The Psychology of Freedom

This book presents a new theory of the will - of our capacity for decision-making. The book argues that taking a decision to act is something we do, and do freely - as much an action as the actions which our decisions explain - and that our freedom of action depends on this capacity for free decision-making. But decision-making is no ordinary action. Decisions to act also have a special executive function, that of ensuring the rationality of the further actions which they explain. This executive function makes decision-making an action importantly unlike any other, with its own distinctive rationality. Pink's original and highly persuasive study uses this theory of the will to provide new accounts of freedom, action, and rational choice. The author argues that, in a tradition that runs from Hobbes to Davidson and Frankfurt, Anglo-American philosophy has misrepresented the common-sense psychology of our freedom and action - a psychology which this book now presents and defends.

THE PSYCHOLOGY OF FREEDOM

THE PSYCHOLOGY OF FREEDOM THOMAS PINK Lecturer in Philosophy, King's College London

CAMBRIDGE UNIVERSITY PRESS

CAMBRIDGE UNIVERSITY PRESS Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo Cambridge University Press The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www. Cambridge. org Information on this title: www.cambridge.org/9780521555043 © Cambridge University Press 1996 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 1996 This digitally printed version 2007 A catalogue record for this publication is available from the British Library Library of Congress Cataloguing in Publication data Pink, Thomas. The psychology of freedom / Thomas Pink. p. cm. ISBN 0 521 55504 3 (hardback) 1. Free will and determinism. 2. Decision-making. I. Title. BJ1461.P55 1996 128'.3-dc20 95-51644 CIP ISBN 978-0-521-55504-3 hardback ISBN 978-0-521-03822-5 paperback

ForJudy

Contents

Acknowledgments

page x

Introduction

i

1

Agency and the will

14

2

Scepticism about second-order agency

33

3

Decision-making and freedom

64

4

The Psychologising conception of freedom

101

5

Decision rationality and action rationality

137

6

Decision-making and Teleology

166

7

The Regress argument

187

8

In defence of the Action model

228

9

The special-purpose agency of the will

247

Conclusion

269

Bibliography Index

276 278

IX

Acknowledgments

Most of this book was written at the end of a research fellowship at Churchill College, Cambridge, for which I owe thanks to the Master and Fellows. My thanks in particular to Tom Baldwin for supporting this project. In addition the following have read through drafts of this book and given many comments: John Bishop, Tim Crane, Rob Hopkins, Mike Martin, Joe Mintoff, David Owens, Paul Noordhof. My thanks also to participants in meetings at Oriel College, Oxford and at Heidelberg, to David Gauthier for some correspondence and comments, to Bill Brewer, Edward Craig, Nick Denyer, Dorothy Edgington, Frank Hahn, Susan James, and Charles Kahn for useful discussions, to various referees for their criticisms, and to Paul Snowdon for his support and advice at a late stage. Acknowledgments are due to the Editor of Mind for permission to use material from two articles: 'Purposive intending', published in 1991, and Justification and the will' published in 1993. Lastly, my thanks to the staff at Cambridge University Press who have helped bring this book to press, and especially to Judith Ayling, Hilary Gaskin and Gillian Maude.

Introduction

This book is about the psychology of our freedom. Byfreedom I mean the freedom of alternative possibilities: the freedom to do things or not do them, or - as I shall also put it - control over whether we do those things or not. It is just this freedom that we think we possess in relation to much of our action. We ordinarily think that we are free, say, to go out in the evening or stay in - that it is within our control, or up to us, which of these actions we perform. The Psychology of Freedom examines what sort of mind this freedom of action requires. It seeks to determine what mental capacities and states we need if we are to be free to act otherwise than as we actually do. In particular, the book examines what sort of will a free agent must possess. By the will I mean our capacity to determine in advance how we shall act through decision-making, thereby forming intentions about which actions we shall perform. Much of the book discusses what goes on when we take a decision to act, and how the nature and rationality of our decisions relate to the nature and rationality of the actions which those decisions explain. This is because our ordinary conception of our freedom of action involves a conception of the will which is very demanding and perhaps not even coherent. It is a controversial and unsolved problem what goes on in the decision-making by which we determine our actions. And this problem about the will is part of a wider problem about the nature of freedom itself. Our ordinary conception of our freedom of action - I argue requires that decisions combine two characteristics which, as we shall see, are not obviously consistent. Our conception of our freedom of action demands, first, that decisions to act serve an executive function - that they serve to apply practical reason as it concerns the actions decided upon. The point of taking decisions about which actions to perform, after all, is to ensure that we end

2

The psychology of freedom

up performing the right actions. And secondly, our conception of our freedom demands that decisions count as deliberate or intentional actions themselves. The book explains why our conception of our freedom requires that decisions combine these two characteristics, why these characteristics are in tension, and how the apparent inconsistency between them can in fact be resolved by a new theory of practical reason and agency. The book's argument is complex. But then the issues which it addresses are not simple either. So this introduction will present, in outline, the book's general argument. I shall begin by saying something about what I take to be our ordinary conception of our freedom. THE PSYGHOLOGISING CONCEPTION OF FREEDOM

In the seventeenth century, Thomas Hobbes asserted a quite novel and very simple theory of freedom - a theory which certainly contains no special obscurity in its psychology. According to Hobbes, an agent is free - has control over which actions he performs - to the extent that he has a capacity to act however he desires. To be free, then, one needs nothing more by way of a will or action-determining capacity, than a capacity to hold desires to act, and to form beliefs about means to performing the actions desired. But Hobbes' theory was deeply revisionary, both of our ordinary conception of our freedom, and of our ordinary conception of the will. There is more to our freedom of action, as we ordinarily conceive it, than Hobbes allowed. And there is correspondingly more to the kind of will which that freedom requires. It is within our control which actions we perform only because we are capable of taking decisions about how we shall act, and it is within our control or up to us which such decisions we take. Is not that a claim about our freedom to which we naturally assent? If so, our ordinary conception of our freedom is Psychologising. By which I mean that our freedom of action is conceived to depend on a further, psychological analogue of that freedom - on a freedom of some at least of the mental events and states which motivate and give rise to our action. Freedom of action is conceived to depend on a prior freedom of will.

Introduction

3

And that means that our freedom of action depends on our possessing a rather special psychology. For some at least of the mental occurrences which motivate and give rise to free actions must constitute yet further actions themselves. Whether I decide to stay in or decide to go out is supposed to be within my control - up to me -just as is whether I stay in or go out. But that means that, just like staying in or going out, taking a decision to stay in, or taking a decision to go out, must be something which I can deliberately do. Decision-making must be a further form of deliberate agency or doing, additional to the agency of the action which it explains. A Psychologising conception of our freedom ties our freedom of action to our possession of a prior capacity for active decision-making - to our possession of an action-generating agency of the will. It is important to distinguish a Psychologising conception of our freedom from any volitional or conational theory of action. Some philosophers think that our bodily actions are explained by or involve volitions or conations whereby we try to move our bodies. But, as I argue in this book, decisions are importantly different from anything which constitutes trying. To take a decision to act is not, like trying, to initiate bodily movement, but rather to form a persisting psychological state - a state of intention in which we are left motivated to act as decided. Decisions are forms of agency which explain intentions; whereas tryings, like bodily actions, are quite distinct forms of agency which intentions explain. Taking a decision to act, then, is not just an action-generating action. Unlike tryings or conations, I argue, a decision is an action which generates action by way of constituting the formation of an intervening action explanatory psychological state. That makes decision-making a highly distinctive kind of action-generating action -what I term a. second-order action, by contrast to thejirst-order actions it generates. It is the argument of this book, that our concept of agency is deployed at two levels. First we have our everyday actions and attempts at action - our first-order agency. Then we have, among the psychological states which explain and rationalise our firstorder agency, some such as intentions whose formation constitutes agency too. That agency, of which the most intuitive case is decision-making, is our second-order agency. By tying our action control to a prior decision control, a Psychologising conception of

4

The psychology of freedom

freedom ties the freedom of our first-order agency to a capacity for second-order agency. The purpose of The Psychology ofFreedom is to set out and defend precisely this Psychologising conception of our freedom of action. The book does this by providing a theory of what second-order agency is, and why free agents need a capacity to perform it. And this theory of second-order agency is proposed within a theory of agency in general - a theory of what unites our decisions and the actions which they explain to make them all actions. The book addresses two crucial questions which the Psychologising conception of our freedom raises, and which any successful defence of it must answer. First, why should freedom of action depend on a prior freedom of will? And secondly, do we actually possess the second-order, action-generating agency of the will which freedom of will involves? I shall now say something about each of these questions in turn.

WHY SHOULD FREEDOM OF ACTION DEPEND ON FREEDOM OF WILL?

It is not simply our ordinary conception of our freedom which is Psychologising. There have been plenty of philosophical accounts of freedom which are Psychologising too. We find Psychologising accounts of freedom within Stoicism, throughout mediaeval Scholasticism, and in Kant. These accounts typically appeal, as I shall also be appealing, to the assumption that freedom of action involves a capacity for rational self-determination - a capacity, which we possess as agents who are practically rational, to reason or deliberate about how to act, and then to apply those deliberations in the exercise of control over our action. Freedom of action then depends on freedom of will, because this capacity for rational self-determination requires the possession of a free will. My own Psychologising theory is of this type. But it is going to differ from many past Psychologising theories of freedom at one important point. In many past theories, freedom of will was often conceived as the freedom of a deliberative capacity - as a freedom of our practical reason. This is true, in particular, of Kant's theory of freedom. Freedom of will, according to Kant, is a freedom

Introduction

5

which we possess as rational deliberators - for, in his view, as rational deliberators we are the sovereign legislators of reason's requirements on action. My theory is rather different. On my theory, the will on whose freedom our freedom of action depends, is not in any sense a deliberative capacity. Rather, the will is an executive capacity a capacity for applying our prior deliberations about how to act. A decision is a second-order executive action - an action by which we ensure that we subsequently perform the first-order actions which, as deliberators, we have judged it desirable to perform. Freedom of action depends on freedom of will, I shall be arguing, because to exercise action control in a way which applies our deliberations about how to act, our action control must extend into the future. We must be able to exercise control in advance over what future actions we perform - a control that we exercise through a prior action-generating agency of decision-making. Our ordinary conception of the will is, I argue, precisely a conception of it as a faculty for second-order executive agency - an agency the function of which is to apply reason as it governs the first-order agency of the actions decided upon. For, as we ordinarily conceive it, decision-making provides a highly distinctive, reasonapplying way of controlling one's future actions. We can, of course, use ordinary, first-order actions to control our future actions. Jon Elster's Ulysses and the Sirens contains a wellknown discussion of various ways of doing just this. Ulysses is a case in point, who had himself bound to a mast to prevent himself from subsequently trying to leave his ship to join the Sirens. But there are other, more everyday and less drastic examples. Concerned for our health, we might perform some first-order action now in order to get ourselves to eat less in future. For example, we might now tell our friends that we are going on a diet. That way we bring the fear of losing face in as a further motive to induce ourselves to eat less in future. But such use of present first-order action to control one's future action is essentially manipulative. One is twisting the arm of one's future self to get it to perform actions which now, prior to the arm twisting, one favours performing on quite other grounds. Fear of losing face is in the future going to be the motive which induces one to eat less. It is a concern for one's health, however, which is presently motivating one to make use of that fear to ensure that

6

The psychology of freedom

one eats less. This arm-twisting use of present first-order action to control future action can even threaten or wholly remove one's future freedom of action. Ulysses' use of present action to control his future action placed a brute physical constraint on what he might do in the future. The use of the second-order agency of our decision-making to control future action is, by contrast, essentially freedom-preserving. Having taken a decision to act - a decision which will cause us to act as we have decided — we none the less always retain a continuing freedom not to act as we have decided. And that is because decisionmaking is also a method of future action control which is essentially non-manipulative. Deciding to perform an action is not a form of arm-twisting. When one takes a decision to act, one is not putting brute constraints on one's future self's capacity for action. Nor is one creating inducements or penalties to get one's future self to act as one now thinks it should. And that is because as second-order agency, decision-making is, quite distinctively and unlike first-order action, a motivation-perpetuating cause of future action. A decision to act affects future action, I shall be arguing, simply by perpetuating the motivating force of the considerations that have already motivated it. The decision ensures that the reasons which motivated one to take it also motivate one subsequently to perform the action decided upon. If I decide to eat less out of a concern for my health, then my decision ensures that that same concern for my health will thereafter motivate me actually to eat less. Taking a decision perpetuates our present motivation into the future, so that our motivation in the future will remain one with which we can, as we are at present constituted, immediately identify. When we take a decision to act, our future self is left a person who entirely shares our present rationality - and who, precisely because of that shared rationality, is motivated and willing to act as we have decided. Decision-making, then, is a very special kind of future actioncontrolling action. By taking a decision to act, we are not only applying practical reason now; we also ensure that we go on applying that same practical reason into the future, in our very performance of the actions which our present deliberations recommend. Decision-making, then, is a form of action by which we exercise future action control in a way that ensures our continuing

Introduction

7

rationality as agents. And it is that form of future action control which, I shall be arguing, we need to be free. Thus the book contains an original explanation of why freedom of action depends on freedom of will. And this explanation is itself based on a new theory of decision-making as a reason-applying, second-order executive action - an action by which we ensure that we act rationally thereafter.

IS THERE AN AGENCY OF THE WILL?

But is taking a particular decision to act a deliberate action itself? Whatever ordinary intuition might say on the matter, from Hobbes onwards plenty of philosophers have been pretty sure it is not. And many more philosophers have simply avoided the issue. This book is endebted to Michael Bratman for his important work on the action-coordinatory and planning function of intention. But in his Intention, Plans and Practical Reason, Bratman notably refrains from claiming that forming a particular intention to act - what we do when we take a decision to act - is itself an intentional action. And such restraint is entirely understandable. It is a serious question whether there really is an agency of the will. It is not hard to see where the scepticism comes from. The actions which our decisions explain are purposive. We perform these actions for the sake of some end — either for their own sake, or for the sake of wider ends extending beyond what we are doing. We perform these actions, then, in the belief that they will attain desirable ends. And that is connected with the fact that the practical deliberation which explains our actions is precisely deliberation about how to act — it is precisely deliberation in which we consider what ends our actions might further. But the forming of intentions to act does not seem to be purposive. It does not seem to be motivated by beliefs about what ends it will further. We form intentions to do A because we believe doing A - the action decided upon - will attain desirable ends, and not because we believe that now intending to do A will attain desirable ends. The deliberation that explains our intentions to act, then, is simply deliberation about which actions to perform. Our intentions are not generally based on practical deliberation about which intentions to form.

