, viewed as institutions which socially construct problem representations.

Introduction Theoretical and experimental research in cognitive psychology and computer science has investigated why and how the difficulty human and artificial agents face when solving a problem relates to the way they represent the problem itself. In particular, robust empirical evidence shows that human subjects can find it either very hard or very easy to solve problems which are perfect isomorphs in their abstract structure, but presented to them under different cover stories which suggest different representations.

372

Luigi Marengo

This effect can be due to different reasons: some representations can entail very different knowledge requirements in building categories, representing the search space and understanding the strategic links (DeGroot 1966), or induce changes of the size of the relevant search space (Hayes and Simon 1977), or invoke domains which are totally unfamiliar or already familiar to the subjects (Wason and Johnson-Laird 1972), or change the nature of the actions required (Kotovsky et al. 1985), or, finally, increase or decrease the amount of knowledge embedded in the environment and thus subtracted from or added to the subject’s cognitive load (Zhang and Norman 1994). This evidence is in sharp contrast with the approach of decision theoretic models adopted by standard economics, which assume that agents are – and ought to be – able to reduce all decision problems to their pure formal structure, irrespective of the context in which they are presented. In fact, if this axiom was true, mathematically isomorphic problems would be solved in exactly the same fashion. According to some scholars (cf. for instance Gigerenzer (1996)) this very same disregard of contextual features undermines also most of the research in experimental economics, including the classic studies of Kahneman and Tversky, who share with the tradition in decision theory the idea that rationality (and limitations and biases thereof) can be defined and analysed without any reference to the specificities of the context in which behaviour is observed. Following this line of reasoning we could even question whether it makes any sense to engage in the quest for context-free notion of rationality. In computer science the representation problem is known to be one of the crucial problems in evolutionary computation. Evolutionary computation, e.g. evolutionary programming (Fogel et al. 1966), genetic algorithms (Holland 1975), genetic programming (Koza 1993), uses simulated natural selection as a powerful search algorithm for complex problems: the space of possible solutions is a data structure upon which ‘genetic’ operators (such as mutation, crossover and the like) produce successive generations of ‘offspring’ solutions which undergo selection according to differential performance. Evolutionary algorithms have often proven extremely powerful in solving hard problems, but in other circumstances they fail. It has been shown that the performance of evolutionary algorithms crucially depends upon how possible solutions are represented in the search space. By acting on the representation of the possible solutions we act upon the topology of the search space and can make it very easy or very hard to be searched by mutation, crossover and other genetic operators (Jones 1995). The representation problem has also recently attracted the attention of biologists. Wagner and Altenberg (1996) observe the very close similarity between the representation problem in evolutionary computation and the problem of evolvability in natural evolution; that is, the capability of generating new viable and fitter solutions out of blind and random mutations. One important aspect of the complexity of a problem which relates to

Problem complexity and representation 373 the way it is represented concerns the possibility of decomposing the problem itself into sub-problems which can be solved independently or quasi independently: in this way very large search spaces can be transformed into a collection of smaller and more manageable subspaces. The importance of this question has been early recognized in the study of human and machine problem-solving (cf. for instance the classical analysis in Newell and Simon (1972) and Simon (1981)). Sub-problem decomposition is perhaps the most fundamental device for rendering difficult problems manageable by computationally limited human or artificial problem-solvers, but there are serious bounds to the extent to which subproblem decomposition can be effective: interdependencies. When, in decomposing a problem, interdependent elements are separated, then sub-problem decomposition brings inevitably sub-optimality, lock-in and path-dependence. A fundamental dilemma emerges: on the one hand decomposing is necessary to make difficult problems tractable, but on the other hand incorrect decompositions greatly reduce the likelihood to discover optimal or even good solutions. Finally, boundedly rational agents are almost inevitably constrained to use incorrect decompositions, since finding the ‘right’ decomposition of a problem is computationally more complex than optimally solving the problem itself, as the space of possible decompositions is obviously vaster than the space of possible solutions. Also from a psychological perspective there is wide evidence that human beings encounter very strong cognitive limitations in correctly analysing all the interdependencies of a problem. Interdependencies cause a combinatorial explosion of the search space of a problem and our mind has very limited power in dealing with such large search spaces, if anything because of the high short-term memory requirements such spaces imply. But this question also has an important counterpart which relates to the division of labour and the design of social and economic institutions: can complex economic problems, such as designing and producing new goods and institutions, be handled by dividing them into subtasks, and can the latter be co-ordinated through anonymous market transactions or are different forms of coordination (authority for instance) needed? The model presented in the next sections addresses these issues and suggests a formal way to characterize these dilemmas precisely. The chapter is organized as follows: the second section discusses how interdependencies constitute a source of difficulty for decentralized problemsolving; the next two sections define, respectively, a class of problems with high interdependencies, which we call puzzles, and show that the difficulty implied by such interdependencies varies according to how the puzzles are framed and represented. The final section draws some implications and conclusions for the theory of individual problem-solving and for the role of economic and social institutions as devices for collective problemsolving.

374

Luigi Marengo

Interdependencies and difficulty If not the only one, interdependencies certainly constitute a major source of difficulty in human problem-solving because local processes of search in one dimension propagate their effects over other parts of the system in ways which are very hard to understand, foresee and control. If interdependencies are widespread, only global search procedures which simultaneously control all the dimensions of the problem are fully effective, while local search procedures get quickly stuck into a local optimum. Unfortunately human beings find it very hard to control for more than one or very few dimensions at a time and are typically bound to use search strategies which are very similar to local trial-and-error and implicitly assume complete separability among dimensions by varying only one (or very few) of them in order to find better solutions (Bruner et al. 1956). The sheer combinatorial problem implied by interdependencies makes them intractable by our mind, whose short memory resources are well known to be highly limited. Local trial-and-error search is, at least as a first approximation, analogous to one-bit mutation algorithms widely studied in biology and evolutionary computation. The limits of one-bit-mutation algorithms in multi-dimensional and highly interdependent problems are well described by Stuart Kauffman’s NK model (Kauffman 1993) which analyses general statistical properties of fitness landscapes, i.e. systems made of N components, each of which can be in a finite number of alternative states, and whereby every combination among the N components is attributed an exogenously determined fitness value (which in our case could be a measure of the ‘goodness’ of a solution to the problem). Kauffman examines the properties2 of one-bit-mutation search processes over classes of randomly generated fitness landscapes and finds that the power of one-bit mutation is limited by the degree of interdependency (‘epistasis’) among components, measured by a parameter K. When the latter is minimum (that is, components are not interdependent and the system is fully decomposable), then one-bit mutations are always able to locate the global optimum. This case is depicted for a landscape of dimension 3 in Figure 16.1, where each vertex of the hypercube represents a binary string with its fitness value in brackets. It can be seen that every other vertex of the hypercube is linked to the one with the highest fitness by a path in which fitness never decreases. Thus one-bit mutation algorithms can always find the global optimum by following such a path. If, on the contrary, K approaches N (that is, the system is characterized by diffused interdependencies), then multiple local optima exist and, depending upon the initial conditions, one-bit mutation search processes can end up in a local optimum. This is the case depicted in Figure 16.2 where, for instance, one-bit mutation starting from 001 or 011 will always

010 (0.63)

110 (0.80)

011 (0.53)

111 (0.70)

000 (0.53)

100 (0.70)

101 (0.60)

001 (0.43)

Figure 16.1 Fitness landscape with independent components

010 (0.70)

110 (0.63)

011 (0.53)

111 (0.43)

000 (0.47)

001 (0.50)

100 (0.83)

101 (0.40)

Figure 16.2 Fitness landscape with interdependent components

376

Luigi Marengo

end up in the inferior local optimum 010 and never locate the global optimum 100. As already mentioned, Kauffman’s results concern statistical properties of randomly generated landscapes, but it may well happen that a particular structure of interdependencies may make the landscape optimally searched by one-bit mutation, or other simple local trial-and-error algorithms, in spite of having a very high degree of interdependency among the constituent components. In the next section we will show that this is the case for a very important class of problems, which we call ‘puzzles’, which are characterized by strong and diffused interdependencies, but of a special architecture – a ‘nested’ one – which still makes one-bit mutation always able to locate global optima. However, we also show that when search is based upon an incomplete representation of the problem, then local optima can be introduced which are not the product of the underlying structure of the problem, but of the representational biases of the problem-solving agents.

Puzzles and their solutions The particular class of problems we are going to analyse is represented by puzzles, i.e. one-player games which can be expressed in the form of deterministic transitions between a finite set of states (for a more extensive treatment of these puzzles, and for some graph theoretic results which are equivalent to the ones presented here, see Egidi (2002)). States are characterized by a full description of the configuration of the game: layout of cards in a solitaire (including those which are not visible to the player), configuration of a Rubik cube, position of disks on the pegs of the Tower of Hanoi, etc. We assume that the game does not involve any random component in the transitions: it might indeed present hidden information and randomness with respect to the initial configuration, as usually happens in card games, but a given move applied to a given configuration must always cause the same transition. Definitions Puzzles can be described by the following elements: 1

2

A finite set of states S 傼 W where S  {s1, s2, . . . , sn} is the set of nonterminal states of the game and W  {w1, w2, . . . , wp} is the set of terminal or ‘winning’ states, i.e. those in which the puzzle is successfully solved and no further moves should be performed. Each state si or wi is a complete description of all the relevant aspects of the puzzle at that stage. A non-empty subset B 債 S of the possible initial states, i.e. configurations which can appear as starting ones. In order to be as general as

Problem complexity and representation 377

3

4

possible we will assume in the rest of the chapter that B  S, i.e. that the puzzle can begin from every non-terminal state. Without loss of generality we also consider only puzzles in which W can be reached from every state in B. A finite set of moves M  {m1, m2, . . . , mk}, i.e. possible (given the physical constraints) and legal (given the rules of the game) actions which determine transitions between states. A transition matrix T  [tij] with tij 僆 S 傼 W for i  1, 2..n and j  1, 2..k where tij is the state which is reached when move i is performed at state j. Some moves can be illegal at some states (i.e. violate the rules of the game), thus the matrix T may contain empty cells. Alternatively, illegal moves can be modelled by introducing an ‘illegal state’ s0 which is reached from any other state when an illegal move is performed and which terminates the game (no further moves are allowed).

Playing algorithms A solution for a puzzle can be thought as a set of moves, one for each state which is encountered when playing the game. In order to keep as general as possible we will assume that every non-terminal state can be a starting state and therefore every non-terminal state can certainly be encountered. Thus, a complete and deterministic playing program P is a string of length n which specifies one and only one move for all the non-winning states: P  ms1ms2 . . . msi . . . msn with msj 僆 M We call ∏ the space of the Kn complete programs. Goven a starting state si 僆 S, the program P determines a unique sequence of states which can end up in: 1

2 3

one of the winning states wj – in this case we also compute the length of the sequence of moves which the program has performed to reach wj; the illegal state s0 if an illegal move has been performed by the program; an infinite loop if the sequence twice visits a state which is neither winning nor illegal.

We suppose that complete programs can be completely ordered according to some measure of their performance. Given any two programs Pi and Pj we assume that we can always state whether Pi Ɑ Pj or Pj Ɑ Pi or Pi ⬇ Pj if Pi is respectively more, less or equally performing than Pj. We also write PiⱰ Pj when Pi is weakly preferred to Pj.3

378

Luigi Marengo

We call the pair (∏, Ɒ), i.e. the set of complete programs with its order relation, the landscape of complete programs. In all but the most trivial puzzles the number n of states is very large,4 therefore agents cannot rely upon a complete playing program, i.e. programs which specify one move for each state. Real agents will explicitly or implicitly rely upon compound representations and incomplete playing programs. Limiting ourselves, for the sake of simplicity, to partitional representations, we call a representation of the set S a partition:

 {1, 2, . . . , h} with i 債 S where h

i  S and i 傽 j  ⵰ i  j 艛 i1 If is formed only by the singletons {s1}, {s2}, . . . , {sn} we say that the agent has a complete representation of the puzzle, otherwise we say that the representation is incomplete. When using an incomplete partitional representation  {1, 2, . . . , h} the agent does not make a distinction between, and therefore treats as equivalent, two states si and sj if they belong to the same class k and therefore the search space is limited to programs which prescribe the same move for both si and sj. An incomplete program is a string of length h  n: P( )  m1m2 . . . mn This reduces the size of the search space ∏( ) 債 ∏ but may well imply that the optimal program(s) does (do) not belong to the space ∏( ) and thus is (are) not attainable. This trade-off between the size of the representation and the optimality of the attainable solutions is discussed in detail in the next section.

Representations and difficulty Compacting representations We will make the hypothesis that agents can move on the landscape of programs only by charging one move at a time, i.e. mutate only one bit of the current program P. This hypothesis may seem rather extreme, but reflects evidence suggesting that human search processes are essentially local: as shown for instance in Bruner et al. (1956) human subjects engaged in search processes over multidimensional spaces usually vary only one or

Problem complexity and representation 379 very few dimensions at a time and therefore are easily locked into local optima if the system presents widespread interdependencies. However, we show in the following that this strictly local nature of search processes is not a limitation per se but only insofar as agents hold incomplete representations. Proposition 1 in fact shows that the landscape of complete programs does not have local optima with respect to one-bit mutation processes; that is to say, that one-bit mutation (i.e. changing only one dimension at a time) can always locate a globally optimum program, provided that agents operate on the space of complete programs, i.e. they have a complete representation of the states of the puzzle. But of course complete representations of even trivial puzzles cannot be held by human beings whose memory bounds are largely overcome by the combinatorial nature of the search space. Human subjects are forced to employ representational heuristics which compress such a space. Typically, in fact, heuristics employed by human subjects explicitly or implicitly define broad classes of states which are considered as equivalent, in spite of their differences, because they share some common key features. For instance, in card games only some cards with higher value are paid attention to; in the Rubik cube only one side is looked at at a time; in the Tower of Hanoi agents try first of all to move the larger disk on the rightmost peg, etc. Proposition 2 states that when search takes place on landscapes of incomplete programs (i.e. generated by incomplete representations) then it is generally subject to lock-in into a sub-optimal solution. Propositions 1 and 2 together describe a trade-off between the sophistication of search algorithms on one side and the completeness of the representation: representations which are detailed enough can support very simple trial-and-error search heuristics, while coarser representations need more sophisticated search heuristics which have to control for interdependencies. In other words we could say that puzzles are inherently simple (i.e. they can be optimally solved by one-bit mutation) and that complexity is introduced by the representational limitations of our mind. We first show that one-bit mutation on the landscape of complete programs can always locate a globally optimum program, i.e. that such a landscape is always fully decomposable. This result amounts to saying that optimal solutions of any puzzle can be found by a totally decentralized trial-and-error search process, i.e. that puzzles can always be decomposed into sub-problems of minimum size which can be solved independently from each other. In order to show this result we have to introduce some preliminary notions (more details can be found in Marengo and Dosi (2003)). Given two complete programs P and P we say that P is a ‘one-bit preferred neighbour’ or simply a ‘neighbour’ of P if P Ɒ P and P differs from P by one and only move.

380

Luigi Marengo

Given a complete program P0 we define a (one-bit) path from P0, denoted by (P 0) a sequence of complete programs P1, P2, P3, . . . such that Pi is a neighbour of Pi1. Given two complete programs P0 and P we say that P is (one-bit) reachable from P0 if there exists a path (P0) such that P 僆 (P0). Let us now characterize the optimal program(s). In order to find optimal program(s) we can partition the set S into subsets of states for which sequences of moves exist which connect them to W by no less than 1, 2, 3, . . . moves. Such a partition can be built recursively: 1 2 3

let G1 be the set of the states si 僆 S such that there exists at least one move mh such that tih 僆 W set h  1 repeat •

hh
1

•

let Gh be the set of states si 僆 S \ 艛 Gk such that there exists at

冦

h1 k1

冧

least one move mh such that tih 僆 Gh1 Sk冣  ⵰ 冢艛 k1 h

until S \

Proposition 1 The optimal program(s) is (are) reachable from every other complete program P 僆 ∏. Proof: Let us suppose that there is a unique optimal program P*  m*s1m*s2 . . . m*si . . . m*sn. Now consider a generic program P  mi1mi2m . . . min and a state sj 僆 G1. By construction a move m*ij must exist which connects sj directly to W. Now either P already has m*ij in the j  th position or it does not. In the latter case the program P which is identical to P except for having m*ij in the j  th position can be reached from P with one mutation and has certainly higher performance than P. If instead P has already m*ij in the j  th position for all the states sj 僆 G1, we can repeat a similar argument for all the states in G2, and so on. Proposition 1 shows that when working with a complete representation (i.e. one which distinguishes among all the states) puzzles are always ‘simple’, in the sense that one-bit mutation is always able to locate a global optimum. Note that this does not imply that there are no interdependencies in the space of complete programs: actually it turns out that interdependencies do exist and can be very strong but they have a rather special structure, i.e. a ‘nested’ one.5 In fact, moves relative to the states in G1 are independent from all the others, moves relative to states in G2 are interdependent with one move in G1, those relative to states in Gi are linked with those in Gi1, Gi2, . . . , G1.