8

The psychology of freedom

Again actions, at least in their uncontroversial first-order form, are subject to the will. We can perform actions A on the basis of having decided to do so. But the decisions which determine our actions are not themselves subject to the will in the same way. We cannot, it seems, take decisions to do A on the basis of having decided to take them. Decisions to act, then, look importantly different from the actions which they explain. Actions are purposive, based on practical deliberation about whether to do them and about what ends they would further, and are subject to the will. Typically, decisions to act share none of these properties. Why should we suppose that these decisions to act are deliberate actions too? These dissimilarities between decisions and the actions they explain, have been the staple of scepticism about second-order agency since Hobbes. The Psychology ofFreedom concedes that these dissimilarities do indeed exist, and uses its new theory of decision-making to explain them. At the same time, the book also provides a new theory of agency - a theory which shows why, these differences from ordinary first-order action notwithstanding, decisions to act are indeed deliberate actions themselves. What might unite decisions to act and the actions which they explain to make them uniformly actions? Perhaps the uniting factor is practical reason itself. It may be that there is a distinctively practical reason - a distinctively practical way in which the justification, or lack of it, for all our practice or agency is determined. In which case decision-making can count as action if, just like the firstorder action which it explains, it is governed by reason in this distinctively practical way. But what is practical reason? One possibility is that reason in its distinctively practical form comes to means-end justifiability. Firstorder actions are means-end justifiable - that is, they are justified in terms of the likelihood that performing those particular actions would further desirable ends. What if decisions to do A are means—end justifiable too? One way of defending the agency of decision-making, then, is by appeal to what I call the Action model of decision rationality. On this model, decisions to act are meansend justifiable - justifiable in terms of the desirable ends which they themselves are likely to further — just as are the actions which those decisions explain. And that makes decisions to act actions too.

Introduction

9

DECISION RATIONALITY: THE ACTION MODEL VERSUS THE PRO ATTITUDE MODEL

However, it is not obvious that the Action model of decision rationality is true. The Action model of decision rationality seems to clash with the belief, which I have argued our conception of freedom also demands, that decisions to act have an executive function - that they serve to apply reason as it governs our subsequent action. For if the function of decisions to act is executive - if the function of decisions to act is to apply reason as it governs first-order action - then should not the rationality of deciding to do A be explained simply in terms of the rationality of doing A? And so we have the Pro Attitude model of decision rationality - a view of decision rationality assumed in the work of philosophers such as Anscombe, Davidson, and Bratman. On the Pro Attitude model, justifications for deciding to do A are simply justifications for eventually doing A. Decisions to do A are justified simply in terms of the desirability of doing A - in terms of the likelihood of doing A furthering desirable ends, and not in terms of the likelihood of those decisions themselves furthering desirable ends. On the Pro Attitude model, whilst the actions which our decisions explain are means-end justifiable, our decisions are not. We have then two competing models of decision rationality. One, the Pro Attitude model, says that decisions to do A are justified in terms of desirable ends which doing A would further. The other, the Action model, says that decisions to do A are justified by desirable ends which deciding to do A would further. And the potential difference between these models emerges clearly if we imagine, as Gregory Kavka has asked us to imagine, a huge cash prize offered simply for taking a decision to perform a particular action A - a prize which we win irrespective of whether we ever carry out the decision thereafter. It is just the taking of a decision to do A which the prize rewards, and not doing A.1 Now on the Pro Attitude model, decisions to act cannot rationally be motivated by prizes awarded just for taking them. For such prizes do not justify the actions decided upon -winning the prize is not a desirable end which the action decided upon would further. See Kavka's 'The toxin puzzle'; Analysis (1983).

io

The psychology of freedom

And so, the Pro Attitude model implies, such prizes cannot justify the deciding to do A. On the other hand, it might appear, the Action model implies that the prize does justify deciding to do A. For is not winning the prize a desirable end which deciding to do A would further? But then, if decisions to do A can be made rational by prizes offered just for taking them, the rationality of taking a decision to do A need have no connexion at all with - and certainly will not guarantee - the rationality of subsequently doing A. The Action model seems inconsistent with the view that the essence of deciding to perform an action is executive - that decision-making serves to apply practical reason as it concerns the actions decided upon. At the heart of the book, then, is a deep tension in our conception of the will - the very conception of the will as an executive agency that our belief in our freedom demands. We think of decisions as executing reason as it governs first-order action. The Pro Attitude model of decision rationality looks true. At the same time, we think of decisions as being actions themselves - and so as justified by the desirability of taking them, just as the actions which our decisions explain are justified by the desirability of their performance. The Action model of decision rationality looks true. The Psychology of Freedom resolves this tension. It argues that the Pro Attitude model of decision rationality is false, and the Action model is true. And that makes decisions actions. On the other hand, it defends a restriction - which I term REASON-APPLY - on the ends which can rationally justify one's taking a decision. The Action model says that justifications for deciding to perform an action are desirable ends which taking that decision would further. But it is consistent with the Action model that not every desirable end which a decision would further can justify taking that decision. According to REASON-APPLY, desirable ends can only justify one's taking a decision to act if, once that decision is taken, they would thereafter be furthered by, and so justify, one's acting as decided. That means that decisions are not justified by prizes offered just for taking them. For such prizes do not thereafter justify acting as decided. The Action model, when supplemented with REASONAPPLY, places restrictions on what can justify decisions that are very like - though, as we shall see, importantly not identical with - those imposed by the rival Pro Attitude model. And so we arrive at a version of the Action model that, I shall be arguing, is fully

Introduction

11

consistent with our belief that decisions are executive in function — that they serve to apply reason as it governs subsequent action - as well as with our belief that decisions are actions themselves. The Psychology of Freedom thus proposes a new theory of decision rationality. The theory, by supplementing the Action model with REASON-APPLY, constitutes a hitherto unsuspected via media between the Pro Attitude model of decision rationality, and the Action model in its unsupplemented form. This new via media, I shall argue, allows us to combine what, as users of common sense psychology, we very much want to combine - the doctrine that decisions serve to apply reason as it governs our subsequent action, and the doctrine that decisions are actions themselves. And so the book provides a convincing defence of the reality of second-order agency - a defence which remains faithful to our ordinary conception of decision-making, while making sense of second-order agency within a wider theory of agency in general. My theory has the merit of leaving wholly intelligible the widespread philosophical disagreement we find about decision rationality. It explains why each of the Pro Attitude and Action models should have had its own band of determined supporters. My theory leaves it no surprise that so many competent philosophers should have wanted to defend the Pro Attitude model of decision rationality. That is the model which at the outset, before we supplement the Action model with REASON-APPLY, appears most consistent with the executive function of decision-making. The Pro Attitude model already provides, just as it stands, a very good approximation to the truth. On the other hand, my theory also leaves it no surprise that so many other equally competent philosophers should have wanted to defend the Action model. For after all, decisions are actions; and the Action model is in fact true. CAUSAL DETERMINISM AND MORAL RESPONSIBILITY

There is another well-known problem about freedom - a problem which is about the metaphysics of freedom, rather than its psychology. This is the dispute between Incompatibilists and Compatibilists about whether freedom of action is compatible with causal determinism. The Psychology of Freedom is not intended to answer this question. If you want to know whether freedom really is compatible with causal determinism, this book will not tell you.

12

The psychology of freedom

But this traditional problem about freedom's metaphysics is still touched on, in chapter 3. My purpose in so doing is to examine the relation between Incompatibilist and Compatibilist views and my own Psychologising theory of freedom. Recently, Incompatibilist writers such as Robert Kane have also begun to argue that our freedom of action depends on a prior freedom of will. It is important to show that my own Psychologising theory is not Incompatibilist though nor, I suspect, is it essentially Compatibilist either. My conclusion, albeit somewhat tentative, is that my Psychologising theory is largely neutral in the dispute between Incompatibilists and Compatibilists. But whether or not I am right about that is, for me, a rather secondary concern. The question of whether freedom of action depends on freedom of will, is a problem about freedom which deserves our attention in its own right. And it is to this particular problem about freedom's psychology which the book is devoted, and not any other. I should also warn that as a book about our freedom, The Psychology of Freedom is not also a book about moral responsibility and blameworthiness. As Harry Frankfurt has reminded us, to be morally responsible for our actions is one thing, and to be a free agent - to have control over whether or not we perform them - is another. Hence, it is an important and sensible question what the precise relation between moral responsibility and freedom will turn out to be. It may well be that our ordinary conception of freedom owes comparatively little to any theory of moral responsibility. After all, such interest as we do have in our own freedom need not be ethical. It need not come from some concern to be truly deserving of morally based praise or blame for what we do. We may simply want to be truly in control. So philosophers must be careful not to moralise the theory of freedom. They should not assume that conceptions of freedom necessarily have an ethical motivation or rationale. And this book will argue that neither our belief in an agency of the will nor our belief in a dependence of freedom of action on freedom of will does have anything whatsoever to do with moral theory. These beliefs are part of a more general theory, implicit in common-sense psychology, not of moral responsibility, but of our rationality. Our belief in a Psychologising conception of our freedom is based on a conception of freedom as a capacity for

Introduction

13

rational self-determination. And this, we shall see, is a conception of freedom which is quite independent of any thoughts about moral responsibility or blame. The key notion in my theory of freedom - as in my theory of agency and of the will - is rationality, not responsibility.

CHAPTER I

Agency and the will

ACTIONS AND PURPOSES

Action is purposive. Whenever someone deliberately performs an action, they are doing something in order to attain some end. The end's attainment might be caused by the action - as when a doctor gives his patient medicine in order to cure him. But the end's attainment might also be constituted, either in part or in whole, by the action's performance. I might be performing an action simply for its own sake. When I twiddle myfingers,the relevant end might simply be that my fingers be twiddled. Purposiveness is a salient characteristic of action. It is part of what makes our action seem so distinctively active, rather than passive - so distinctively something which we do, as opposed to something which happens to us. To do something as a means to an end is, at the very least, to do something. So what makes action purposive? What makes it true that I am performing some action A purposively, as a means to an end E? There is a familiar answer to this question. The purposiveness of action is constituted, at least in part, by the psychological states or attitudes which cause and explain it. These attitudes are desires and beliefs which constitute the agent's reasons for the action's performance. They are desires and beliefs which rationalise the action. Any action is going to be explained and rationalised by some desire to perform it. To be motivated to perform an action A, an agent must want to do A at least as much as he wants to perform any other option available to him. And, where the action is not being performed simply for its own sake, this desire to perform it is in turn to be explained in terms of a desire for some further end, and a belief that the action would further that end. If I am doing A in order to attain some end E, then this is going to be because I 14

Agency and the will

15

desire to attain E, and combine this desire with a belief that doing A would further E. Such an action-rationalising combination of desire with means-end belief I shall call a purpose for doing A. The combination counts as a purpose for the action, in that it rationalises the action as a means to an end. This familiar account of how actions are explained by our psychological states I take to be true, as far as it goes. There surely is, in particular, a sense of'desire' that applies whenever we are motivated to act. In what follows, at any rate, I shall be questioning not the truth of this account of action, but its completeness. For our actions are not just explained by beliefs and desires. Our actions are often explained by what appear to be further actions - actions that are performed in the head, and which somehow generate the motivation for the actions which they explain. SECOND-ORDER ACTIONS

That our actions are often explained by further actions is suggested by our beliefs about ourfreedom. We suppose that we are free to act otherwise than as we actually do. For many actions A, we are both free to do A and free not to do A. This freedom to act otherwise - the freedom of alternative possibilities, as it is sometimes called - is what I mean when I write of our freedom of action. I shall also refer to this freedom as control over which actions we perform. To be free to do A or not, is to possess control over whether or not one does A. By control over our actions, I shall again always mean, in what follows, a freedom to perform them or not to perform them - a freedom of alternative possibilities.1 Some of the time, then, we have freedom of action. We have control over which actions we perform. Or so we ordinarily think. But this control over which actions we perform is not the only control which we think we possess. 1

Some philosophers - particularly those interested in providing a theory of moral responsibility - are happy to write of our performing an action 'freely' or of our 'exercising control' over an action even in cases where we are not free to act otherwise, and so lack control over whether we perform that action or not. We shall meet such a use of the term 'freely' in chapter 3. I shall not be using the terms in this way. 'Freedom' and 'control' and related terms always imply, as I use them, a freedom to act otherwise - a freedom of alternative possibilities. For it is in freedom so understood that I am interested.

16

The psychology of freedom

Besides our capacity for action, we also believe ourselves to have a will - a capacity for decision-making or intention formation. Prior to performing actions, we can take decisions or form intentions to act. These intentions, when formed, then persist to the time of action and cause us to act as intended. And this capacity of will, we believe, is also one which we are free to exercise otherwise than as we actually do. The will is also a capacity over which we believe ourselves to have control. Suppose that this afternoon you will have the opportunity to perform some action, such as going to the dentist. Besides any control you will have this afternoon over whether or not you go, there is another control which you also have now. You have control over which action you now decide to perform - over whether you now form an intention to go to the dentist, or instead form an intention not to go. It is as much up to you whether you now decide to visit the dentist as, subsequently, it will be up to you whether you actually do visit the dentist. Not only do we suppose ourselves to have a freedom to act otherwise; we suppose ourselves to have a freedom to decide and intend otherwise as well. We have freedom of will as well as freedom of action. And that means we have two forms of action control. Not only do we have and exercise control over our actions as we perform them. We also have a control over our actions which we can exercise in advance. By taking a decision now to go to the dentist this afternoon, I can ensure that from that time onwards I intend or remain decided on going - an intention which persists to the time of action, so that I eventually go. On the other hand, had I decided not to go to the dentist this afternoon, that decision would have prevented me going. I should have formed a persisting intention not to go to the dentist. So, given its effects on subsequent action, decisionmaking can be used to exercise control in advance over which future actions the decision maker performs. Decision and intention control gives us future action control. Where does our control over our decisions and intentions come from? In general, our control over anything comes from what we do. Control presupposes the possibility of deliberate agency - some deliberate doing or action in which that control could be exercised. If we are to have any control over an occurrence O, there must be some action over which we have control in turn, where our deliberately performing or refraining from the action would either