Problem complexity and representation 381 The next proposition, on the contrary, shows that obviously if representations are incomplete and therefore agents are searching the space of incomplete programs ∏( ) then the optimal program(s) is (are) not always reachable from every initial condition. Proposition 2 The optimal program(s) is (are) not always reachable from every incomplete program P 僆 ∏( ). Proof: The proposition is trivially true when the optimal program does not belong to ∏( ): this happens when the optimal program P* prescribes different moves for some states which belong to the same class in the partition . It is easy to compute which local optima are reachable for a given representation and, vice versa, classes of representations which are compatible with observed local optima. In the appendix we provide some examples based on a well-known and widely studied puzzle known as the Tower of Hanoi. Expanding representations Let us now make a thought experiment: we showed in the previous section that the search space of puzzles does not present local optima for simple local trial-and-error search processes if search is based upon a complete representation (i.e. if agents are able to distinguish and treat independently every state of the puzzle). Then we showed that local optima can be a product of the incompleteness of representations. Thus, in general, we could draw the implication that every local optimum is in a sense the product of a coarse representation of the underlying search space. By building finer and finer representation we should be able to avoid any kind of lock-in into local optima. Considering landscapes of puzzle-solving programs helps us to understand clearly where such local optima come from and how they can be avoided by properly refining a representation. Local (and not global) optima are due to a representation which does not distinguish between two (or more) states which in the optimal strategy require two (or more) different moves. Thus by adding one (or more) new states to the current representation they can be eliminated. Let us explain this point by means of an example. Consider the landscape of dimension 3 (Table 16.1), where strings are ranked from the most preferred (rank 1) to the least preferred (rank 8). Suppose that this landscape represents the ranking of possible solving programs of a puzzle with three non-winning states (s1, s2, s3) and two moves (labelled 0 and 1). Consider for instance the first bit: we can read that in the optimum program at state s1 move 1 should be performed. However, we also read that switching from move 1 to move 0 at state s1 does not always improve the overall

382

Luigi Marengo Table 16.1 A landscape with local optima Strings

Ranking

100 010 110 011 001 000 111 101

1 2 3 4 5 6 7 8

performance of the program: for instance, it worsens the performance when move 1 is performed at s2 and move 0 at s3. The program 010 is in fact a local optimum into which (depending on the initial condition) search processes can be locked into.6 It is easy to check that we cannot find any puzzle whose complete representation can generate a landscape of complete programs like the one of Table 16.1; in fact, we cannot write a unique transition matrix T which can generate this ranking of programs.7 Thus the landscape of Table 16.1 cannot be a landscape of complete programs and must necessarily be one of incomplete programs, which is based on an incomplete representation  {1, 2, 3} whereby 1 is certainly not a singleton. Since only two moves are possible, we can divide 1 into two separate states, 1A and 1B, whereby the former covers all the cases in which for the first bit move 1 dominates move 0, whereas 1B covers all cases in which the opposite happens. Thus we derive from the landscape of Table 16.1 a new landscape of dimension 4, built upon the representation  {1A, 1B, 2, 3}, which is fully decomposable and can be optimally solved through one-bit mutation. The derived landscape is presented in Table 16.2.

Some implications for individual and collective problemsolving The simple model presented so far has some obvious testable implications for individual problem-solving: a precise framework which maps representations into degrees of difficulty and states that any kind of problem can be made as difficult or as easy as one wishes by properly modifying its representation has been outlined. In future work it is planned to use the framework to help design rigorous experiments in which, for a given puzzle, keeping invariant its formal structure in terms of states, moves and transition matrix, subjects can be induced to adopt different representations by properly manipulating the way it is presented to subjects, and check whether the optimal or suboptimal solutions subjects tend to find

Problem complexity and representation 383 Table 16.2 A derived landscape with local optima Strings

Ranking

1100 1000 1010 0010 1110 0110 1011 0011 1001 0001 0100 0000 1111 0111 1101 0101

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

are consistent with the predictions of our model. The appendix briefly outlines how a puzzle widely used in experimental research in cognitive psychology can be rendered in our framework. But our framework can be used the other way round also to derive some properties of the representations human subjects actually adopt when solving problems. Representations are clearly unobservable and their features are largely opaque, as they are to the subjects who actually hold them, but our framework allows the deduction of some formal features of such representations by observing which local optimum agents tend to be locked into when solving a given problem. Perhaps more interestingly, our model also addresses important questions related to collective problem-solving and to the theory of social and economic institutions. In this chapter it has been shown that if agents base their search upon representations which are detailed enough (i.e. they approach the one-to-one mapping to problem states) they can find optimal solutions by means of the simplest local trial-and-error search heuristics. It has been acknowledged that human beings, because of their memory limitations, cannot possibly hold complete representations of any problem but the most trivial ones. However, an obvious observation is that such representations need not be held in the mind of a single individual: representations are in fact typically distributed among a multiplicity of individuals as well as social norms, rules, conventions, physical artefacts, and so on. We have here a clear picture of the long-debated role of economic and social institutions as devices for collective problem-solving, and in particular our framework suggests that their primary role is that of providing ways of collectively sharing and

384

Luigi Marengo

distributing representations in order to circumvent the limitations of individual representational capabilities. A framework like the one outlined in this chapter can be used to address, also experimentally, questions about the problem-solving effectiveness of different institutional arrangements differing in their division of knowledge, rules of interaction, physical objects involved, and so on. An interesting aspect which emerges from the analysis here concerns the importance of the ‘division of representational labour’: in the framework presented here, when search is based upon a detailed enough representation, then problem-solving can be carried out by local processes of trial-and-error co-ordinated only through exogenous efficiency signals. The latter mechanism is closely related to a decentralized market mechanism, and the results presented in the previous section can be viewed as a sort of extension – and qualification – of the Coase theorem (Coase 1960) to problem-solving situations, the claim being that the more correct a representation of the problem is the more we can rely on local search co-ordinated by an anonymous selection mechanism. However, it also argues that any rationally and computationally limited entity cannot be supposed to hold correct representations of non-trivial problems whose optimal solution is not known, as finding a correct representation is computationally more difficult than solving the problem. Note that if the representation is not the correct one then totally decentralized local search is inefficient, irrespective of the existence of frictions, uncertainty and other sources of transaction costs. Coase himself pointed out (Coase 1988) that his so-called theorem holds only in an ideal and unreal frictionless world: in the problem-solving environment presented here it can be added that it holds only in an equally ideal and unreal world in which cognitive and representational problems have been optimally solved.

Appendix The Tower of Hanoi This appendix illustrates the main propositions of the chapter by means of a well-known example: the Tower of Hanoi (TOH henceforth). TOH’s playing board is made of three fixed pegs on which three disks of different sizes can be moved. The winning position is the one depicted in Figure 16.A1, while the starting positions can be, in general, any other legal configuration. The puzzle has two rules: 1 2

only one disk can be moved at a time (therefore if there are many disks on a peg only the one on the top can be moved); only configurations in which no disk is on the top of a smaller one are

Problem complexity and representation 385

Figure 16.A1 Tower of Hanoi: the winning configuration

Figure 16.A2 Tower of Hanoi: two starting configurations

legal (therefore a disk can be moved onto a peg only if it will be the smallest on that peg). Given the two rules there are 27 legal states. Table 16.A1 lists all the states with the entire transition matrix. When more than one disk is on a single peg, we indicate on the leftmost position the one at the bottom of the stack. Moreover, we put an X in the transition matrix when the corresponding move is impossible (because the peg from which a disk should be moved is empty) and s0 when it is illegal (it violates rule 2 above). The state s27 is the winning one. The optimal complete program is the following (we use the abbreviation LC to indicate the move ‘left to centre’, and so on): LR, CR, LC, CL, RC, RL, RC, LC, LR, RC, CL, LC, RL, . . . . . . , RL, CR, LR, CL, RC, LR, CR, CL, RL, CR, LC, LR, CR This optimal complete program is always found via one-bit mutations from every starting complete program with a small (linear in the number of states) number of mutations. Suppose now that we are using a representation with five classes:



 1  {s1, s2, s3, s4, s5, s6}   2  {s7, s8, s9, s11, s13, s15}   3  {s10, s12, s14, s16, s17, s20}   4  {s18, s19, s21, s22, s23, s24}   5  {s25, s26} 

386

Luigi Marengo

Table 16.A1 States and transitions of the Tower of Hanoi States

Left peg

s1 s2 s3 s4 s5 s6 s7 s8 s9 s10 s11 s12 s13 s14 s15 s16 s17 s18 s19 s20 s21 s22 s23 s24 s25 s26 s27

L,M,S L,M L,M L L L L,S L,S L

S

S S M M S

M M,S M,S M

Centre peg

Right peg

S S M

M M,S L M,S L L,S M M M,L L,S

S M S M,S M

M,S L M M L,M L

L,S L,M L,M,S L,M L L

S S

S S

L L L,M L,M,S

Left to centre

Left to right

Centre to left

Centre to right

Right to left

Right to centre

s2 s0 s5 s0 s0 s10 s4 s9 s0 X X s13 X X s11 s20 s0 s14 s26 X X s21 s17 s25 s0 X

s3 s4 s0 s0 s0 s0 s6 s5 s11 X X s10 X X s14 s21 s13 s0 s27 X X s0 s22 s18 s26 X

X s1 X s7 s3 X X s0 s8 s6 s15 s0 s12 s18 s0 s0 s23 X X s16 s22 s0 s0 X s24 s19

X s3 X s6 s0 X X s7 s5 s0 s14 s0 s10 s0 s19 s12 s22 X X s21 s0 s0 s24 X s18 s27

X X s1 s2 s8 s7 s0 X X s12 s9 s0 s17 s15 s0 X X s24 s0 X s16 s23 X s0 s0 s25

X X s2 s0 s9 s4 s8 X X s13 s0 s16 s0 s11 s0 X X s25 s15 X s20 s17 X s23 s0 s0

Now the search space consists of programs of only five moves (one for each class i) and search is therefore much faster, but one-bit mutation search processes are now locked into a sub-optimal solution which, expressed in the five classes, is either: CL, CR, RL, LC, CR or (with the same performance): LR, CR, RL, LC, CR

Problem complexity and representation 387

Notes 1

2 3

The work presented in this chapter has been greatly inspired by long discussions with Massimo Egidi: I thank him for his insights and suggestions. I am also indebted to Giovanni Dosi, Koen Frenken, Paolo Legrenzi, Scott Page, Corrado Pasquali, Herbert Simon and Marco Valente, who, at different stages of this work, provided suggestions and criticisms. The usual caveat applies. This research is part of the project NORMEC (SERD-2000-00316), funded by the European Commission, Research Directorate, 5th framework program. The author alone is responsible for the opinions expressed, which are not necessarily those of the European Union or the European Commission. See also Frenken et al. (1999) for a more detailed presentation of the NK model and for a critical evaluation of its relevance for the study of problem difficulty. A natural measure of performance which establishes such an order relation is the following. We define for each program Pi the pair: (APi, BPi) 

4 5

6

7

冢 冱 Nfail , 冱 Nmoves 冣 si僆B

si

si僆B

si

where the first number is the sum of illegal moves and infinite loops performed by a program Pi and the second is the sum of the number of moves made by Pi to reach a winning state whenever neither infinite loops nor illegal moves are encountered. Then we can use lexicographic orders on these pairs and say that Pi Ɑ Pj if APi APj or, if APi  APj, then BPi BPj. Even a puzzle as simple as the three disks Tower of Hanoi, which is presented in the appendix, has 27 states, a number which far exceeds our short-term memory capacity. Using Kauffman’s terminology, puzzles can have maximum K (i.e. the highest level of interdependency) and nevertheless be optimally solved by one-bit mutation. If instead we use the difficulty measures presented in Page (1996) we can say that puzzles have a large cover size and ascent size equal to 1. Note that the worst local optimum 010 has a larger basin of attraction as it covers all possible programs except the global optimum itself. Thus, only a search which starts at the global optimum will (trivially) stop at the global optimum itself with certainty, while for four initial configurations search might end up in either local optima (depending on the sequence of mutations) and for the remaining three initial configurations search will end up at 010 with certainty. Multiple conditional transition matrices would be required, in which the state which is reached from state s1 when move 1 is performed depends upon the path followed to reach s1 itself.

References Bruner, J.S., Goodnow, J. and Austin, G. (1956) A Study of Thinking, New York: Wiley. Coase, R.H. (1960) ‘The Problem of Social Cost’, Journal of Law and Economics, 3: 1–44. —— (1988) ‘The Nature of the Firm, 1. Origin, 2. Meaning, 3. Influence’, Journal of Law, Economics and Organization, 4: 3–47. DeGroot, A.D. (1966) ‘Perception and Memory Versus Thought’, in B. Kleinmuntz (ed.), Problem-solving, New York: Wiley.

388

Luigi Marengo

Egidi, M. (2002) ‘Biases in Organizational Behaviour’, in M. Augier and J.G. March (eds), The Economics of Choice, Change and Organization: Essays in Memory of Richard M. Cyert, Cheltenham: Edward Elgar. Fogel, L.J., Owens, A.J. and Walsh, M.J. (1966) Artificial Intelligence Through Simulated Evolution, New York: Wiley. Frenken, K., Marengo, L. and Valente, M. (1999) ‘Interdependencies, NearlyDecomposability and Adaptation’, in T. Brenner (ed.), Computational Techniques for Modelling Learning in Economics, Dordrecht: Kluwer Academic Publishers. Gigerenzer, G. (1996) ‘Rationality: Why Social Context Matters’, in P.B. Baltes and U.M. Staudinger (eds), Interactive Minds, Cambridge: Cambridge University Press. Hayes, J.R. and Simon, H.A. (1977) ‘Psychological Differences Among Problem Isomorphs’, in N.J. Castellan, D.B. Pisoni and G.R. Potts (eds), Cognitive Theory, Hillsdale, N.J.: Erlbaum. Holland, J.H. (1975) Adaptation in Natural and Artificial Systems, Ann Arbor: University of Michigan Press. Jones, T. (1995) ‘Evolutionary Algorithms, Fitness Landscapes and Search’, Ph.D. thesis, Albuquerque, N.Mex.: University of New Mexico. Kauffman, S.A. (1993) The Origins of Order, Oxford: Oxford University Press. Kotovsky, K., Hayes, J.R. and Simon, H.A. (1985) ‘Why Are Some Problems Hard? Evidence from Tower of Hanoi’, Cognitive Psychology, 22: 143–183. Koza, J.R. (1993) Genetic Programming: On the Programming of Computers by Means of Natural Selection, Cambridge, Mass.: The MIT Press. Marengo, L. and Dosi, G. (2003) ‘Division of Labor, Organizational Coordination and Market Mechanisms in Collective Problem Solving’, Pisa, St Anna School for Advanced Studies, Working Paper LEM 2003/04. Newell, A. and Simon, H.A. (1972) Human Problem Solving, Englewood Cliffs, N.J.: Prentice-Hall. Page, S.E. (1996) ‘Two Measures of Difficulty’, Economic Theory, 8: 321–346. Simon, H.A. (1981) The Sciences of the Artificial (2nd edn), Cambridge, Mass.: The MIT Press. Wagner, G.P. and Altenberg, L. (1996) ‘Complex Adaptations and the Evolution of Evolvability’, Evolution, 50: 967–976. Wason, R.A. and Johnson-Laird, P.N. (1972) Psychology of Reasoning, Cambridge, Mass.: Harvard University Press. Zhang, J. and Norman, D.A. (1994) ‘Representations in Distributed Cognitive Tasks’, Cognitive Science, 18: 87–122.

Index

Page numbers followed by n indicate endnotes; those followed by f indicate figures. abduction, exploration 218, 219, 225 Abramowitz, Moses 356, 367n abstraction 226; and classification 33–8, 48 adaptation 364n; and resistance to change 6; and theory of evolution 348 adaptive toolbox, evolutionary psychology 67 Administrative Behavior (Simon) 138–9, 140, 141, 148, 153n, 168 ‘administrative man’ 158 affective: and cognitive 71–2, 77n, see also motivation; sentiment; sympathy agency: equilibrium theory 320–1; Hayek on agent’s heterogeneity and knowledge 326–32; Veblen on inertia and active subjects 346–52, 366n, see also individuals agglomeration effect 286 airline industry: cycles 232; innovation 197; technological knowledge 272–3 Alchian, Armen 3, 14n, 146, 186 algorithms 72–3, 77n; evolutionary algorithms 372; problem-solving 377–8 Allais, Maurice 2, 20, 85, 96 Altenberg, L. 372 altruism, evolutionary psychology 64, 69–70 American Hospital Supplies (AHS) 239 Anderson, P. 224, 232 Andreoni, J. 104n Annals of Mathematics, The (Wiles) 31 anthropology 361n Antonelli, Cristiano 268–79, 292 Aoki, M. 235 Arena, Richard 316–36

Arrow, Kenneth J. 139, 146, 167, 207, 302 Arrowian paradox of information 172 Arthur Andersen 223, 238 Arthur, Brian 2, 10, 201 ‘as if ’ assumption 22, 23 ATT, innovation 198 Augier, Mie 133–52 Aumann, R.J. 64, 247, 251–2, 260 authority: and controversies 29–32; as defect in our system of knowledge 31–2; Simon on 141 autopoiesis 14n Axelrod, R. 59, 105 axioms, and disagreements in scientific community 30–1 Bacdayan, P. 38, 84 Bach, George Lee 145 Bacon, Francis 28–9 Bagolini, L. 126n Bandura, A. 284 Baumol, W. 167 Bausor, R. 208 Bayesian method 247, 305 Beccaria, Cesare 127n ‘Behavioral Model of Rational Choice, A’ (Simon) 140–1 behaviour: deviant behaviour 284; game theory, ‘accepted standard of behaviour’ and ‘established social order’ 252–5; knowledge and social learning 180–1; learning models as behavioural 312; optimal behaviour, decision theory 21–5 behavioural economics, Simon’s development of 145–51