Agency and the will

17

cause or constitute the occurrence of O. As for control in general, so for decision control in particular. So do we exercise control over our decisions through performing some prior decision-causing action? Surely not. As I shall put it, our control over which actions we decide to perform is direct - that is, it can be exercised without having to do something else first which then causes a particular decision to be taken. To exercise decision control, we only need to take a particular decision or refrain from taking it. The agency which gives us decision and intention control is agency which constitutes the taking of a particular decision to act, rather than agency which causes that decision to be taken. Taking a particular decision to act is itself a deliberate action - an action-generating mental action. And so we ordinarily suppose decision-making to be. Taking your decision to visit the dentist is something you deliberately do, just like the subsequent visit to the dentist which that decision explains. That you take that particular decision is not something which simply happens to you. Like the visit which it explains, your decision is not a passion but an action. Taking a decision to do A is a psychological action with intentional content - a content which specifies a further action, doing A, which that decision serves to cause and explain. Our concept of agency is deployed at two levels. We perform actions which, in turn, are explained by various action explanatory attitudes — by attitudes such as desires and intentions which rationalise the actions which they explain. These actions constitute our first-order agency. But then the formation of some at least of these action explanatory attitudes counts as agency too. These mental, action explanatory actions count as our second-order agency. By second-order agency I shall mean the formation of any particular action explanatory psychological attitude which itself counts as our doing, and counts as such other than because it results from some prior first-order action. I need to be able to refer to our agency as a whole, whether first or second-order, collectively. On the other hand, there is an obvious danger of confusing decisions, and any other second-order actions which we might perform, with the first-order actions which they cause and explain. I need therefore to adopt some conventions about what, when used without the explicit qualifiers 'first order' and 'second order', terms such as 'action', 'activity', and 'agency'

18

The psychology of freedom

are to mean. So when I use the term 'action' without further qualification, and without explicitly applying it to decisions, I shall mean by it simply first-order action. But when I use the terms 'agency5 or 'activity' without qualification, I shall mean agency in general, of whatever order it might be. That we apply the concept of agency at two levels - that we conceive the psychology which explains our actions to contain a further, action explanatory agency within it — is a remarkable fact. We need to explain what could justify such a conception of our agency; or if no such justification is to be had, why we do in any case conceive of our agency in such terms. That common-sense psychology - our everyday psychological belief - includes agency within the psychological process that explains agency, raises the question of that psychology's own character and function. INTENTION AND AGENCY

Decision-making counts as action of the second order because it is the active formation of an action explanatory psychological state the state of intending to act. This state of intention is a psychological state which is tightly linked to our agency. Its close connexions with agency are what immediately distinguish intention from ordinary desire. First, intention is always directed at agency. An intention is always an intention to do something, or to refrain from doing it. If I intend that E, I must be intending to bring E about through my performing or refraining from performing some action. Whereas I can want E, without wanting E to be brought about through my agency. Secondly, intentions are not only directed at agency. They ensure that one is motivated outright to act as intended. That is, one cannot be failing to act as intended because, though one still intends to act, one's intention is being overridden by a desire to do something else. If I fail to do something simply because I want more strongly to do something else, then I cannot really be intending to do what I fail to do. Whereas ordinary desires can certainly be so overridden. I can want to get up, but be staying in bed because I also want to do that. Desires incline one to act; but they may fail to motivate one outright to act as desired. Thirdly, intentions tend to persist until the time intended for

Agency and the will

19 2

action. Intentions have inertia, as Bratman has put it. Precisely under what conditions intentions do persist until the time intended for action is a matter which we shall be discussing in a later chapter. Roughly, I shall be arguing that an intention should tend to persist for as long as the beliefs on which its formation was originally based are not lost or revised. Since intentions ensure that one is motivated outright to act as intended, and tend to persist until the time intended for action, or have inertia, intentions tend to cause one to act as intended. And that means intentions are characteristically accompanied by the expectation that one will or at least might act as intended. Indeed some such belief is arguably a condition of holding the intention.3 There is no such belief condition on desire. Finally, as we have seen, intentions not only cause agency. They also arise out of agency. We characteristically form intentions by taking decisions to act; and these decisions to act seem to be our agency too. And this is an obvious contrast with the case of desire. Desire formation is not generally itself a case of agency. Forming a desire seems, in general, to be something that happens to us, and not something we do. Desire formation, characteristically, is passive passion, not active action. What, then, is the state of intention? I shall be arguing that an intention is a psychological state which is formed actively, through the second-order agency of the will - a capacity for second-order action by which we affect which first-order actions we are thereafter motivated to form. To say that someone holds an intention is to say that a motivation-affecting second-order action has been performed, and that this action's performance continues to produce its characteristic effects. In particular, it is to say that the agent still retains the relevant motivation to first-order action, and concomitantly refrains from further exercising the will in any way which would, by constituting a change of mind, remove that motivation to act. 2 3

Sec Intention, Plans and Practical Reason, p. 16. It is sometimes claimed that intentions to do A imply, or are actually a form of, outright or full belief that one will do A. On this view, someone who intends to do A cannot be in doubt about whether they will in fact do A. At no stage, however, does my argument require such a strong belief condition on intention - a belief condition which is anyway very controversial. I assume only that someone intending to do A at least believes that they might do A.

20

The psychology of freedom

The state of intending, then, is the state of remaining decided. It is the state of remaining motivated by one's past second-order action and so, in particular, of refraining from further second-order action that would undo that motivation. And that makes intention an inherently active psychological state. An intention's formation consists in our doing something; and its persistence consists in our refraining from further agency which would counteract the effects of what we have done. Many philosophers would deny that intention is inherently active. Some philosophers deny this because they deny that there is any such a thing as second-order agency at all. In their view, no intention formation is active. Others allow that second-order agency can occur, but deny that intentions need always be formed by it. Our decision-making, they agree, does count as agency, and intentions formed through it are formed actively. But not all intentions are formed through decision-making. And some or all of those intentions which are not formed through decision-making are passive. They are intentions which simply come over us.4 I need to argue against both rival views. I shall shortly be considering the view that denies second-order agency outright. But I want to say something now about the compromise view- the theory that intention formation is sometimes active, but is not always or inherently so. It is perfectly true that not all intentions are formed through decision-making. Intention formation counts as decision-making only if at the time of the intention's formation the agent actually entertains alternatives to what he intends. And this he might not do if, for example, he were forming an intention out of a habit - as when, after awaking, he drowsily forms an intention to turn on the light. But, I shall argue, the distinction between intentions formed through decision-making and other intentions is not a distinction between active and passive intentions, and has no significance for the theory of agency. So although the claim that all intending is a remaining decided is a tad oversimple, it does not materially 4

Thus Mele: 'a decision is a deed . . . Intentions, however, are not deeds, and some intentions are not products of deeds. People in the habit of answering knocks at their office doors typically do not perform the action of forming an intention to answer such a knock. But when they answer these knocks, they typically intend to answer them. Generally, these intentions are products, not of decision, but of habit' ('Intention, belief and intentional action', p. 25).

Agency and the will

21

misrepresent the relation between intention and second-order agency. Such distinctions between intentions are irrelevant to the theory of agency because all intentions are governed by rationality in the same way. The same general theory of intention rationality - of what justifies forming one intention rather than another - applies to intentions generally, whether or not their formation counts as decision-making, and so whether or not the intender actually entertains alternatives to what he intends. And that means, I shall be arguing, that intention formation is either inherently and universally active, or is inherently and universally passive. For our agency, as I shall be claiming in the next chapter, is distinguished by the inherently practical way in which it is governed by reason - by its distinctively practical rationality. If agency is determined to be such by its rationality - by the way in which it is governed by reason - then intention formation could not be variously active or passive unless the rationality of its formation were various too. The compromise view, then, will not be my principal target. The most serious rival to my theory that intention is inherently active is the theory that intention is inherently passive - the theory that second-order agency does not occur at all. And that means that there is one special condition which I will not be placing on second-order agency. The entertainment of alternatives at the time of an intention's formation is no more necessary to second-order agency - to an intention formation's counting as doing - than it is to first. If, when drowsy in the morning, my turning on the light can be a genuine action without my entertaining alternatives to its performance, then so too can my first forming an intention to turn on the light. SECOND-ORDER AGENCY, RATIONALITY AND FREEDOM.

Does it matter whether we are second-order agents? Two things in particular have been thought to depend on second-order agency. A capacity for second-order agency has been seen as a condition on our practical rationality. And it has also been seen as a condition on ourfreedom. It is very natural to think, and many philosophers have taught, that a special psychology is needed both for rationality and for freedom. This psychology, which not all agents are sophisticated enough to possess, involves a capacity for active decision-making.

22

The psychology of freedom

Intentions do seem to require more intellectual sophistication than do desires. It is unclear that merely because they are moved to act by desires and beliefs, less intelligent animals such as sharks and mice must also be capable of forming intentions about how they will act. This is not surprising if intention formation is inherently active - if a capacity to form intentions is a capacity for a secondorder, motivation-affecting agency of the mind. For why should all agents be so developed as to possess a capacity for this kind of higher order agency in particular? But there is another action explanatory attitude that also seems to require some special intellectual development. As well as forming intentions about how we shall act, we also form judgments about what ends and actions it would be desirable to attain or perform. Such judgments, which I shall term practicaljudgments, are formed in the course of our deliberation about how to act — our practical deliberation. As well as holding desires for ends, and forming desires to act on their basis, we also judge that the attainment of various ends might be desirable, and infer from these judgments to judgments about which actions are desirable. Practical judgments seem to be beliefs of a relatively sophisticated kind. By making practical judgments one comes to cognise, or mistake, justifications for one's action -justifications that lie in the fact that the actions would further desirable ends. And that means that to make practical judgments, one must deploy a concept of desirability. But the deployment of such a concept is surely not required for ordinary desire. To want something is not the same as to think it desirable. And so one can want something, and realise that one wants it, while still wondering whether what one wants is really worth wanting. Some agents may be incapable of thinking things desirable because, although they possess desires, they altogether lack the concept of the desirable. Animals such as sharks and mice may want things, and may be moved to act by their wants. But does their capacity to want things imply a capacity to judge them desirable? Both our capacities for practical judgment and intention formation seem, then, to require more intelligence than is needed for holding simple desires. It also seems intuitive that the two kinds of capacity should come together. If one can form intentions about how one will act, it is plausible that one is also going to be able to base those intentions on prior deliberation — and so on prior

Agency and the will

23

judgments about what ends it would be desirable to attain. And vice versa: if one can arrive at judgments about what ends are desirable, will not one also be able to form intentions about which actions to perform on the basis of these judgments? So much do practical judgment and intention seem to come together, that some philosophers - the most recent example is Donald Davidson - have simply taken intention to be a form of practical judgment.5 To form an intention to act, on their view, just is to make a judgment about how, overall, it is most desirable for one to act. To form an intention to do A, is just to judge that doing A would be more desirable than any alternative. Many philosophers have conceived of the will as a straightforwardly deliberative capacity. But there is an alternative way of conceiving of the relation between practical judgment and intention - a view of their relation which I shall be defending. On this view intentions are quite distinct from practical judgments. One can form intentions which are not accompanied by any conviction that what is intended is desirable. One can form intentions which are not actually explained by concomitant practical judgments at all. Nevertheless, the capacities for practical judgment and intention are still possessed together, because the core function of intention formation is to apply or execute our practical judgments about how we should act. Our capacity for practical judgment is a deliberative capacity, while our capacity for intention formation is an executive capacity. Between them, whether they are the same or distinct, our capacities for practical judgment and for intention intuitively constitute our capacity for practical rationality. To be practically rational, it is not enough simply to pursue desired ends on the basis of beliefs about how those ends might be attained. One must also be able to deliberate and make judgments about which ends and actions are desirable, and - either in the very making of those judgments, or, as I prefer to maintain, in applying them - to form intentions about which actions to perform. Practical rationality requires a capacity for practical judgment and intention. Or so it is very natural to suppose. But if intention is, as I have suggested, an inherently active psychological state, then it follows that, in so far as practical 5

See Davidson's 'Intending'.

24

The psychology of freedom

rationality requires a capacity for intention, it also requires a capacity for second-order agency. If practical rationality comes to more than being moved to act by beliefs and desires, if in particular it requires a capacity for decision-making or intention formation, then it follows that practical rationality also requires a capacity for active self-determination - a capacity to cause oneself to perform actions by performing prior second-order actions in the mind. But a capacity for second-order agency may not be a condition of our practical rationality alone. It also seems to be a condition of our freedom. It is equally natural to suppose that we are free agents that it is up to us which actions we perform - only because we have control of the decision-making by which we determine how we act. In which case freedom of action is going to depend on a prior freedom and second-order agency of the will. There is going to be a special psychology of freedom - a psychology which involves a capacity for second-order agency, and which a mere desirepossessing animal could lack. In the remainder of the book we shall be examining the connexions between second-order agency, freedom, and practical rationality. We shall be looking at the part which a second-order agency of the will might play both in our practical rationality, and in our freedom. We shall be determining whether practical rationality really does require a capacity for second-order agency and whether there really is a dependence of freedom of action on freedom of will. The belief that second-order agency is a condition both of rationality and freedom was once widely held. Mediaeval and early modern Scholasticism held that our rationality and freedom, which was supposed to distinguish us from the lower animals, depended on our special psychology - on our possession of special faculties or capacities of intellect and will. Intellect, at least as a specifically practical intellect, or capacity to make practical judgments, and will were possessed by us together. If one had the one capacity, one had the other. For Aquinas, a will - our decision-making capacity just was the distinctively rational appetite that came with a capacity to make practical judgments. The will was carefully distinguished from the merely sensitive appetite - the capacity for holding desires - which we and the lower animals shared. This faculty of will was exercised by us freely - just as freely as the actions which its exercise explained.