390

Index

Behavioural Theory of the Firm (Cyert and March) 149 beliefs: belief co-ordination, social harmony 121; belief spaces, co-ordination 120; belief structures, and co-ordination 107, 120; communication beliefs and moral sense 125; cultural beliefs, translating 119 Bell Laboratories, innovation 198 Bellandi, M. 285–6 Ben-Ner, A. 55 Benetton 233, 237 Bennis, W.G. 237 Bettis, R.A. 214n Binmore, K. 247 biotic, and social science 360–1n Birner, Jack 326 Boas, F. 356 Boettke, P. 316 Boone, C. 97 Borel, E. 244, 245, 248–9, 253 Börgers, T. 313n Boulding, Kenneth 2, 6, 14n, 146 bounded rationality 22, 50n, 179, 283, 371; cognitive elements 153n; contract theory 171, 173n; game theory 152n, 173n; incomplete contract controversy 162–4; and irrationality 68; and learning 305–7, 310; as negatively defined 168; organizational economics 160–1, 164–5, 170–2; and organizational structure 148; renamed ‘procedural rationality’ 168; rhetorical dimensions 158–73; and routine 169; Simon’s development of concept 133–52; Simon’s Ely and Nobel lectures 165–9; simplified model 153–4n; social knowledge 108–9; surprise 109; transaction cost economics (TCE) 54, 65, 66, 161–2, 171–2, see also organizations; procedural rationality brain, path-dependence 10 Brandenburger, A. 251 bridge building, innovation 200 Bruner, J.S. 378 business conceptions: cognitive leadership 183–4; communication 183; competitive genesis of the firm 184–6; and division of labour 181–4, 190; nature of the firm 181–4 business cycles, Hayek on 326–9

Buss, D.M. 60, 62 Butler, Joseph 125 Caldwell, B. 15n, 332 Camerer, C. 171 Campbell, D.T. 206 capability: capabilities approach, knowledge 207–11; as relative 212 capitalism: innovation 203; invisible hand 297n; as process of creative destruction 193 car industry, technological knowledge 273–4 Carnap, Rudolph 138, 153n Carnegie Mellon University, Simon’s development of behavioural economics 144–51 Carter, C. 194, 195 Cartesian doctrine 28 causal analysis 344–6, 360n, 363n cement industry, cycles 232 Center for Adaptive Behavior and Cognition (ABC) 67 Central Book House (Netherlands) 236 change: explanations for 5; organizational and institutional 11–12; resistance to 6–7, 10–11, 15n, see also habit; inertia; innovation cheating, evolutionary psychology 60 chemicals 233–4 chess, theory of problem-solving 22–3, 25 Clark, M. 198 classification, problem-solving 33–8, 48 Clausius, R.J.E. 344 co-operation: evolutionary psychology 58–9, 76n; Moulin’s modes of co-operation 259–61 co-ordination (social): beliefs 107, 120, 121; co-ordination games 108, 126n; commensurability and diversity 119–20; game theory, agreement and co-ordination 262–5; language 116–17; and rationality 333; and sentiment 120–3, 125; shared cognitive practices 110–11; social knowledge 107–10; sympathy 122, 126n, see also social knowledge Coase, R. 281, 357, 384 Coasian theory of the firm 178, 281 cognition: cognitive and affective 71–2, 77n; cognitive dissonance 14n; and evolution 3–5; instrumental

Index 391 rationality and cognitive rationality 304; social learning and cognitive processes 178–81; theories and ideologies as cognitive traps 20–50 cognitive experimental economics 82–104; and cognitive economics 83–6; individual learning, experiment 87–94; integrating experiments and simulations 101–3; team learning, experiment 95–101 cognitive leadership: business conceptions 183–4; cognitive leadership regime and organizational development 187–9; and governance 234 cognitive psychology, and transaction cost economics 172 Cohen, M.D. 38, 84, 85 collective knowing 292–4 commensurability, and diversity 119 Commons, John 361n communication: business conceptions 183; co-ordination games 108, 126n; cognitive processes 179–80; and collective knowing 292; communication beliefs and moral sense 125; imagination and sympathy 111–12; knowledge channels 276–7; and learning 284, 311, see also coordination; language comparison: categories and abstractions 33–8; cognitive traps 38–42 ‘Comparison of Organization Theories, A’ (Simon) 142 competition: creating demand 12; evolutionary approach 3; exploitation and exploration 231; ‘as if’ assumption 22; and rationality 333; Veblen on 348 competitiveness: and capabilities 207; competitive advantage as transient 212–13; competitive genesis of the firm 184–6 complexity theory 172 computer industry: cycles 232; economy of scale 233 conceptual spaces, diversity 117–20 consolidation 225; and cycle of integration and disintegration 229–31; danger of inertia 227 Constant, E.W. 213 constructivism, intelligence 220 context, theory of the firm 285–9, 293

Continental school, knowledge 28 contract theory, and bounded rationality 171, 173n controversies: and authority 29–32; incomplete contract controversy, bounded rationality 162–4; settling of within scientific community 30–1 Cooper, A.C. 197, 198 Cooper, William 145 Copp, N.H. 200 Cosmides, Leda 58–62, 66, 67, 73–4, 76n, 356 Cowles Commission 139–40, 142, 143, 153n, 154n creativity: creative destruction, capitalism as process of 193; and understanding 105, see also imagination; innovation creodic processes, and positive feedback 14n crises, theory of economic crises 322 critical thinking 50 Critique of Judgement (Kant) 341–2 cultural beliefs, translating 119 culture: exaptation 12; organizational culture 220; translating cultural beliefs 119 cumulative prospect theory 171 custom 116–17 Cusumano, M.A. 196 cybernetic systems, economic systems as 6 cycles 232; discovery process 226f, 232–4; Hayek on business cycles 326–9; innovation, cycle theory 222–5; integration and disintegration, cycle 229–31, see also equilibrium theory Cyert, Richard 146, 149, 167 Darnanoni, V. 84 Darwin, Charles 250, 365n; influence on Veblen 343–6, 347, 348, 349, 350, 362n Dasgupta, Partha 108 David, Paul 2, 10, 15n, 108 Dawkins, R. 63–4 De Fermat, Pierre 31 Debreu, Gerard 139 decentralization 230, 235 decision-making 67; decision theory, optimal behaviour 21–5; definition of decision-maker 265n; game theory,

392

Index

decision-making continued interactive decision processes and individual decision-making 244–50; influence of environment 142; innovation and rational decisions 198–200; and mental representations 24–5, see also problem-solving decomposing problems 373 default classification, induction 48 Demsetz, H. 186 Descartes, René 28 deviant behaviour 284 deviation, as error 27–9 Dewey, John 341 differentiation 225, 227, 231 discontent, cause of 365n discontinuities 224–5 discovery process: cycle of exploitation and exploration 226f; cycles of discovery and (dis)integration 232–4; exploration 225–9; generalization 225, 227, see also innovation disintegration, integration and disintegration of organizational forms 229–31 dissipative structures 14n district see industrial districts; localized character diversity: and co-ordination 120; and commensurability 119; conceptual spaces 117–20; freedom of choice 127n division of labour: and business conceptions 181–4, 190; and division of knowledge 293; economies 295; and growth of knowledge 203–4; learning processes 109–10 Dorfman, J. 340, 361n Dosi, Giovanni 173n Dow, G. 162 Dreze, J.H. 260 Duhem, Pierre 33 Dyer, A.W. 366n ecological rationality 67 Economic Theory of Evolutionary Change, An (Nelson and Winter) 356 Economics and Institutions: A Manifesto for a Modern Institutional Economics (Hodgson) 357 economy of scale 233 Edgell, Stephen 340, 355, 356, 365n Eff, A. 356

efficiency, and bounded rationality 171 Egidi, Massimo 20–50, 84–5, 92, 208, 376 egoism, enlightened egoism 55 ‘Einstellung effect’ 38 Eldredge, N. 225 Ely, Richard 165, 166 empiricism: empirical observation 22; and rationalism 28 employment relation 141 entrepreneurial activity: the firm and district 294–6; and learning 311–12 entropic, economic systems as 6 environment: evolutionary psychology 70–1; firms, surviving selection environment 221–2; influence on decision-making 142; norms, mind’s handling of environmental variables 2 epistemic learning 307 equilibrium theory 165; business cycles, Hayek on 326–9; equilibrium and order, Hayek on 332–5; knowledge, Hayek on 317–21; and tâtonnement 321–5; temporary equilibrium theory 319, 321, see also cycles Erev, I. 305, 308, 312 errors: deviation as error 27–9; knowledge, trial and error 206; and learning 313n; problem-solving 26 Essay on the Nature and Significance of Economic Science (Robbins) 21 evolution, theory of 173n, 348; and adaptation 348; and cognition 3–5; and learning 313n; and technological progress 348–9, see also Darwin evolutionary algorithms 372 evolutionary economics: agents’ inertia 346–52; and cognition 3–5; evolutionary game theory 304, 313n; exaptation 7; guessing theory, Veblen 340–6; Hayek on evolution 7, 8–10, 310, 311; Hodgson 357–8; Indirect Evolutionary Approach (IEA) 55–7, 64, 65, 68, 69–70, 72, 74–5, 76n; institutions 365n; knowledge, Veblen’s sabotage-based approach 353–4; and learning 313n; neurognosis 6–7; self-organization 5–6; transaction cost economics (TCE) 53, 54–5; Veblen 338–60 evolutionary psychology (EP) 58–62, 69–71; genes 63–4; motivation 62–75;