Agency and the will

25

There is much in this scholastic theory which I do not wish or need to defend. Particularly debatable is its dualist metaphysics the location of practical judgment and intention in immaterial capacities. Equally debatable is the scholastic assumption that the capacities for practical judgment and intention were lacked by all non-human animals, without exception. But I shall certainly be defending the core Scholastic doctrines - the doctrines that as humans we, at least, are capable of second-order agency, and that such agency is a condition both of our rationality and of our freedom. And I shall be defending these doctrines against the arguments of their most formidable opponent - an opponent whose dismissal of second-order agency still influences philosophical, psychological, and economic theory today. HOBBES

According to Thomas Hobbes, as rational and free agents we need not and do not possess any special faculties of intellect and will. The practical deliberation which explains our actions does not ever consist in the formation of any practical judgments distinct from desire. All deliberation ever consists in is the combination of desires for ends - as Hobbes put it, appetites or hopes for their attainment - with beliefs about which actions would further those ends, to constitute desires to perform those actions: When in the mind of man, Appetites, and Aversions, Hopes, and Feares, concerning one and the same thing, arise alternately; and divers good and evill consequences of the doing, or omitting the thing propounded, come successively into our thoughts; so that sometimes we have an Appetite to it; sometimes an Aversion from it; sometimes Hope to be able to do it; sometimes Despaire, or Feare to attempt it; the whole summe of Desires, Aversions, Hopes and Fears, continued till the thing either be done, or thought impossible, is that we call DELIBERATION. {Leviathan, p. 44)

As Hobbes asserted against a scholastic opponent, Bishop John Bramhall, nothing goes on in human deliberation and action that does not go on in animals too: And first your Lordship's own experience furnishes you with proof enough, that horses, dogs, and other brute beasts, do demur sometimes upon the way they are to take, the horse retiring from some strange figure that he sees, and coming on again to avoid the spur. And what else doth a man that deliberateth, but one while proceed toward action, another while retire from it,

26

The psychology of freedom

as the hope of greater good draws him, or the fear of greater evil drives him away. ('Of Liberty and Necessity', pp. 244-5) What philosophers had called our will - our intentions and our capacity to form them - Hobbes again explained in terms of desire. A will to perform some action A consists simply in one's desire to do A being left, at the conclusion of one's deliberation about how to act, strong enough to override any desires to act otherwise, and so strong enough actually to move and dispose one to do A: In Deliberation, the last Appetite, or Aversion, immediately adhaering to the action, or to the omission thereof, is that wee call the WILL . . . Will therefore is the last Appetite in deliberating. (Leviathan, pp. 44-5) In Hobbes' view, then, the action explanatory attitudes found in humans may have more complex contents than those found in animals. But they are not psychological attitudes of a different kind. Human rationality and freedom required no special capacity for a second-order agency of the will. A free agent is simply an agent who is capable of acting as he wants to act: A FREE-MAN, is he, that.. .is not hindred to doe what he has a will to ... from the use of the word Free-will, no Liberty can be inferred of the will, desire or inclination, but the Liberty of the man; which consisteth in this, that he finds no stop, in doing what he has the will, desire, or inclination to doe. (Leviathan, p. 146)

So Hobbes denied both that second-order agency ever occurs, and that it is of any importance either to our rationality or to our freedom. And this denial has proved enormously influential since. In particular, it has marked Anglo-Saxon theory, from Hume through to Ramsey, the pioneer of modern rational choice theory, and Ryle. Much Anglo-Saxon theory has tended to embrace what I shall term an Enlightenment psychology - a psychology which either ignores second-order agency, or which actually denies its occurrence. All these thinkers ignored or denied the importance of second-order agency either to our rationality or to our freedom. And this Anglo-Saxon tradition of disbelief or at least doubt in second-order agency still has plenty of supporters. Daniel Dennett, for example, has wondered whether there really is an agency of decision-making which explains our actions: Are decisions voluntary? Or are they things which happen to us? From some vantage points they seem to be the preeminently voluntary moves in

Agency and the will

27

our lives, the instants at which we exercise our agency to the fullest. But those same decisions can also be seen to be strangely out of our control. We have to wait to see how we are going to decide something, and when we do decide, our decision bubbles up to consciousness from we know not where. We do not witness it being made) we witness its arrival. {Elbow Room, P . 78) And in his moral psychology, as we shall soon see, Bernard Williams denies that action arises out of a second-order agency of the will. Why, more than agency in general, should a second-order agency of the will pose a particular philosophical problem? There are very good reasons why - reasons which we shall be examining in depth in the next chapter. For the agency of the will, at least as we ordinarily conceive it, is something of an anomaly. That is, there are obvious and intuitive differences between formation of intentions to act, as we ordinarily conceive them, and the actions which those intentions explain. And these differences make it hard to see how the formation of a particular intention to act could ever constitute a deliberate action itself. It is these differences which have generated the scepticism about second-order agency that we find in the work of Hobbes and his successors. In the next chapter I shall be exploiting these differences between decision-making and action to present a series of arguments for the passivity of the will. We find versions of these arguments recurring in philosopher after philosopher, from Hobbes in the seventeenth century to Davidson in the twentieth. These arguments so often recur precisely because they do look like good arguments - and they have yet to be answered. If Hobbes and his successors were wrong, it has yet to be shown why. The best-known modern defence of an agency and freedom of the will is to be found in the work of Harry Frankfurt, whose views we shall also be discussing. As we shall see, Frankfurt has tried to explain what freedom of will might come to by modelling the agency of our will very closely on the agency of our action. In so doing he denies the very differences between decisions to act and actions which sceptics about second-order agency have always wanted to emphasise. But these differences really do exist. If there had not been these differences, there would never have been so much doubt and debate about will agency in the first place. It is Frankfurt's refusal to recognise these differences which, I shall be arguing, undermines his account of freedom of will. Frankfurt's

28

The psychology of freedom

refusal to recognise these differences means that his defence of our freedom of will fails to address the concerns of the sceptic - and it fails to provide an account of second-order agency as we ordinarily conceive it. To answer scepticism about second-order agency, we need a developed theory of agency in general - a theory that then explains precisely what second-order agency is, and how consistently with its status as agency, it can still differ in certain salient respects from much of the first-order action which it explains. Scepticism about second-order agency has not yet been refuted, because such a theory of agency has not yet been developed. It is the task of this book to provide the general theory of agency required. It is important that, in opposing Hobbes, I shall also be expounding and defending the common-sense psychology of the will - the psychology of the will to which our everyday beliefs about decisionmaking and intention commit us. Hobbes and other proponents of Enlightenment psychologies have often claimed to be defending or analysing common sense. Theories of second-order agency, they maintain, are an unwarranted philosophical invention, and not part of our ordinary conception of the mind. But in this Hobbes and his followers misrepresent what common-sense psychology actually says. Enlightenment psychologies are radical revisions of commonsense psychology, and not faithful interpretations of it. Common-sense psychology, I shall argue, is a theory of our own rationality and freedom. It characterises intentions and other psychological states in normative terms - in terms of what would count as a rational justification for holding them. And the possible relations, as conceived in common-sense psychology, between intentions and other events and states such as actions, practical judgments, and desires, all depend on these facts about rational justification - about the various ways in which the rationality of these occurrences is determined. It is the way in which reason governs our mental states and actions which, I shall be arguing, explains why intention formation, in particular, counts as second-order agency. Again, it is the way in which reason governs our mental states and actions which, I shall also argue, explains why both our rationality and our freedom as agents depend on our having this capacity for second-order agency. And it is the way in which reason governs our mental states and actions which, I shall show, explains the scope of our second-order

Agency and the will

29

agency - why there are some things which we can do at will, just on the basis of deciding to do them, while other things cannot be subject to the will at all. Since I shall be expounding and defending the common-sense psychology of the will, let me say more about what I take the content of that psychology to be.

DECISION-MAKING AS AN EXECUTIVE AGENCY

I have already mentioned one account of the role played by decision-making in our practical rationality. Many philosophers, I noted, have simply identified intentions with practical judgments; they have simply conceived of the will as a deliberative capacity. To decide to do A is to form a judgment that the doing of A is desirable overall. Davidson's well-known theory of intention is only the most modern version of such a deliberative conception of the will. We find versions in Stoicism and in mediaeval Scholasticism. Such an identification commits us to conceiving of freedom of will as a deliberative freedom - a freedom of practical judgment. In Stoic theory, human agency is distinguished from animal agency by the human capacity for sunkatathesis, or making assent to propositions about how one should act. And this capacity for assent is a capacity over which we have an assured control. The Stoic then takes second-order agency to be an agency of practical judgment. Our will control is a practical judgment control. Aquinas similarly understood our freedom of will in terms of a freedom of practical judgment that comes with our rationality. He claimed that, in general, where the truth is not self-evident, a rational agent can appreciate that fact, and is left with control over which judgment he eventually arrives at: If the objects [of judgment] are such that the mind assents to them of its nature, thus first principles, the giving or withholding of assent does not lie within its power but arises from the nature of things, and is therefore not properly speaking subject to our command. However there are some objects perceived which for some cause or other do not so convince the mind, so that the mind can assent, dissent or suspend judgment. In such cases each of assent or dissent lies within our power, and is subject to our command. (Summa Theologiae, ia2ae, question 17, article 6, p. 196)

30

The psychology of freedom

As for judgments in general, so for practical judgments in particular. Usually, there is conflicting evidence about whether the performance of a given action would be desirable. A rational agent can appreciate that fact, and is left free in respect of his final judgment about whether or not he should perform the action: the universal [the good] contains many particulars potentially; so that the universal conception can be applied to many and diverse things. For this reason the judgment of the intellect concerning things to be done is not determined to one thing only . . . And this means that [rational] beings have freedom of choice, which is defined as the free judgment of reason. (Summa contra Gentiles, book 2, chapter 48, p. 146) Our freedom of will is identified with, or is at any rate linked to, a freedom of practical intellect. It is based on a deliberative and judgmental freedom. But it is far from obvious that intention is actually a form of practical judgment - that to take a decision is actually to arrive at a conclusion about how overall it is most desirable to act. Our freedom of will does not appear to be a deliberative freedom. There are some familiar considerations that suggest why the will is not a deliberative capacity. For one thing — and this is an often made point — the options between which one has to decide may be judged equally desirable. Faced by an array of identical biscuit packets in a supermarket, one can judge that each biscuit packet is as desirable as the other. So if one has to decide between them, one's decision to pick out one packet from all the rest cannot be a practical judgment made about that biscuit packet in particular. Ex hypothesi one does not judge that biscuit packet to be any more desirable than the rest. Taking the decision to pick that biscuit packet out is not, then, a deliberative move. One's deliberation has already been completed in one's conclusion that there is nothing to choose between all the biscuit packets. The decision - the exercise of one's will - must be the exercise of a non-deliberative capacity. Secondly, it appears that one can decide on an option despite judging it to be less desirable than alternatives that are also available to one. One can decide acratically - that is, one can take a decision which irrationally flouts one's judgment about how, overall, it would be desirable to act. One can decide to smoke another cigarette. The intuitive possibility of acratic decision-

Agency and the will

31

making presupposes a distinction between decisions and intentions on the one hand, and practical judgments on the other. There is a third and less commonly proffered consideration. But it is no less relevant than the others. It is very much more intuitive that we have a freedom of decision than that we have a freedom of practical judgment. Whether I decide to work harder may be at least as much within my control as whether I work harder. But is it as much within my control how desirable I think working harder to be? Prima facie, then, our conception of the will seems to be a conception of a kind of second-order agency which, like the action that it explains, is non-deliberative. Or so I shall be arguing. And I shall be using my account of the rationality of intention formation to establish the point. Common-sense psychology distinguishes intention from practical judgment because the two states have a different rationality - they are governed by reason in crucially different ways. So when we take a decision to do A, this is an action which is just as distinct from the making of any practical judgment, as is the action of doing A which results thereafter. As I put it earlier, decision-making is not a deliberative, but an executive, agency. Decisions are actions by which we apply our deliberations. This idea that decision-making is a second-order executive agency has often been met with a special scorn. Bernard Williams has recently claimed that: a decision is not a special kind of action . . . {Shame and Necessity, p. 36) and has attacked in particular the idea that decision-making is a non-deliberative or executive agency: Homer has no word that means, simply, 'decide'. But he has the notion. For he has the idea of wondering what to do, coming to a conclusion, and doing a particular thing because one has come to that conclusion; and that is what a decision is. He also has the idea of coming to a conclusion about what to do later, and doing that thing later because of that conclusion. All that Homer seems to have left out is the idea of another mental action that is supposed necessarily to lie between coming to a conclusion and then acting on it: and he did well in leaving it out, since there is no such action, and the idea of it is the invention of bad philosophy. {Shame and Necessity, p. 36)

But, whatever Homer may have thought, we ourselves do appear to have the very conception of the will which Williams decries. We do

32

The psychology of freedom

think of taking a decision as a non-deliberative action. And whether or not this non-deliberative agency of the will 'necessarily' intervenes between deliberation and action, we think that it characteristically does so. What is more plausibly a peculiarity of 'bad philosophy', I shall be arguing, is the attempt to conceive the freedom and agency of our decision-making in deliberative terms, as a freedom and agency of practical judgment. We need now to look at why so many modern philosophers, from Thomas Hobbes to Bernard Williams, have doubted whether second-order agency occurs at all, be it in deliberative or executive form.

CHAPTER 2

Scepticism about second-order agency

THE REDUCTION ARGUMENT

We ordinarily assume that we do have a freedom to decide otherwise — that which actions we decide to perform is within our control - and that, therefore, taking a particular decision to act is as much something which we deliberately do as is the action which that decision explains. We believe that we have a capacity for secondorder agency. But philosophers have often denied that second-order agency occurs at all. Hobbes was a pioneer sceptic. His scepticism about second-order agency, we have seen, was part of a more general scepticism about the existence of intentions and practical judgments as action explanatory attitudes distinct from desire. Hobbes denied that there was any special psychology required for rational or free agency. What separates us humans as agents from desire-possessing sharks and mice, is not our possession of quite new kinds of psychological attitude which those animals lack - but merely the fact that our desires have far more complex and varied contents. We can extract from Hobbes' psychology of action the following Reduction argument both against second-order agency and freedom of the will. Ignoring cases where we affect which desires we form through doing something else first - as when we go for a walk to work up an appetite - we seem to lack control over our desires. Desire formation seems to be passive rather than active. So we have: (1) Desire formation is passive. And then it follows from Hobbes' psychology of action that: (2) All intention formation is desire formation. 33

34

The psychology of freedom

For a Hobbesian intention is just a desire to act which overrides contrary desires to dispose us to act. Therefore (3) There is no agency of intention formation. And since freedom of will presupposes second-order agency: (4) There is no freedom of will. To refute the argument, we need to prove the falsehood of one of the two premisses (1) and (2). Now run-of-the-mill desire formation - desire formation that our intuition immediately recognises as such - does seem to be passive. So one obvious strategy is to concede that (1) is or may well be true, but maintain that at any rate (2) is false. We have freedom of will because intention is an active psychological attitude that is distinct from desire. But it could also be that (2) is true, and (1) is false - that all intentions are desires, but that some desire formation is active. Nothing has yet ruled out the possibility that there is some desire formation which is not run-of-the-mill - that what we think of as the formation of an intention may really be a rather special case of the formation of a desire, and that such desire formation is active. Freedom of will might yet turn out to be a freedom of desire. For the moment, then, I shall continue to leave it open that freedom of will could turn out to be a freedom of desire, and that second-order agency may after all turn out to be consistent with Hobbes' psychology of action. But there is one particular and highly influential conception of freedom of will as just such a freedom of desire which we should regard with great suspicion from the very outset. Frankfurt has a theory of the freedom and agency of the will which, I shall suggest, is radically revisionary of our ordinary conception of second-order agency. FRANKFURT ON FREEDOM OF WILL

Frankfurt gives the same account of the will as did Hobbes. A desire to do A constitutes our will if it is strong enough to override competing desires and move and dispose us outright to do A. But Frankfurt also claims - and here he differs from Hobbes - that desires can be explained by second-order desires, or desires to desire. In

Scepticism about second-order agency

35

particular, that a desire is strong enough to count as our will can sometimes be explained by the fact that we want it to be that strong - by the fact that we want it to constitute our will. So, according to Frankfurt, just as our actions are explained by desires to perform them, our desires can also be explained by desires to hold them — by second-order desires. And this is what falsifies premiss (1). Some desire formation is active. When our first-order desires, or their strengths, are explained by secondorder desires that they be held, we count as second-order agents — as agents who are active with respect to our desires as well as our actions. An agent with freedom of action, according to Frankfurt, is an agent who can act as he desires. Just so, Frankfurt infers, an agent with freedom of will is an agent who can desire as he desires. Now freedom of action is (roughly at least) the freedom to do what one wants to do. Analogously, then, the statement that a person enjoys freedom of will means (also roughly) that he is free to want what he wants to want. ('Freedom of action and the concept of a person', p. 20) Freedom of action consists in our action depending on our will - on our action explanatory desires to act. Freedom of will consists in those desires depending, in turn, on our desire explanatory desires to desire. On Frankfurt's view, then, even if the will is reducible to desire, we can still possess freedom of will. Freedom of will does not require a new kind of psychological attitude - a state of intention which is quite distinct from any desire. What freedom of will does require, however, is desires with a second-order content. For second-order agency we need second-order motivation.

More recently Frankfurt has described decisions to act as being mental actions by which we identify with and regiment our desires. Hence decision-making involves a motivation that is second-order: In making up his mind a person establishes preferences concerning the resolution of conflicts among his desires and beliefs. Someone who makes a decision thereby performs an action, but the performance is not of a simple act that merely implements afirst-order desire. It essentially involves reflexivity, including

desires and volitions ofa higher order. Thus, creatures who are incapable of this volitional reflexivity necessarily lack the capacity to make up their minds. They may desire and think and act, but they cannot decide. ('Identification and wholeheartedness', p. 176 (my emphasis))

36

The psychology of freedom

A second-order agent, according to Frankfurt, is concerned not simply with how he acts in the future, but also with the desires that motivate and dispose him to act. In taking a decision to go to the dentist, I am concerned not only with whether I go, but also with whether beforehand I hold desires that are going to move me outright to go. So when I take a decision to go to the dentist, I am not motivated by a mere first-order desire to go to the dentist. Rather I must be motivated by a second-order desire that my firstorder desires be such as would move and dispose me to go to the dentist. One advantage of Frankfurt's theory, is that it does justice, or appears to do justice, to our intuition that decision-making requires a special intellectual capacity - a capacity of which simple desirepossessing agents, such as sharks and mice, may be incapable. It is quite plausible, after all, that simple desirers, such as sharks and mice, fail to possess the concept of a desire, and so fail to conceive of themselves as desirers. In which case, though they may possess desires, they cannot take decisions. For they lack the second-order motivation - the desires to desire - which, according to Frankfurt, decision-making requires. But, at the same time, our intuition that decision-making requires a special intellectual capacity is shown to be consistent with a psychology of action that is fundamentally Hobbesian — which identifies the will as a form of desire. Frankfurt's theory suffers, however, from a major disadvantage. It fails to provide an account of second-order agency as we ordinarily conceive it. Frankfurt makes two crucial assumptions. First, he assumes that second-order agents need to be concerned not only with how they act, but also with the psychological states that motivate their action. Secondly, Frankfurt assumes that the action explanatory attitudes which compose a free and active will are explained by a second-order version of the very same attitudes - by a second-order version of the will. But both these assumptions are false of second-order agency as we ordinarily conceive it. What matters to a second-order agent?

Consider your decision to go to the dentist. Why might you take such a decision? Surely your reasons for deciding to go to the dentist are going to be just your reasons for actually going to

Scepticism about second-order agency the dentist. You have a tooth-ache, want to cure it, and believe going to the dentist a means thereto. The reasons that move you to decide to perform an action are, characteristically, simply reasons for performing that action. Hence the concerns that you have as someone taking decisions about how to act need be no broader than the concerns that you have as a first-order agent. All you need be interested in is, what first-order actions you perform, and what consequences those actions have. So Frankfurt is just wrong when he says: Someone who makes a decision thereby performs an action, but the performance is not of a simple act that merely implements a first-order desire. For a simple act which implements a first-order desire is precisely what a decision appears to be. Frankfurt would have it that, simply as someone taking decisions about how to act, you must be concerned not only with which firstorder actions you perform, but also with which first-order desires move you to act. But it is unclear why you need have this further concern. It is perfectly true that your present decisions will only lead you to act as decided if by the time of action you are motivated to act as decided. But why should this aspect of the mechanics of decision-making- an aspect which we shall be returning to consider in later chapters - be a matter you care about? When you decide to go to the dentist to cure your tooth-ache, all you need care about is whether you get to the dentist, and whether your getting there leads to a cure. Frankfurt, then, foists onto decision makers concerns which, simply as decision makers, they can perfectly well lack. But we find a similar error made in a corresponding theory of practicaljudgment. According to this theory, a judgment that a given action A is desirable is simply a special kind of desire. It is a second-order desire - a desire to desire to do A. Though it identifies practical judgment with desire, this theory avoids the implausibilities of identifying practical judgments with first-order desires. I can want to want something without actually wanting it; and I can want something without wanting to want it. So the theory is consistent with the possibility that I should want something without yet judging it to be desirable - that I should even want it while actually judging it undesirable. The theory that

37

38

The psychology of freedom

practical judgments are second-order desires also allows for the possibility of acrasia. Since it is on our first-, and not our secondorder desires that our motivation for action depends, an agent can perform an action whilst judging that action less desirable overall than others available to him. Furthermore, on this theory sharks and mice can lack a capacity for practical judgment for the same reason as, on Frankfurt's theory, they lack a capacity for decisionmaking. Sharks and mice can lack a capacity for practical judgment because they lack a capacity to conceive of themselves as desirers. So, although the theory holds practical judgments to be desires, it does not implausibly award all desirers a capacity for practical judgment. But the theory that practical judgments are second-order desires faces the same problem as Frankfurt's account of decision-making. The theory is thoroughly ad hoc. It attributes to the makers of practical judgments concerns which, simply as makers of practical judgments, they need not have. When you form a judgment that it is highly desirable to get rid of your tooth-ache, it is false that ipso facto you must care about your own motivation - that you must care about whether you desire to get rid of your tooth-ache. Gripped by your tooth-ache, all you are likely to desire is to be rid of it. You need not also desire to desire to be rid of it. (If you did reflect about your desires, it might even occur to you that the pain of the toothache could, if anything, be worsened by your so strongly wanting to be rid of it.) Again, if, on seeing you in danger, I decide to help and deliberate about the best or most desirable way to do so, I need have absolutely no interest in whether or not I am motivated to help you. All that need concern me is whether I do in fact help you. The same is true if I am faced by a dilemma - as when I am forced to decide between helping you, or helping somebody else also in danger. I shall be concerned with which of you I help. But provided I end up helping the right person, what does it matter to me which of you I am motivated to help? It might be argued that a rational deliberator needs to be able to resolve conflicts among his first-order desires - and resolve them in favour of performing the action which he judges desirable. Perhaps second-order agency, motivated by practical judgments qua secondorder desires, is necessary to achieving such a resolution. A rational deliberator could use such an agency to reinforce his first-

Scepticism about second-order agency

39

order motivation for the actions he judges desirable. A rational deliberator, Velleman has argued, must be motivated by a desire to act in accordance with reasons - a motive which Velleman takes to be second-order: What it is for this motive to operate is just this: for potential determinants of behaviour [first-order desires to act] to be critically reviewed, to be embraced or rejected, and to be consequently reinforced or suppressed. ('What happens when someone acts?', p. 480) But why do rational deliberators need second-order desires to resolve their first-order desire conflicts? The conflicts could perfectly well be resolved at the first-order level - and resolved rationally, in favour of performing the actions judged desirable - without second-order desires being involved at all. An agent's practical judgments could simply be beliefs that certain actions are desirable. Then, for desires to act otherwise to be overridden, thereby rationally resolving all conflict, these beliefs would merely have to motivate the agent to perform only desirable actions. In other words, the agent need only hold a strong enough first-order desire to perform only desirable actions. Practical judgments and decisions have a first-order content. At least so things appear, and, until we have compelling reason for doing otherwise, we should go with appearances. Making judgments or decisions about how to act implies nothing more than a concern with what those judgments and decisions are about — a first-order concern with action. Second-order agency characteristically has afirst-order motivation.

It may yet turn out to be true that any agent who is capable of making practical judgments, and basing decisions on them, must have a sufficiently sophisticated conceptual capacity. He may in particular need to have a concept of desire, and so at least a capacity to hold second-order desires. For a capacity to deliberate about what actions it is desirable to perform may well require that one can conceive of oneself as an agent — and so also that one can conceive of oneself as motivated by desires. But even if this does turn out to be true - and I shall not argue the point here - that does not imply that a practical deliberator's actual concerns ipso facto extend, in every case of deliberation, beyond his actions to how he desires to act. Theorists who identify practical judgments with second-order desires, might be right about the capacities required

40

The psychology of freedom

for making practical judgments. But that does not make them right about what practical judgments are. But if practical judgments are not second-order desires, then they are not desires at all. For we have already rejected as implausible the view that practical judgments are first-order desires. Hence practical judgments must indeed be beliefs - beliefs which motivate us to act in so far as we have a sufficiently strong desire to do whatever it is most desirable to do. Does the will ever take a second-order form?

For Frankfurt free will involves the explanation of will-constituting attitudes by further attitudes of the same kind - but in secondorder form. Freedom of action involves the dependence of our action on our will. Correspondingly, freedom of will involves the dependence of our will on a higher order version of the same. But can the attitudes which compose the will as we ordinarily conceive it - our intentions to act - be explained by a higher-order version of the same? Consider my intention to go to the dentist. Is it plausible that I should have formed this intention on the basis of a higher-order intention to hold it? Surely this is not how our intentions are formed. We do not form intentions to go to the dentist on the basis of prior intentions to form those intentions. I can form an intention that tomorrow at io o'clock I shall go to the dentist, and then tomorrow, on the basis of that intention, act as I have intended. But I cannot now form an intention that tomorrow at io o'clock, I shall then form an intention to go to the dentist — and then, simply on the basis of today's higher order intention, tomorrow at io o'clock form the intention intended. We do not form particular intentions to act on the basis of prior intentions to form them. Our actions may be explained by, and subject to, the will. But the attitudes which, as we ordinarily suppose, make up the will - our intentions to act - are not themselves explicable in the same way by higher-order intentions. The will is not subject to the will. It is a considerable puzzle why our intentions are not subject to the will, and I shall be producing an explanation in a later chapter. But it is worth noting that not only are our intentions not subject to the will. It seems that many of our other action explanatory attitudes are not subject to it either. I cannot intend today to form

Scepticism about second-order agency an intention tomorrow to go to the dentist, and expect to form that intention tomorrow simply on the basis of today's intention to do so. No more can I intend today that tomorrow I shall form a desire to go to the dentist, or intend today that tomorrow I shall come to think that going to the dentist would be desirable, and expect to form that desire or make that practical judgment tomorrow simply on the basis of my present intention to do so. Actions may be subject to the will. But it is a quite general characteristic of our intentions, desires, and practical judgments that they are not. We may be able to act as we decide. But we cannot intend things, or want things or think them desirable simply as we decide. Summary

On Frankfurt's conception of it, second-order agency is clearly very different from the way we ordinarily conceive it to be. Not that this fact makes our ordinary conception of second-order agency any easier to defend. For much of the appeal of Frankfurt's account of second-order agency lies in the fact that it faithfully generalises what we know to be true of action. Frankfurt straightforwardly models second-order agency on the first-order kind. First-order action is motivated purposively, by desires to perform it. So too, with its second-order motivation, is Frankfurt's second-order action. First-order action is subject to the will. Again, so too is Frankfurt's second-order action. But second-order action as we ordinarily conceive it is neither of these things. And that, as I warned in the last chapter, is precisely what generates so much philosophical scepticism about second-order agency. From Hobbes onwards, many philosophers trying to devise theories of agency have noticed the differences between decisions to act, as we ordinarily conceive them, and the actions which those decisions explain. And they have inferred from these differences that decisions to act are not actions themselves. And so we arrive at two further sceptical arguments. THE ARGUMENT FROM NON-PURPOSIVENESS

I have claimed contra Frankfurt that the reasons which move us to take decisions and form intentions are, characteristically, the very

41

42

The psychology of freedom

same reasons as would motivate our performance of the action intended. An intention to do A is explained by beliefs about what ends doing A might attain, combined with desires for those ends. But then we arrive at a second ground for scepticism about second-order agency as we ordinarily conceive it. This is the nonpurposiveness of intention formation. The deliberation which leads us to form intentions to act is, characteristically, deliberation about which action to perform - not deliberation about which intention to form. In deliberating we simply consider the actions between which we must decide, and what ends those actions might further. And so the formation of particular intentions to act is explained by beliefs about what ends the actions intended might attain, and not by beliefs about what ends those intentions might themselves attain. What, once we have deliberated, normally leads us to form an intention to do A, are purposes for doing A - for performing the action intended - not purposes for forming and holding the intention to do A itself. Actions are performed as means to ends. But intentions to act do not seem to be formed as means to ends. As for intentions, so for desires and practical judgments. The deliberation which leads us to form desires to act, or to arrive at judgments about which actions are desirable, is again, simply, deliberation about how to act. We form these desires and judgments just on the basis of considering what ends would be furthered by the relevant actions - and not on the basis of considering what ends would be furthered by our forming those desires or making those judgments. Our reasons for desiring to do A, or for judging that it would be desirable to do A are, simply, reasons for doing A. Our desires and practical judgments are non-purposive too. And this might make one doubt whether the formation of particular intentions, or indeed the formation of particular desires or practical judgments, could possibly count among what we deliberately do. For is not purposiveness a condition of agency? Our first-order actions count as agency if anything does. And it is their purposiveness, I have claimed, which clearly establishes these actions as doings. To do something as a means to an end is precisely to do something. It seems, then, that what is distinctively active about our first-order actions, is the way in which they are motivated. They are motivated by rationalising psychological attitudes in a distinctively practical way - by attitudes which rationalise them as