Index 393 rationality 64–9; selfishness 63–4; Tit-for-Tat 59 exaptation 7, 9, 10, 12; penguins’ wings 15n exceptions, problem-solving 48–9 expectations, and ‘as if’ assumptions 22 expected utility (EU) theory 21–2, 24, 170–1 experience: language ordering experience 127n; neurognosis 6–7; social knowledge and morality 114, see also knowledge experiments: experimental economics, and cognitive economics 83–6; experimentalism 84, 85; individual learning, experiment 87–94; law of effect 308; and simulations 101–3; ‘Target the Two’, routinized thinking experiment 38–42; team learning, experiment 95–101, see also cognitive experimental economics exploration/exploitation 218–39; abduction 218, 219, 225; cycle of 226f; cycles, contingencies that affect 232–4; discovery process 225–9; firm strategy 234–7; innovation, cycle theory 222–5; integration and disintegration, cycle 229–31; knowledge and learning 219–21; selection environment 221–2; standardization 237–9, see also innovation external economy, and the industrial district 285–7 extroversion 97–9 Eysenck, H.Y. 97, 98 falsification, and verification 32–3 farsightedness: farsighted contracting, transaction cost economics (TCE) 54; and selfishness 63 feedback, negative and positive 6, 14n Ferguson, Adam 126–7n Fermat’s Last Theorem 31 firms: Coasian theory of the firm 178, 281; competitive genesis of the firm 184–6; district, nature of 281–97; division of labour and business conceptions 181–4, 190; exploration/exploitation, firm strategy 234–7; industrial districts 285–7; information and knowledge 282–3; knowing in firms 289–91; knowledge, capabilities approach

207–11; knowledge, Veblen’s sabotage-based approach 353–5; learning, exploration and exploitation 218–39; organizational development, opportunities and problems 186–9; reconciling exploration and exploitation 234–7; selection environment, surviving 221–2; theory of (‘F-theory’) 142, 160, 174n, 177–90, 285–9, 293, see also institutions; organizations Fleetwood, S. 316, 333 focus, and danger of myopia 221 Foray, D. 270 Ford Foundation 135, 145–7, 149, 154n Ford, Henry 227 ‘Formal Theory of the Employment Relation, A’ (Simon) 141–2 Formulario Matematico (Peano) 30 Foss, Nicolai J. 152n, 339, 359; evolutionary theories of the firm 353, 355; organizational economics of Simon 158–73; time irreversibility 14n Foster, J. 5, 13n, 14n framing effects 24–5 Fransman, M. 282 freedom of choice: and co-ordination 107; diversity 127n Freeman, C. 194, 195, 213 Friedman, D. 88 Friedman, Milton 22, 23, 143, 165, 166, 307 Gaffard, Jean-Luc 268–79 Galilean method 32 game theory 84, 243–65; ‘accepted standard of behaviour’ and ‘established social order’ 252–5, 266n; agreement and co-ordination 262–5; and bounded rationality 152n, 173n; evolutionary game theory 304, 313n; Greenberg’s social situations 255–8; hazard 245; interactive decision processes and individual decisionmaking 244–50; learning 304; Moulin’s modes of co-operation 259–61; Robinson Crusoe and social exchange economies compared 245–6, 246t; selfishness 265n; Simon’s adaptation of 143; and social organization 250–65; uncertainty 265n; Von Neumann on 243, 244–5, 246–7, see also theories

394

Index

Garrouste, Pierre 302–13 Gaussian distribution 22 Gellner, Ernest 127n General Electric, innovation 197 generalization: and cycle of integration and disintegration 231; discovery process 225, 227; law of effect 308 genes, evolutionary psychology 63–4 Georghiou, L. 194 Gersick, C.J.G. 224 Gigerenzer, G. 65–7, 372 Gintis, H. 55–6 Gödel, Kurt 30 Gould, S.J. 7, 15n, 225, 227 governance: bounded rationality 161; and cognitive leadership 234; governance regime and organizational development 187–9 Graduate School of Industrial Administration (GSIA) 145, 147 Graham, Margaret 195 Greenberg, J. 255–8, 262 Grossman, S. 163 growth theory 367n guesswork, Veblen on 340–6 Guetzkow, Harold 146, 147 Güth, W. 56, 57, 68, 76n habits: Veblen on 349–50, 365n, see also inertia; routine Halliday, Michael 127n Harris, Abraham 355 Harsanyi, J.C. 247, 261 Hart, Oliver 161, 162, 163 Hayek, F.A. 2, 4, 12; agent’s heterogeneity and knowledge 325–32; evolution 7, 8–10, 15n, 310, 311; knowledge 27, 329–32; knowledge and equilibrium 316–36; learning 220, 307; path-dependence 10; quantity theory of money 326–7; rules of conduct 308; technological knowledge 270–1 hazard, game theory 245 Henderson, R. 196 heterogeneity, Hayek on agent’s heterogeneity and knowledge 326–32 heuristic problem-solving 29; evolutionary psychology 65, 67 Hey, John D. 84, 85 history: path-dependence 10; as possible impediment to innovation 201

Hodgson, G.M. 13n, 14n; Veblen’s evolutionary economics 339, 340, 341, 343, 344, 346, 355, 356, 357–8, 359, 360n, 365n Holland, J.H. 44, 218 Holt, Charles 146 Horowitz 316 Horwitz, S. 15n Huck, S. 57 hypertext organization, exploration and exploitation 236–7 IBM 230 ICI, innovation 196 ICT, exploration and exploitation 237 ideologies: as cognitive traps 20–50, see also theories images: ability to change 6; mind building image of itself 8 imagination: and communication 111–12; and language 113–17; morality 114–15; and rationality in innovation 200–1; and social coding 124, see also creativity imitation, learning 307, 308, 310 ‘impartial spectator’ 111 imperfect resemblance 117–20, 123–4, 126n incomplete contract controversy, bounded rationality 162–4 indeterminacy 344 Indirect Evolutionary Approach (IEA) 55–7, 64, 65, 68, 69–70, 72, 74–5, 76n indirect knowledge 109 individual learning, experiment 87–94 individuals: learning 307–9; limited knowledge 305, 313n, see also agency induction 28; default classification 48; inductive reasoning 343, 362n; social knowledge 113–14 industrial districts: entrepreneurial activity and the firm 294–6; exploration and exploitation 235; the firm and external economies 285–7; the firm and institutional setting 287–8; institutional setting 287–8; nature of the district 281–97, see also localized character inertia 210; danger after consolidation 227, see also change; habits information: Arrowian paradox of information 172; cognitive processes 179; denial of validity 6; information

Index 395 structure, horizontal and vertical 235; and knowledge 204–5, 282–3; and learning 309, see also knowledge innovation: airline industry 197; ATT and Bell Laboratories 198; and capitalism 203; competence enhancing and competence destroying 224; competitive advantage as transient 212–13; cycle theory of 222–5; cycles of discovery and (dis)integration 232–4; discovery process 225–9; exploitation and exploration 218–39; history as possible impediment to 201; innovation research 194–8; integration and disintegration of organizational forms 229–31; localized character of innovation processes 268–79; and rational decisions 198–200; rationality and imagination 200–1; RCA, failed innovation 195–6; role of firm 206; ship building 197; technological knowledge 197–8, 201, see also change; discovery process; habits; inertia; knowledge institutional economics see new institutional economics institutions: definition 297n; and evolution of the mind 9; evolutionary economics 365n; institutional setting and industrial district 287–8, see also firms; new institutional economics; organizations instrumental rationality, and cognitive rationality 304 integration: integrated structure 223–4; integration and disintegration of organizational forms 229–31; positive and negative aspects 221 INTEL 230 intelligence, as internalized action 220 interaction, learning 310 internationalization, product life cycle of 223 interpretation, ordering of perception 8 introversion 97–9 invisible hand, social capital 297n irrationality, and bounded rationality 68

Jewkes, J. 195

James, William 343, 344, 348, 351, 352, 353, 356, 358, 364n

Lachmann, L. 316 Lakatos, I. 21, 50

Kahneman, D. 2, 24, 65, 66, 67, 85, 150, 166 Kant, Immanuel 341–2, 343, 352, 353, 361n Katona, George 146 Kauffman, S.A. 376, 387n Kelley, H. 88 Kelly, George 15n Keynes, Maynard 117 Kirman, A. 306 Kliemt, H. 57, 68, 76n Knight, F.H. 4, 85, 218, 294 knowledge 251; authority, as defect in our system of knowledge 31–2; behaviour 180–1; capabilities approach 207–11; Cartesian doctrine 28; collective knowing 292–4; communication, knowledge channels 276–7; Continental school 28; definition 309; deviation as error 27–8; direct and indirect knowledge 113, 123, 124; and division of labour 203–4, 293; evolution of the mind 8–9; exploration/exploitation 219–21; firms, capabilities approach 207–11; Hayek on agent’s heterogeneity and knowledge 326–32; Hayek on equilibrium theory 317–21; indirect knowledge 109, 113; individual’s limited knowledge 305, 313n; and information 204–5, 282–3; knowing in firms 289–91; learning and technological knowledge 268–79; localized character 269, 271–9, 283; Plato 27; and production 208; social division of 27, 29–30; social processes 202–4; source of, differing views 27–9; trial and error 206; and understanding 202–6, 212; Veblen on 338–60; Veblen’s sabotage-based approach 353–5, see also bounded rationality; experience; information; learning; social knowledge; technological knowledge Koopmans, Tjalling 139 Koppl, R.G. 306 Kuhn, Thomas 127n Kunkle, G.C. 194