Scepticism about second-order agency

43

believed means to desired ends. If our intentions, desires, or practical judgments are not likewise motivated in this distinctively practical way, why should we suppose that forming these attitudes can be a deliberate action too? Davidson has proposed this very argument from Non-Purposive ness. He has noted that intention formation is not normally purposive as action is. And this has led him to be cool towards the idea of second-order agency. In reply to question, Is intending an action?, Davidson replies: The coming to have an intention we might try connecting with desires and beliefs as we did other intentional actions ... But the story does not have the substantial quality of the account of intentional action. ('Intending', p. 90) As Davidson puts it, if someone performs an action of type A for a reason, that he does A must be explained by the fact that the agent has: a pro attitude toward actions of type B . . . and a belief that in performing an action of type A he will be (or probably will be) performing an action of type B ... The description of the action provided by the phrase substituted for 'A' gives the description under which the desire and the belief rationalise the action . . . There must be such rationalising beliefs and desires if an action is done for a reason. ('Intending', pp. 86-7) Now in Davidson's view, intentions are formed, and formed for reasons — explained by beliefs and desires which motivate and rationalise them as intentions - without being motivated and rationalised by beliefs about what one will be doing informing those intentions to act. The beliefs which motivate and rationalise the intention are, rather, beliefs about what the agent will be doing in performing the action intended:

The reasons an agent has for intending to do something are basically of the same sort as the reasons an agent has for acting intentionally [i.e. doing that intended something]; they consist of both desires (and other pro attitudes) and beliefs. If someone intends to polish his right shoe, it must be because there is some value he wants to promote by polishing his right shoe (perhaps he has already shined his left shoe and wants the two to match), and he believes that by shining his right shoe he has a chance of promoting what he wants. ('Replies', Essays on Davidson, pp. 213-14 (my emphasis)) In Davidson's view, then, we are not performing any action in forming a particular intention to act. And that is because we are

44

The psychology of freedom

not motivated to form the intention purposively, as a means to an end. THE REGRESS ARGUMENT

We noted, contra Frankfurt, that intentions and decisions to act, indeed many of our action-rationalising attitudes generally, are not subject to the will. We cannot take decisions or form intentions to do A simply by deciding so to decide or intend. Hobbes shared this intuition, and sought to use it in a Regress argument against the possibility of freedom of will. Hobbes began with the plausible assumption that, in order to have freedom of action, one must have a capacity to act as one wills. He inferred that, therefore, in order to have freedom of will, one must have a capacity to will as one wills. But, Hobbes claimed, there is no such capacity: I acknowledge this liberty, that I can do if I will, but to say, I can will if I will, I take to be an absurd speech. ('Of liberty and necessity', p. 240) In Hobbes' view, our freedom of will - of decision - depended on the possibility of second-order decisions - decisions to decide determining our decisions to act in the same way that our decisions to act determine our actions: The question is whether the will to write, or the will to forbear, come upon a man according to his will. ('Of liberty and necessity', p. 240) And Hobbes' view, from which he concluded to a denial of freedom of decision, was that second-order decisions cannot determine our decisions to act in the same way that our decisions to act determine our actions. Whether we accept Hobbes' conclusion, his premiss that the will is not subject to the will - seems thoroughly sound. Hobbes' argument is against freedom of will. But surely it is also an argument against a second-order agency of the will. For if a second-order freedom of will is impossible, not contingently, but of necessity, and is impossible because of the very nature of the will itself, what is left of the idea that taking a particular decision or forming a particular intention is something which we deliberately do? Agency does not imply the actual possession of control over whether we perform it. We can lack the freedom to do otherwise. What we are doing may be explained by some addiction or

Scepticism about second-order agency compulsion. Or we may be being manipulated into doing it. But all these conditions that take away our control over our agency-which take away our freedom to do otherwise — are contingencies. What Hobbes is alleging is that a second-order control over our intentions is a control which we lack, not contingently, but of necessity, because of the very nature of intention and intention formation. But if Hobbes is right, then there can be no such thing as a secondorder agency of intention. It must at least be consistent with the nature of anything which counts as a genuine doing, that one should have control over whether one does it. If whether we X lies, by X-ing's very nature, outside our control, then that we X must be something which happens to us, and not something which we can deliberately do. WHY DO WE BELIEVE IN SECOND-ORDER AGENCY?

We believe in a second-order agency of the will. That we take a particular decision to act - that we form a particular intention - is as much our deliberate doing as is the action which that decision explains. Yet intention formation seems, in certain crucial respects, to be quite unlike action, and very much more like intuitively passive occurrences, such as run-of-the-mill desire formation. Actions are purposive and may be subject to the will. But intention formations, like desire formations, are neither purposive, nor are they subject to the will. The implication of the Non-purposiveness and Regress arguments seems clear. We believe that the taking of particular decisions counts amongst what we deliberately do. But nothing else in our conception either of the will or of agency seems to support that belief. And this is a marked contrast with the case of first-order action. Our belief in the agency of what we count as our firstorder actions has a reasonably clear basis in our beliefs about what those occurrences are like in other respects. It has a clear basis, in particular, in our beliefs about how those occurrences are motivated and explained. There seems no comparable basis for our conviction that decision-making is something we do. Where then might our belief in second-order agency come from? Bernard Williams has noted the tendency for philosophers to introduce a second-order agency of the will into their theories of action - to take the view that:

45

46

The psychology of freedom

The self can act (at one time rather than another, now rather than earlier) only by doing something - the thing it does, willing. ('Nietzsche's minimalist moral psychology', p. 71) Associated with the view that these philosophers have of the will, then, is what Williams calls a 'doubling of action'. To perform any ordinary action we have first to perform another action - a willing. And Williams is convinced that it is moral theory - and mistaken moral theory at that - which gives rise to this belief in an actiongenerating agency of willing: We start with a supposed psychological phenomenon, willing, associated with the conception of the self in action . . . Since the picture [of the will] is neither coherent nor universal, yet has this authority, we need to ask where it comes from and what it does. It is not manifestly tied to morality, offering rather a picture of voluntary action in general, but there is a moral phenomenon, a certain conception of blame, which it directly fits . . . The fit between the special psychological conception and the demands of morality enables us to see that this piece of psychology is itself a moral conception, and one that shares notably doubtful features of that morality itself. ('Nietzsche's minimalist moral psychology', p. 74) But belief in a second-order agency of the will is not a philosophical idiosyncracy. It is a part - as I shall argue, a perfectly coherent part - of the common-sense psychology to which we are all committed. And this very general belief in second-order agency is not a 'moral conception5 either. It is not the result of a dubious intrusion of ethics into action theory. Belief in second-order agency certainly does not need some dodgy moral theory to motivate it. Contrary to appearances so far, our belief in second-order agency fits perfectly well with, indeed even follows from, our other beliefs both about the nature of the will and about agency in general. These other beliefs do indeed have a normative content. But, as I have already claimed, this is because our ordinary conception both of agency and of the will are part of a wider theory of our rationality - the theory that forms much of our common-sense psychology. And this theory of rationality has nothing to do with blame, guilt or any other distinctively moral concepts. A full defence of our ordinary conception of second-order agency is going to take up much of the remainder of the book. But an outline of the general strategy can be given here.

Scepticism about second-order agency

47

AGENCY AND PRACTICAL REASON

Davidson's theory of agency is a theory of its motivation. Agency is distinguished by the fact that its motivation takes a distinctively practical form. Active actions are distinguished from what Davidson regards as passive intentions and desires, by the fact that actions are performed as means to ends. Whilst, characteristically at any rate, intentions to act are not motivated by purposes for so intending, any more than desires to act are motivated by purposes for so desiring. Davidson's assumption that intentions to act are characteristically motivated non-purposively - by the same psychological attitudes as would explain and rationalise performing the action intended - seems right. The deliberation which guides our intention formation is indeed characteristically about nothing other than how to act. A decision to go to the dentist can perfectly well be taken on the basis of nothing more than, say, a judgment that it is desirable to cure one's tooth-ache, and a belief that going to the dentist would be a means to that end. A convincing defence of the intuitive agency of such a decision, then, should be consistent with its equally intuitive nonpurposiveness. If one has taken a decision to go to the dentist, taking that particular decision is going to be something which one has deliberately done. But it does not follow that in taking it one has done anything purposive - that one has done anything as a means to some end. So, to defend the agency of the will on such terms, we must abandon Davidson's theory of agency as essentially purposive. What other warrant might there then be for our belief in the agency of our decisions to act, if not the way in which we have been motivated to take them? The warrant is unlikely to lie in the phenomenology of decision-making. For Dennett's observation that this phenomenology is all too elusive is, I think, perfectly just. To repeat his claim: From some vantage points [decisions] seem to be the preeminently voluntary moves in our lives, the instants at which we exercise our agency to the fullest. But those same decisions can also be seen to be strangely out of our control. We have to wait to see how we are going to decide something, and when we do decide, our decision bubbles up to consciousness from we know not where. We do not witness it being made; we witness its arrival. (Elbow Room, p. 78)

48

The psychology of freedom

We certainly believe that taking a particular decision is something we do. But I doubt that there is any phenomenological feature of our decision-making that gives rise to and justifies that belief, and which is generally present when we take a decision to act. As Dennett observes, our decisions may perfectly well seem to arrive or bubble up. Is there really an 'actish feel' that we experience, not only whenever, say, we deliberately climb the stairs, but even whenever we simply decide so to act? It is a little implausible to suppose that there need be. At any rate, in my case, and I have no other to go on, deciding to climb the stairs does not feel anything like actually climbing them. Our clear conviction that taking decisions to act is something we do is surely not based on something so elusive as feeling. Of course, we are aware of our intuitively active decisions as they occur. But equally, and in the same way, we can become aware of many of our intuitively passive desires as they occur. In both cases, the relevant mental event occurs; and simply because it has occurred, one comes to believe in its occurrence. The conviction that, unlike much desire formation, taking a particular decision is something which one does must, then, have some basis other than in feeling or mode of awareness. The basis for our belief in the active nature of our decisions lies, I suggest, not in the purposiveness of our decisions - in how our decisions have been motivated — nor in what decision-making feels like or in how we become aware of it. The basis lies, instead, in the rationality of our decisions - in the way in which decision-making is governed by reason. What quite generally distinguishes agency from non-agency is how agency is justified. Philosophers have long discussed the nature of what they call practical reason. And the naturalness of this terminology suggests that reason or rationality might take a distinctively practical form. But perhaps it does. An hypothesis that I shall be adopting is that there is indeed a distinctively practical rationality — a distinctive way in which reason or rationality governs our agency or practice. The true theory of agency, then, specifies agency in terms of distinctively practical canons of justification: what all agency has in common is being governed by these canons. So if it turns out that, say, desire formations are passive, but intention formations and actions are uniformly active, then on my theory that will be because of a fundamental difference between

Scepticism about second-order agency desire rationality on the one hand, and intention and action rationality on the other. There is going to be one non-practical way in which justifications for desires are determined; and then, whatever other differences between intention and action rationality there might be, there is going to be another uniformly practical way in which justifications both for intentions to act and actions are determined. But if the way agency is governed by reason is distinctively practical, it need not follow that the way in which that agency is actually motivated and explained must be distinctively practical too. Intentions to act may be governed by practical reason, as actions are, even in cases where the formation of those intentions is motivated in a non-purposive and desirelike way. What is rationality in its distinctively practical form? By way of answer, let me give a general intuition - an intuition which any characterisation of practical reason should respect — and then I shall make a specific proposal. The general intuition is one which we have already mentioned in connexion with Hobbes' Regress argument. This is the intuition that the nature of anything which counts as our agency - as something we deliberately do - must at least be consistent with our actually exercising control over and through its occurrence. Not that the performance of agency guarantees either the actual possession of control or its exercise. Agents can be doing things without actually being free to do otherwise - without actually possessing or exercising control over what they do. They might be subject to compulsion or manipulation, or they may otherwise fail to meet various conditions on freedom which we shall be examining in the chapters which follow. But, if some occurrence - say, the taking of a decision - does count as a genuine doing, then the nature of that occurrence, including the way that it is governed by reason, should not rule out its ever constituting the exercise of control. Perhaps then a distinctively practical rationality is that kind of rationality that must govern any exercise of control. THE MEANS-END MODEL OF AGENCY

How, in very general terms, does reason govern the first-order agency of our action? What rationally justifies action is its likely desirability - which consists, in turn, in the fact that it is likely to further desirable ends. If you want to persuade anyone into

49

50

The psychology of freedom

performing an action, you will have to persuade them that their performing the action would or might have some desirable outcome. What more precisely the desirability of ends might come to we can for the moment leave an open question. Whatever the desirability of ends does consist in, actions, at any rate, are means-endjustifiable - they are justified as likely means to desirable ends. In chapter 5 we shall be considering other ways in which reason might be thought to take practical form. But the most obvious candidate for constituting reason in its distinctively practical form is, precisely, means-end justifiability. Perhaps reason governs all our agency, the taking of particular decisions to act and the actions decided upon alike, by reference to the desirability of the ends which performing it is likely to further; and our agency just is those motivated occurrences in our lives which are so governed by reason. There is a compelling reason for supposing that means-end justifiability is indeed reason in its distinctively practical, agencygoverning form. Remember that, as I have just observed, it is essential to agency, and distinctive of it, that it provide the possibility of exercising control. So reason as it governs our agency must be reason as it governs the exercise of control. But means-end justifiability just is reason as it governs the exercise of control. For what does the rational exercise of control involve, but producing outcomes which are desirable, and preventing outcomes that are undesirable? And means-end justifiable doings are made rational precisely by the fact that they would produce desirable outcomes and prevent undesirable ones. Notice that there is an important link between means—end justifiability and purposiveness. If X-ing is means-end justifiable - if X-ing can be justified by the fact that it would further some desirable end - then it must surely be possible for an agent rationally to X on that basis, motivated by the belief that X-ing would further that end. That X-ing is justified as a means to ends does imply, then, that X-ing can also occur in purposive form, motivated by beliefs about what ends X-ing would further. But, as I shall show shortly, occurrences that are means-end justifiable can be motivated in non-purposive form as well. From the facts that X-ing is means-end justifiable, and that an agent is motivated to X, it importantly does not follow that the agent is ipso facto being motivated by beliefs about what ends X-ing would further - that the

Scepticism about second-order agency agent is X-ing as a means to an end. The doctrine that agency is means-end justifiable, then, does accommodate and explain the common intuition that there is an important connexion between agency and purpose - between doing something, and doing it as a means. But the connexion may be looser than Davidson supposes. Doing something may imply merely the possibility of doing it as a means. If means-end justifiability is reason in its distinctively practical form, then we can use it to define agency. We can adopt what I shall call the Means-End model of agency and agency rationality. On this theory of agency, some sort of motivation by rationalising psychological attitudes is still necessary to the occurrence of agency, as Davidson has supposed that it is. So an agent still counts as active - as actually doing something - only if (a) for some X, they are X-ing for reasons - and so are being motivated to X by psychological attitudes which in some way or other rationalise their X-ing. But agency is no longer distinguished from, say, motivated but passive desire formation by the manner of its motivation. Purposiveness - Davidson's distinctively practical mode of motivation - is no longer necessary to the occurrence of something we do. Just like non-agency, agency can be motivated in non-purposive form. We are characterising and distinguishing agency in terms of its distinctively practical justification, not its distinctively practical motivation. Given that condition (a) is met, then, what makes an agent's motivated X-ing suffice for genuine agency, according to the Means-End model, is its being true that (b) justification for the agent's X-ing depends on the likelihood that X-ing would further desirable ends. On this Means-End model of agency, an agent who has been moved to take a decision to do A by nothing more than beliefs about what ends doing A would further - whose reasons for intending to do A have been, simply, reasons for doing A - can still have acted in taking that particular decision. For though the decision to do A is not taken as a means to any end, the justification for taking it might still depend on the likelihood that deciding to do A would further desirable ends. Though as non-purposive in its motivation as any passive desire to do A, the decision to do A might still be means-end

51

52

The psychology of freedom

justifiable. In which case in taking that particular decision the agent could be exercising, albeit non-purposively, a capacity for second-order action. This strategy for defending the agency of the will immediately raises two questions. First, the defence only works if, as I have claimed, there in fact can be agency which is means-end justifiable while being non-purposive in its motivation. But can we be moved to perform agency which is justified as a means to an end without ipso facto being motivated to perform it as a means to an end? If we cannot be, then the Means-End model of agency will not take us beyond Davidson's theory - and the intuitive non-purposiveness of decisions to act will be just so much evidence for their not being means-end justifiable either. Secondly, is it plausible that decisions to do A, in particular, are means end-justifiable - that the taking of particular decisions to act is justified as a means to ends? For unless decisions to act can be shown to be justified as means to ends, we may have a new account of agency in general; but it will not be much use to the defence of the agency of the will. DECIDING AND TRYING

There is at least one kind of agency which, though means-end justifiable, can be - indeed, often is - motivated in non-purposive form. The agency in question seems very like that of taking a decision to act. Indeed, the two forms of agency have often been identified. What happens when I purposively perform some bodily action — when, for example, I purposively move my hand? Many philosophers would maintain that, when I move my hand, there is something else that I do - a form of agency which I perform prior to my hand's moving, and which causes my hand to move. This agency is a conation - what we would ordinarily describe as my trying to move my hand. And so for bodily action - one's purposive moving of one's limbs and body parts - generally. Whenever I perform a bodily action, I shall perform a conation. There is going to be a way in which I am trying to move my body; and the bodily action is performed provided that, as a result, my body then moves in the appropriate way.1 1

See Jennifer Hornsby's Actions for a sustained argument for this view.

Scepticism about second-order agency One argument for the occurrence of these conations is as follows. Suppose someone purposively moves their hand. Had unbeknown to them their hand been restrained at the time of the action, so that they were prevented from moving it, they would still have done something. They would still have tried to move their hand. But then surely they do this anyway, whether or not their hand is restrained. Suppose then that conations do occur when we perform bodily actions, as they certainly do when our limbs actually are restrained - when we try to move them but fail. When conations occur then, whether successful or not, they count amongst what we do. If I try to do something but fail, I shall at least have done something. And so I shall have performed an action - not the bodily action which I attempted, but an action none the less. And what I have done will have been done for reasons. Beliefs and desires will have moved me to attempt the bodily action, rationalising or rendering my attempt intelligible. But, though motivated actions themselves, conations need not be purposive. My reasons for trying to move my hand may be, simply, reasons for moving my hand. The attitudes that move me to try to move my hand, and render my making such an attempt intelligible, may simply be desires for ends and beliefs about how moving my hand would further those ends. I may be moved to try to move my hand simply by my desire to signal, and my belief that to move my hand would be to signal. For when one tries to move one's limbs, one need not consider, in particular, the consequences of trying to move them. One's beliefs, if one holds any - and it should not be assumed in every case that one need actually have formed a n y - about what will happen if one tries, need play no role in the motivation of one's attempt. For one's attempts at moving one's limbs are typically successful. So one can usually afford to ignore one's attempt at moving one's limbs and the question of whether it will succeed or fail, and make the attempt just on the basis of considering what outcomes actually moving one's limbs would have. Attempts at bodily action, therefore, may intelligibly be based solely on one's beliefs about the possible consequences of the relevant bodily action. One need not have considered the attempt and its possible consequences at all. Tryings, then, can provide clear cases of non-purposive agency. When we try to perform some bodily action A, our trying to do A is itself an action which we are motivated to perform. But the

53

54

The psychology of freedom

deliberation which motivates us to try to do A may be, simply, deliberation about whether to do A. It need not be deliberation about whether to try to do A. Tryings to do A are actions which may be motivated exactly as Davidson supposes decisions and intentions to do A to be motivated - by desires for ends and beliefs that doing A would further those ends. Conations, like decisions, are actions themselves. But again, as with decisions, we need to avoid confusion. This time the confusion to be avoided is between attempts at bodily action and the bodily actions attempted. From the fact that someone has tried to perform a bodily action, it follows that they have indeed performed an action. But, of course, it does not follow that they have performed the action attempted. Nor, as I have just argued, does it follow that they have done anything purposive. I need therefore to make it very clear when I am talking about conations, and when I am talking instead about the actions which conations explain. So, unless the term 'action' is being applied to conations explicitly, unqualified uses of it should not be understood to include conations. What then do these often non-purposive tryings all have in common with the purposive actions attempted? One thing they do have in common is, crucially, their rationality. Tryings, whether purposive or non-purposive, and the actions attempted are all alike means-end justifiable occurrences. Each activity is justified in terms of desirable ends which doing it is likely to further. In most cases there is little difference between the ends likely to be furthered by trying to move my hand, and the ends likely to be furthered by actually moving it. And so I make my attempt simply on the basis of deliberating about whether to perform the action attempted, and not on the basis of deliberating about whether to attempt it. My attempt is made non-purposively. But that does not alter how the justification for my making the attempt is determined. It is determined by reference to the ends likely to be furthered by making the attempt, and not by reference to the ends likely to be furthered by the action attempted. Even where these two sets of ends are more or less the same, that still makes the attempt means-end justifiable. Trying to do A is not justified simply in terms of the likely desirability of doing A. To see this, just consider cases where there is a discrepancy between the ends likely to be furthered by trying, and the ends likely to be furthered by succeeding. No matter how likely it is that

Scepticism about second-order agency moving my hand would further desirable ends, I may yet have little justification for trying to move it if the likelihood of the attempt being a disastrous failure is high. And that is precisely because justifications for trying depend on what ends the trying is likely to further, and not on what ends the action attempted is likely to further. Is there a possible parallel here between trying and deciding - between conation and will? Decisions and tryings seem quite similar. Compare my decision to move my hand with my attempt at moving it. Each is a kind of agency - something I do - which can cause my hand to move, and thereby explain the successful performance of a first-order action — my action of moving my hand. Indeed, each form of agency is identified in terms of its explanatory relation to this very first-order action. Moreover, each form of action-generating agency is typically motivated and rationalised by just the attitudes which would motivate and rationalise the firstorder action it generates. My reasons for deciding or trying to move my hand are, typically, nothing but reasons for moving my hand. Perhaps, then, deciding will turn out to be, like trying, a form of means-end justifiable, but characteristically non-purposive agency. Perhaps, indeed, deciding and trying are one and the same form of agency. And so some people certainly have thought. Much talk in early modern philosophy about volition and willing made no distinction between someone's taking a decision to act, and their trying to act. Consider, for example, Locke's theory of the will. In Locke's view the will was an 'active power' by which we cause both mental and bodly actions. The will was a power to begin or forbear, continue or end several actions in our minds, and motions of our Bodies, barely by a thought or preference of the mind ordering, or as it were commanding the doing or not doing such or such a particular action. (An Essay Concerning Human Understanding, p. 236) We exercise the will by performing mental acts of volition: Volition, 'tis plain, is an Act of the Mind knowingly exerting that Dominion it takes it self to have over any part of the Man, by imploying it in, or withholding it from any particular Action. (An Essay Concerning Human Understanding, p. 241)

And it is clear that Locke made no serious distinction between deciding to perform an action and trying to perform it. Both kinds

55

56

The psychology of freedom

of agency were, in his view, equally cases of volition. In the case of a bodily action a Lockean volition shares many of the characteristics of an attempt. When I move my hand, the action, according to Locke, is performed in virtue of the fact that a volition that my hand move causes my hand to move. On the other hand in other respects volitions are more like intentions or intention formations. A volition, according to Locke, is a kind of action explanatory preference that the action be performed. That makes it a kind of action explanatory psychological attitude. Moreover, like a decision or intention, a volition can sometimes occur long before the motion which it causes - though Locke thinks that most volitions occur, as do tryings, at or around the time of action: For considering the vast number of voluntary Actions, that succeed one another every moment that we are awake, in the course of our Lives, there are but few of them that are thought on or proposed to the Will, till the

time they are to be done. {An Essay Concerning Human Understanding, p. 246)

Locke's failure to make any distinction at all between decidings and tryings must, however, be a mistake. Whatever the similarities between deciding and trying turn out to be, deciding and trying are still importantly different kinds of agency. The decision to move my hand is the formation of an action explanatory psychological state - it is the formation of an intention to move my hand — and so is a case of second-order agency, separated from the action decided upon and the hand motion alike by that intervening intention. The decision to move my hand, then, causes my hand to move by way of influencing my intervening motivation - by way of causing me to intend, and so be motivated, to move it. Contrast the attempt at moving my hand. The attempt is not the formation of a persisting psychological state. Trying to move my hand does not cause an intention to move my hand which then persists to cause the action. What my trying to move my hand does cause, instead, is extra-mental nerve signals and muscle motions and it is simply by means of these nerve signals and muscle motions that the conation then causes my hand to move. Since it does not constitute the active formation of any action rationalising psychological state, an attempt at moving one's hand is, like the action of moving one's hand, a case of[first-order agency. Trying to move my hand, then, is motivation-independent in its

Scepticism about second-order agency effects. It causes my hand to move other than by way of causing me to be motivated to move my hand. Unlike decisions, then, conations take effect in a motivation-independent way. But note - it is quite consistent with this claim, and evidently true, that success in what one tries to do can, and often does depend on one's motivation over time. For trying to do something can consist not, as does trying to move one's hand, in a single conation, but rather in a whole series of such performed over time. In which case, the occurrence of all these conations, and so too one's success in one's attempt, will of course depend on one's continuing to possess the motivation which is needed to explain them all. But the distinction I have drawn between conations and decidings still holds. Each conation making up the attempt as a whole takes effect in a motivation-independent way. Consider my attempt at pushing a piano to the other side of the room. That attempt consists in, and depends for its success on, a whole series of conations - in my trying, at each stage of the attempt, to push the piano a little bit further across the room. As before, each of these conations takes effect, causing the piano to move a little bit further across, in a motivation-independent way by way of vigorous muscle motion, and not by way of causing some intervening intention. The agency which causes the continuing motivation which the occurrence of all these conations requires which causes me to intend, and so remain motivated, to push the piano across - is the second-order agency of my initial decision to push the piano to the other side of the room. We must carefully distinguish the first-order agency of these conations, each of which takes effect in a motivation-independent way, from the essentially motivation-affecting second-order agency of my initial decision to push the piano across - the agency which gave rise to the motivation needed for all these conations to occur. Some modern philosophers, such as Hornsby, have proposed that we identify a bodily action such as moving one's hand, when one succeeds in performing it, with one's successful attempt -with the conation which causes one's hand to move. And this identification of bodily actions with the successful attempts at performing them, is not obviously absurd. But it would clearly be absurd to identify bodily actions with the decisions to perform them. The actions and the decisions obviously have different causes and effects. For there is an action-explanatory attitude, the intention to act, which is

57

58

The psychology of freedom

an effect of the decision but a cause of the action decided upon. Decisions, as second-order doings, are indisputably mental events, like the states of which they constitute the formation. But, as firstorder agency, not second, conations may not be mental events at all. It is a sensible matter for discussion whether attempts at bodily action occur within the mind, or outside it - and whether their status as mental or non-mental is determinable a priori at all. Decisions and conations, then, are importantly different kinds of event. Decisions cause actions by influencing one's intervening motivation, and conations do not. And that fact may turn out to imply further differences in the rationality of these two kinds of event. Decisions and conations may even turn out to be governed by reason in ways that are radically different. There is one striking difference, in particular, between decisions and conations which - 1 shall be arguing - certainly does imply some concomitant difference in their rationality. We have seen that our decisions to act are not themselves subject to the will: that, for example, I cannot now decide that at 10 o'clock tomorrow I shall take a decision to move my hand - and then, simply on the basis of that decision to decide, expect that tomorrow at 10 o'clock I shall indeed take the decision to move my hand. But if decisions are not subject to the will, the same is not obviously true of our conations. There is nothing to stop me trying to move my hand on the basis of a prior decision so to try. I might, for example, be a physiologist interested specifically in what happens when I make a conation — in what goes on when I try to move my hand. So I might decide that at 10 o'clock tomorrow, under careful observation, I shall indeed try to move my hand - and then, at 10 o'clock tomorrow actually try to move it just on the basis of that earlier decision so to try. Now if it is true that conations are subject to the will when decisions to act are not, that is going to be because there are indeed some differences at least in the rationality of decisions and conations. For, as I have promised, I shall be developing a theory of the scope of the will — of what we can do just on the basis of deciding to do it. And this theory is going to appeal precisely to the differing rationality of various kinds of occurrence - actions, conations, decisions, desires and practical judgments - to explain why each kind of occurrence is or is not subject to the will. So, if one kind of occurrence, conation, is subject to the will, while another, deciding to act, is not, then, on my theory of what determines the

Scepticism about second-order agency will's scope, that is going to be because of important differences in what would rationally justify these two kinds of occurrence. It remains to be seen, then, if decisions and intentions to act really are means-end justifiable as both conations and actions are. But if decisions to act and the intentions they form are not means-end justifiable, how else might they be governed by reason? The obvious answer is that decisions and intentions might be governed by reason in exactly the same way as desires and practical judgments - occurrences which, as we have seen, resemble decisions and intentions in not being subject to the will either. In other words, intentions might be pro attitudes. PRO