396

Index

Lamarck, Jean-Baptiste 349, 350 Langlois, R.N. 21–2, 200, 291, 305–6 Langrish, J. 194 language: and imagination 113–17; meta-language, attempt to settle disagreements 30–1; and social coding 124 Laughlin, C. 6–7 Lazarsfeld, Paul 145 leadership, cognitive leadership, organizational development 187 learning 219–21; Austrian School 302–13; behaviour and social learning 180–1; behavioural, learning models as 312; and bounded rationality 305–7, 310; and communication 284, 311; division of labour, learning processes 109–10; and economic behaviour 4–5; and entrepreneurship 311–12; epistemic learning 307; and errors 313n; and evolution 313n; evolutionary game theory 313n; and fictitious play 304, 308; firms, exploration/exploitation 218–39; game theory 304; imitation 307, 308, 310; individual learning, experiment 87–94; individuals 307–9; and information 309; interaction 310; learning economy 268, 269, 270; learning models as behavioural 312; learning processes, division of labour 109–10; and market process 312; organizational learning 218; and rationality 302, 303–5, 307, 310; rulefollowers 306; social cognitive learning theory 15n; social learning, cognitive processes 178–81; and subjectivism 311; team learning, experiment 95–101; and technological knowledge 268–79, see also knowledge Leibowitz, S.J. 10 Leonard-Barton, D. 209 Liepert, J. 15n life sciences industry, technological knowledge 273 Lindblom, Charles 146 Loasby, B.J. 172, 293, 295 localized character: innovation processes 268–79; knowledge 269, 271–9, 283, see also industrial districts Loeb, Jacques 340, 356

Lucas, R.E. 167 Luchins, A.S. 38 Luchins, E.H. 38 Lundvall, B.A. 270 3M 230 McBride, W.M. 197 McCloskey, D. 158 MacCloughry, R. 316 McDougall, William 341, 356, 358, 361n Machlup, Fritz 165, 166 McNulty, P. 208 Malerba, F. 233 Malinowski, B. 356 March, James G. 22, 138, 144, 146, 149, 167, 218 Marengo, Luigi 371–86 Margolis, S.E. 10 market: business conceptions and market transactions 181; and learning and market process 312 Marr, David 61 Marschak, Jacob 139, 146 Marshall, Alfred 1, 165, 166, 283; industrial district 285–6, 288; internal and external economies 297n; time irreversibility 14n Marshallian approach 4 Maskin, Eric 162–3, 164, 165, 171 Mason, Edward 165 mathematics, Veblen’s attitude to 362–3n Maturana, Humberto 14n Maxwell, J.C. 344, 363n Mayhew, A. 339, 340, 361n Mead, G.H. 220 Menger, Carl 2, 4, 21; individual’s limited knowledge 313n; learning 308, 310; New Institutional Economics 177 mental representations, and decisionmaking 24–5 meta-language, attempt to settle disagreements 30–1 Metcalfe, J. Stanley 193–214 Milgrom, P. 162 Mill, James 318 Mill, John Stuart 126n Miller, George 145 Miller, Merton 146 Minirubrik 43–8 Mintzberg, H. 224

Index 397 Mirage of Social Justice (Hayek) 334 Mitchell, W. 362n Modigliani, Franco 146 Mokyr, Joel 199 Momigliano, F. 12 monetary economy, tâtonnement and equilibrium 322–3 Monetary Theory and the Trade Cycle (Hayek) 317–18, 322–3, 326, 327 monitoring regime, organizational development 188 morality: imagination 114–15; moral sense and communication beliefs 125; sentiment and social harmony 120–3 Morgan, C. Lloyd 340, 341, 361n, 365n Morgenstern, Oskar 83, 143, 146, 154n, 171; ‘accepted standard of behaviour’ and ‘established social order’ 252–5, 262; game theory 243, 245, 246–7 Moss, Lawrence 334–5 motivation, evolutionary psychology 62–4, 75 Moulin, H. 259–61, 262, 264, 265n mutual consent 126n Narduzzo, A. 38, 40, 84–5 Nash equilibrium 249, 254, 260, 302, 304–5; Nash situation 256–8 Nash, J.F. 243, 249–50 natural selection 5, see also evolutionary economics nature, Kant on 341, 361n negative feedback 6, 14n Nelson, Richard R. 3, 12, 133, 166; economic growth 367n; evolutionary theory 356–7, 358; innovation 198; social technologies 204 networks for communication, collective knowing 292 neurognosis 6–7, 11, 15n new institutional economics 173n, 177–8, 287 Newell, Allan 146, 373; problem-solving 22, 154n Nickerson, J. 172 Nike 237 Nonaka, I. 236, 237, 282, 296 Nooteboom, Bart 218–39 norms, mind’s handling of environmental variables 2 North, Douglas 108, 204, 285; institutions 297n Novarese, Marco 82–104

O’Connor, J.J. 31 oil 233–4 oligopoly theory 167 ‘On Formally Undecidable Propositions in Principia Mathematica and Related Systems’ (Gödel) 30–1 operations research 144 opportunism, transaction cost economics (TCE) 54, 58, 63, 64 optimal behaviour see suboptimal behaviour orchestration 238 order: definition 333; and equilibrium, Hayek on 332–5 organizations: bounded rationality, organizational economics 148, 160–1, 164–5, 170–2; change, organizational and institutional 11–12; co-operation, Moulin’s modes of co-operation 259–61; cognitive leadership regime and organizational development 187–9; and game theory 250–65; integration and disintegration of organizational forms 229–31; learning, organizational learning 218; organization theory (‘O-theory’) and ‘F-theory’ 142; organizational culture 220; organizational designs and localized character of technological knowledge 274–6; organizational development, opportunities and problems 186–9; Simon on organizational economics 147, 148–9, 155n, 158–73; social situations, Greenberg on game theory 255–8; standard of behaviour and established social order in game theory 252–5, 266n, see also firms; institutions Organizations (March & Simon) 147, 149 Orléan, A. 308 Page, S.E. 387n Pareto’s optimality 260 path-dependence, resistance to change 10–11, 15n Peano, Giuseppe 30 Peirce, Charles Sanders 338, 341, 352, 356; evolutionary theory 343–4, 346, 347, 348, 350–1, 353, 362n, 364n, 365n; induction 362n; knowledge 366n

398

Index

Penrose, Edith 14n, 207, 208, 211, 214n, 270 perception: interpretation as ordering 8; knowing in firms 291 pharmaceutical industry: exploration and exploitation 235; innovation 196 Piaget, Jean 220 Planning Production, Inventories, and Work Force (Simon) 144 Plato, knowledge 27 play, fictitious play and learning 304, 308 Popper, K. 21, 225; critical thinking 50; knowledge 27–9; settling disagreements 30; verification and falsification 32–3, 50 Porta, Pier Luigi 107–36 Porter, M.A. 221 positioning view, selection 221 positive feedback, and creodic processes 14n pragmatism, influence on Veblen 347, 351–2, 366n Prahalad, C. K. 214n preferences 77n; as guide to behaviour 57 prices: equilibrium theory 318–19; Hayek on business cycles 326–8; price system and rationality 333 Prices and Production (Hayek) 323, 326–7 Prigogine, I.V. 14n prisoners’ dilemma game 102 problem-solving 371–86; abstraction and classification 33–8, 48; chess 22–3, 25; competing solutions 25–7; controversies and authority 29–32; individual and collective 82–4; induction 48; prisoners’ dilemma game 102; and Protocol Analysis 22–3; puzzles 376–8; representations 34–5, 378–82; Rubik cube 34, 36; specification 48–9; Tower of Hanoi 384–6, 387n, see also decision-making procedural rationality 168–9, 283, see also bounded rationality procedural uncertainty 25 product differentiation 233 production: cycle of production system 232; and knowledge 208 productive contingencies, organizational designs 275 profit maximization, Simon’s questioning of 143

profitability, organizational development 187 Prospect theory 24 Protocol Analysis, and problem-solving 23 Prychitko, D. 316 psycho-neurobiological processes, choice 4 Pure Theory of Capital, The (Hayek) 324, 325, 326 Putnam, Robert 108 Putterman, L. 55 puzzles: problem-solving 376–8, see also problem-solving quantity theory of money, Hayek on 326–7 Quéré, Michel 268–79 Quinn, J.B. 237 Rabin, M. 151, 313n Radner, Roy 139, 154n, 167, 169 Raffaelli, T. 14n Rand Association 84 ‘Rational Choice and the Structure of the Environment’ (Simon) 142 rationality: bounded rationality 22; capacity to get rid of prejudices 50; and competition 333; ecological rationality 67; empirical observation 22; and empiricism 28; evolutionary psychology 64–70; Hayek on order 333; in innovation 198–201; instrumental and cognitive rationality 304; irrationality and bounded rationality 68; and learning 302, 303–5, 307, 310; and satisficing 141; substantive rationality 77n, see also bounded rationality; procedural rationality RCA, failed innovation 195–6 Reber, A. 14n reciprocity 55–6, 74, 227, 231 regret theory 24 representations: problem-solving 34–5, 378–82; representation problem 372–3 research and development (R&D) 194; RCA, failed innovation 195–6, see also innovation Richardson, G.B. 208, 211, 271, 294 Ridley, Clarence 137, 153n Rizzello, Salvatore 1–13, 15n, 84, 85, 87, 88, 92

Index 399 Robbins, Lionel 21, 166 Roberts, J. 162 Robertson, D.H. 281 Robertson, E.F. 31 Robinson Crusoe economy, compared with social exchange economy 245–6, 246t Robson, A.J. 64 Romanes, Georges 340, 344, 360n, 363n Roth, A.E. 305, 308, 312 routine 210; and bounded rationality 169; knowing in firms 291; routinized thinking experiment, ‘Target the Two’ 38–42, see also habits Rubik cube, problem-solving 34, 36 Rubinstein, Ariel 152n, 309 rule-followers, learning 306 Samuelson, Paul 154n Sargent, Thomas 158, 167 satisficing: in contrast to rational choice 141; theory of satisficing search 168 Say, J.B. 318 Scazzieri, Roberto 107–36 Schelling, Thomas C. 108, 126n Schendel, D. 197 Schmidt, Christian 243–65 Schneider, Louis 361n Schumpeter, Joseph A. 193, 356; entrepreneur 220; evolutionary approach 4; theory of innovation 201 Schumpeterian view, evolutionary approach 3 Sciences of the Artificial, The (Simon) 142 scripts, hierarchy of innovations and institutions 228 search, theory of satisficing search 168 selection, firm strategy 221–2 self-interest 365n self-organization, evolutionary economics 5–6 selfishness: evolutionary psychology 63–4; game theory 265n Selten, R. 67, 261 Sensory Order, The (Hayek) 8, 15n Sent, E.M. 158 sentiment, and social co-ordination 120–3, 125 Shapley value 254, 264 Shell 234 ship building, innovation 197 Silverberg, G. 13n

Simon, Herbert A. 2, 4, 305, 373; analysis 86; behavioural economics 145–50; biographical sketch 133–6; bounded rationality 22, 50n, 66, 67, 108–9, 133–52; Ely and Nobel lectures 165–9; organizational economics 147, 148–9, 155n, 158–73; problem-solving and Protocol Analysis 23; procedural rationality 168–9, 283; rationality, limits of 136–9, 356; substantive rationality 77n simulations 2, 101–3 Smith, Adam 203, 294, 295, 297; imagination and sympathy 111–12; imperfect resemblance 123–4; knowledge creation process 293; language and imagination 113–17; moral sense 125; mutual consent 126n; social harmony and sentiment 121–3 Smith, C.G. 198 Sober, E. 77n social capital, invisible hand 297n social closure 127n social co-ordination see co-ordination (social) social coding, and imagination 124 social cognitive learning theory 15n social harmony, belief co-ordination 121 social interaction, effectiveness of 122 social knowledge 107–36; bounded rationality 108–9; and co-ordination 107–10; imagination 113–17; imperfect resemblance 117–20; learning processes 109–10, see also co-ordination (social); knowledge social organizations see firms; institutions; organizations social protest, as lacking true foundation 347 social science, and biotic 360–1n socialist planned economy, Hayek on 327 sociobiology 76n Solow, R.M. 356 Some Neglected Points in the Theory of Socialism (Veblen) 347 specialization 235 specification, problem-solving 48–9 spectator, Smith on 114 Spencer, H. 347, 349, 350, 355, 356 Sperber, D. 71

400

Index

stag-hunt game, Moulin’s modes of co-operation 261 standardization 226, 237–8, 286 state, transaction costs 287 Storper, M. 289 subcontracting, business conceptions 182 subjectivity: Hayek on agent’s heterogeneity and knowledge 325–32; and knowledge 202; and learning 311; subjective utility theory 167 subjects, Veblen on 351–2 suboptimal behaviour: decision theory 21–5; suboptimal solutions as stable 49; suboptimal strategies 37–48 Suppes, Patrick 127n supply and demand: tâtonnement and equilibrium 322–3; telecommunications and aircraft industries 272–3 surprise, bounded rationality 109 Sylos Labini, Paolo 108 sympathy: and communication 111–12; social co-ordination 122, 126n Tagliagambe, S. 332 Tahler, Richard 65 Takeuchi, H. 236, 237, 282 ‘Target the Two’, routinized thinking experiment 38–42 tâtonnement, theory of 321–5 team learning, experiment 95–101 team theory 160, 173n technological knowledge 268–79; aircraft industry 272–3; car industry 273–4; communication, knowledge channels 276–7; and innovation 197–8, 201; learning 268–79; life sciences industry 273; localized character 269, 271–9; organizational designs 274–6; organizational designs and localized character of technological knowledge 274–6; rationality and imagination in innovation 201; telecommunications industry 272, see also knowledge technological progress, and theory of evolution 348–9 Teece, D.J. 196 telecommunications industry, technological knowledge 272 temporary equilibrium theory 319, 321 Tesfatsion, L. 102 theories: as cognitive traps 20–50;

comparing competing theories 32–8; controversies and authority 29–32; critical thinking 29; and suboptimal strategies 37–48; unrealistic assumptions, Simons criticisms 142–3, see also game theory Theory of Games and Economic Behavior (von Neumann and Morgentsern) 243, 245, 246–7; ‘accepted standard of behaviour’ and ‘established social order’ 252–5 Theory of the Leisure Class, The (Veblen) 351 Theory of Moral Sentiments (Smith) 121 Thornton, S. 29, 33 thought: critical thinking 50; Veblen on 362n Tilman, R. 355, 356, 361n time irreversibility 14n Tirole, Jean 162–3, 164, 165, 171 Tit-for-Tat, evolutionary psychology 59 Todd, P.M. 67 Tooby, John 58–62, 66, 67, 76n, 356 traditions, co-ordination 108 transaction cost economics (TCE) 53, 54–5, 65, 160; and bounded rationality 54, 65, 66, 161–2, 171–2; and cognitive psychology 172; incomplete contract controversy 162–4; transaction cost theory 173n translation, conceptual spaces 119 trial and error, growth of knowledge 206 Turvani, Margherita 15n, 281–97 Tushman, M.L. 224, 232 Tversky, A. 2, 24, 65, 66, 67, 85, 150, 166 Twomey, P. 339, 340, 347, 353, 355, 358, 359 Tylor, E.B. 356 uncertainty, game theory 265n understanding: and capability 209–10; and knowledge 202–6, 212; Peirce on 343, see also knowledge; learning uniqueness 289 utility maximization: Simon’s repudiation of 136–7, see also expected utility (EU) theory Utterback, J.M. 197 Varela, F.J. 14n Veblen, Thorstein 2; agents’ inertia and subjects’ response 346–52; guessing

Index 401 theory 340–6; sabotage-based approach to knowledge 353–5 Veracitas Dei 28 verification, and falsifcation 32–3 Verri, Pietro 127n Viano, Frencesca Lidia 338–60 Vincenti, W.G. 197 Volberda, H.W. 234 Volterra, V. 248–9, 265n Von Neumann, John 83, 143, 154n, 171; ‘accepted standard of behaviour’ and ‘established social order’ 252–5, 262; game theory 243, 244–5, 246–7 Vromen, J. 58 Vygotsky, L. 220 Wagner, G. P. 372 Wallis, Allen 145 Walliser, B. 304 Walras, Marie 318, 320 Wason, Peter 59 Wason Selection Task 59–60 Wealth of Nations, The (Smith) 203 Weibull, J. 307, 308 Wellsprings of Knowledge (LeonardBarton) 209

Weyl, Hermann 363n Wiles, Andrew 31 Williams, B.R. 194, 195 Williamson, Oliver E. 133, 151–2, 167n, 169, 173n, 357; transaction cost economics (TCE) 53, 54–5, 58, 63–4, 65, 66–7, 69–70, 72, 74, 75, 147, 161, 163, 171–2, 287 Wilson, D.S. 77n, 201 Wilson, G. 97, 98, 167 Winter, S.G. 3, 12, 133, 166, 169, 356–7, 358 Witt, Ulrich 14n, 15n, 234; entrepreneurship and learning 311–12; learning and communication 284, 311; theory of the firm 177–90 Yaari, M.E. 56, 57 Young, H.P. 307, 308, 310, 312 Zanella, A.W. 200 Zenger, T. 172 Zwick, R. 304–5 Zywicki, T. 55