ATTITUDES

If our actions are means-end justifiable, what justifies the desires and practical judgments that move us to act? A natural hypothesis is that our justifications for holding these attitudes lies simply in the desirability of their objects. Justification for desiring to do A, or for making a judgment that doing A is desirable, consists simply in the fact that, whether in some particular respect or overall, doing A is likely to be desirable - and so in the fact that doing A is likely to further desirable ends. Let us call apro attitude any psychological state which is rationally justified simply in terms of the likely desirability of its object. Desires and practical judgments, it seems, must be pro attitudes. What else could make it rational to desire to perform an action A, or to judge that it is desirable to do A, than the likely desirability of doing A? That is why the deliberation which leads us to make practical judgments or to form desires, concerns itself solely with the desirability of their objects. The deliberation which leads me to form a desire to do A or to conclude that doing A is desirable is going to be, simply, deliberation about what ends doing A might attain — the very same deliberation that would lead me to do A. Of course, there can be advantages to holding desires or to making practical judgments - advantages which consist in the desirability of holding the desire or making the judgment itself, and which have nothing to do with the desirability of the desire's or judgment's object. Suppose, for example, that it is in no respect desirable that you remain loyal to your boss in the future. Your boss

59

60

The psychology of freedom

does not deserve your loyalty, and you only stand to benefit from well-timed future disloyalty. It may still be desirable that, at least at present, you continue to want to be loyal to him, or that you continue to think loyalty to him to be desirable. For your boss may be very sensitive to what you now want and think desirable. Only by wanting to be loyal to him or by thinking loyalty desirable, will you appear to your boss to be loyal - and it is very desirable that, for the moment at least, you appear to be loyal. But practical judgments and desires are not like actions. They are not made rational by their own desirability - by the ends they themselves are likely to further. Even if I know that it would be useful to want to be loyal to my boss, or useful to think such loyalty desirable, that will not make me want to be loyal, or think loyalty desirable. And that is because I am moderately rational, and such considerations are quite irrelevant to the rationality of these desires and practical judgments. The deliberation that rationally guides judgment about the desirability of things is deliberation about, simply, the desirability of those things. Again, standard decision theory is not ignoring any justifications for our desires when it assumes, as it does assume, that rational desires are a simple function of the likely desirability of their objects - and never a function of the likely desirability of holding the desires themselves. The advantages to be got from holding a desire are simply irrelevant to the question of its justification. So if while deliberating I do notice some advantage to be gained from holding a desire, or from making a practical judgment, that will not get me to hold that desirable attitude. At best, my deliberation will lead me to desire to perform some action that would cause me to hold it. Now my deliberation can perfectly well lead me to form that desire to act because such action and so too the desire to perform it may be perfectly rational, even if the attitude which the action caused might not be. For actions, we have agreed, are justified by the advantages they are likely to bring - by their likely desirability; and action which caused the desirable desire or practical judgment would be desirable in turn. It might be argued that it is rational to perform an action which causes one to form some psychological state - and if the action is justified by the very desirability of that psychological state - then the psychological state caused must itself be rational. But this is not a plausible principle. Parfit has already provided a vivid counter-

Scepticism about second-order agency

61

example (see Reasons and Persons, pp. 12-13). Consider the following case invented by Schelling. A criminal might threaten to shoot my children, one by one, unless I give him the combination of my safe. It might therefore be justified for me to take a drug that makes me go crazy so that, for a while, I actually want the man to shoot my children. After all, to hold this desire may, in the circumstances, be very desirable. If the criminal realises that I am crazy enough to hold this desire, he will realise too that his threat has become pointless. But the desire itself is deeply irrational, because there is obviously nothing desirable in my children being shot by the criminal. It may be highly desirable for me to have the desire; and so it may be overwhelmingly rational for me to perform some action to cause myself to form it. But the desire itself is quite mad - quite without rational justification. CONCLUSION

In this chapter we have discussed a range of occurrences, mental and otherwise, that explain our actions - occurrences whose similarities to and differences from the actions which they explain are displayed in the following chart (see Table 1, The psychology of agency -a first sketch).

Part of the way along the decision process, we arrive at occurrences over which - intuition suggests - we exercise the same control as we have over the actions which the process eventually explains. That point seems to be the point of taking particular decisions to act - the stage of our second-order agency. We believe that it is within our control which actions we decide to perform, just as it is within our control which actions we perform. Is this belief in a second-order agency of decision-making warranted? It is easy to see why philosophers might well doubt that it is. As yet we have not shown that the taking of particular decisions to act shares any significant characteristics - whether purposiveness, subjection to the will or means-end justifiability with the actions that those decisions explain. The analogy with non-purposive conations — our attempts at bodily action — has shown one thing, at any rate. It has shown that the characteristic non-purposiveness of our decisions to act is not enough to prove their non-agency. For our conations are all actions or doings themselves, even though many of them are also motivated

62

The psychology of freedom

non-purposively. But that point does not on its own provide much positive support for an agency of the will. For, whether motivated purposively or not, our conations still so obviously count as doings or actions themselves precisely because they are, like the actions attempted, means-end justifiable occurrences. Like the actions attempted, attempts at action are justified as means to desirable ends. And that means, in particular, that even if many conations are not actually purposive, it is at least possible for them to take purposive form. We can perfectly well make conations or attempts at bodily action on the basis of prior deliberation about whether, if we made them, they would be successful. It is always possible, then, for our conations to be motivated, and quite rationally so, by beliefs about what ends those conations would further, and not just by beliefs about what ends the relevant actions attempted would further. But while conations can occur in purposive form, this is not so obviously true of decisions to act. And there is one further striking difference which we have noted between decisions to act and conations - a difference that might well imply a rather deep difference in their rationality. This is the fact that while conations seem to be subject to the will - it looks as if we can make attempts at action just on the basis of deciding to make them - decisions and intentions seem not to be. And in this decisions and intentions resemble our non-means-end justifiable desires and practical judgments. That decisions and intentions are like desires and practical judgments in not being subject to the will, suggests then that decisions and intentions to act may not be means-end justifiable at all. Like desires to do A and practical judgments that doing A would be desirable, intentions to do A could instead be pro attitudes. They could instead be justified simply in terms of the desirability of doing A. In which case decisions to do A would seem to have very little in common with the actions A which they explain — not their motivation, not their relation to the will, and not even the general form of their rationality. Scepticism about second-order agency has certainly not been refuted yet. As we shall see, there are going to be further and, apparently, very compelling reasons for adopting what I shall term a. Pro Attitude model of intention rationality — for holding that, like desires and practical judgments, intentions are pro attitudes. Further support

Scepticism about second-order agency Table i. The psychology of agency - a first sketch Judging that Desiring to it is desirable move one's to move hand one's hand

Deciding to move one's hand

Trying to move one's hand

Moving one's hand

Yes?

Yes, as first order agency

Yes, as first order agency

Typically not

Possibly, but typically not Yes

Yes

Yes

Yes

No?

No?

Other relevant features Purposive? No

• No

Means-end justifiable?

No

No

?

Subject to the will?

No

No

No

Action?

Yes

for the Pro Attitude model is provided by the powerful intuition that decisions and intentions to act have the purely executive function of applying reason as it concerns first-order action. The point of taking decisions about which actions to perform, surely, is to ensure that one performs the right actions thereafter. And it is that fact that makes a Pro Attitude model of decision and intention rationality look so plausible. In the next chapter we shall be considering further this executive function which decisions and intentions have in relation to the actions which they explain. For this executive function is of vital importance to our freedom of action.

CHAPTER 3

Decision-making andfreedom

DEPENDENCE

So far we have been discussing whether there is a second-order agency of the will - whether taking a particular decision to act is something which we deliberately do. But what hangs on our possessing such an agency of the will? Hobbes thought nothing much - and certainly not our freedom of action, of which, as we have seen, he gave a strikingly economical account: A FREE-MAN, is he, that.. As not hindred to doe what he has a will to . . . from the use of the word Free-will, no Liberty can be inferred of the will, desire or inclination, but the Liberty of the man; which consisteth in this, that he finds no stop, in doing what he has the will, desire, or inclination to doe. {Leviathan, p. 146)

In Hobbes' view we lack freedom of action only to the extent that hindrances would prevent the successful execution of our decisions or, what came to the same thing for Hobbes, the satisfaction of our desires. On the Hobbesian theory, then, there is no dependence of our action control on any prior control over which actions we decide to perform. No agency of the will is needed for freedom of action. But that is not what we ordinarily suppose. Our freedom to act otherwise than as we actually do does seem to us to depend on our also having a freedom to decide otherwise than as we actually do. The decisions to act which precede our actions often causally determine how we shall act, or at least render the action decided upon very highly probable; and, in determining how we shall act, the same decisions also give us clear forewarning of the action determined. Now, for whatever reason, it is hard for us to view such 64

Decision-making and freedom

65

decisions as passive non-agency - as events which just happen to us and so lie outside our control - without viewing the actions which those decisions determine and of which they forewarn us as being outside our control also. An agent must have control of the decisions which determine how he acts, if he is to have control over his actions. I shall refer to the claim that our freedom of action depends on a prior control over which actions we decide to perform as Dependence. Whatever Hobbes might have supposed to the contrary, Dependence does seem to us to be true. The last two chapters showed that our ordinary conception of our agency is not Hobbesian. We conceive of our agency as occurring, as decision-making, in second-order form; and this despite the fact that our decisions differ from the actions which they explain in being, apparently, neither purposive nor themselves subject to the will. Our intuition that Dependence is true is one important sign that our ordinary conception of freedom is not Hobbesian either - that the Hobbesian conception of freedom is weaker than the conception which we ordinarily deploy. But intuition alone will not be enough to refute Hobbes' theory of freedom, any more than intuition alone could refute his theory of agency. In both cases intuition needs the defence of its own rival theory - a theory which justifies our naturally anti-Hobbesian convictions by explaining why they are true. It is easy to see why the case for Dependence, in particular, cannot rest on the mere intuition that Dependence is true. After all, at some stage our control of what happens, if we have any at all, must simply begin. If we are to avoid a vicious regress, our having control of a given event cannot in every case depend on our having control of a further and prior event. If freedom of action does depend on a freedom of prior decision, that had better not mean that freedom of decision depends, in turn, on a yet further freedom of something else prior to decision, and so on backwards in time. But then if our control of what happens has to start somewhere, why cannot it start with our actions? Why should our control of our actions depend at all on our having control of our prior decisions to act? Hence the truth of Dependence cannot be a brute fact. If Dependence does hold, we badly need some explanation for it an explanation which avoids regress by appealing to something

66

The psychology of freedom

peculiar to the relation between decisions to act and actions. In this chapter I shall begin to lay the foundations for such an explanation - an explanation which will be given in full in the next chapter. To obtain our explanation, we shall obviously need to develop a theory of freedom. But we shall also need to develop a theory of the function of decision-making in relation to action - an explanation of why actions are preceded and caused by decisions to act. We shall need to develop both these theories together. To take matters further, then, we need to examine the will's function in relation to action. We need to arrive at some idea, at least in outline, of the point of taking decisions and forming intentions to act. THE EXECUTIVE FUNCTION OF THE WILL, THE IDENTITY OF AGENTS, AND THE UNITY OF AGENCY

What role do decisions and intentions play in the life of an agent? The natural intuition, I have claimed, is that they serve an executive function in relation to our subsequent action. The function of taking decisions and forming intentions about which actions to perform, is to facilitate the application of reason as it governs those actions - to ensure that we end up performing the right actions. But how do decisions and intentions do that? The answer is that decisions and intentions are agency unifiers. They ensure that the actions which an agent performs over time have a rational unity: that the actions which an agent performs at any one time are those which he is justified in then performing justified, in particular, given his actions at other times. We do not only deliberate about our actions at the time we perform them. We commonly deliberate about our future actions too. By 'future actions', I do not just mean the future consequences of our present actions. I do not just mean, say, communicating a message to someone tomorrow by posting a letter now. By 'future actions' I mean actions, such as actually holding a conversation with someone tomorrow, which depend on ourfuture motivation - actions which we will perform only if, by the future time of action, we are then motivated to perform them. Deliberation about whether to perform these future actions is therefore immediately followed, not by the performance of any action - that is yet to come - but instead by a prior motivation-fixing decision to act. This decision ensures

Decision-making and freedom

67

that we remain motivated to act as we have judged that we should, thereby causing us eventually so to act. That we commonly deliberate about our future as well as our present actions - that much of our deliberation is immediately followed not by actions but by decisions to act - reflects our status as rational agent continuants. We are rational agents who persist identical through time and are wholly present at each time at which we exist. At any time, therefore, our concern for ourselves is not limited to what is then happening to us and to what we are then doing. Our present concern for ourselves is not limited to the present but extends, as we do, through time - and so, in particular, into the future. In the case of the future as opposed to the past, our concern for ourselves is practical. It is practical, not only because it is a concern about how we should act - our concern for the past, after all, can perfectly well be a concern about how we should have acted - but because it can and does make a difference to how we shall act. It is practical because it has effects on what we are concerned about. As rational agent continuants, it is our deliberations about how we should be acting in the future, as well as our deliberations about how we should be acting now, which most clearly count as exercises of our practical reason - a practical reason which is concerned with the future as well as the present. As agent continuants, our interests at any time depend on the course of our lives over time. Hence the value to us of an action in the present can and characteristically does depend on how we act at other times, including the future. So, as practical reasoners, we do not generally judge that we should be performing a given action now, without at the same time forming a view of what actions we should be performing in the future. Consider action which meets a basic need, such as searching for food. Finding and consuming food takes time; and food searching is pretty generally valueless unless it does lead eventually to the finding of food and its consumption. So the value to an agent as she searches of her then searching will typically depend entirely on her going on searching thereafter, and on her finally consuming the food once she has found it. This is so obviously true, that no rational agent would begin food searching on the basis of deliberation just about whether to search for food now. She will not deliberate only about the merits of

68

The psychology of freedom

performing that action now compared to other actions now. She will not ignore, even temporarily, how she should act in the future. Instead, she will deliberate about whether to carry out a food search - a series of actions, concluding in food consumption, which extends at least some way into the future. And she will have compared that series with rival extended series of actions - like going on working in her room, or visiting the cinema. She will, therefore, have derived any conclusion about how to act now from a prior conclusion about the relative merits of these temporally extended series of actions. So practical deliberation about how to satisfy basic needs is deliberation between series of actions extending into the future, the performance of which depends on the agent's future motivation. In other words, practical deliberation about basic need satisfaction is deliberation between what I shall term plans.1 But we do not need to consider only action aimed at need satisfaction. There is hardly any deliberatively based action which does not turn out, on examination, to have been based on deliberation between what are, in effect, alternative plans. Consider, for example, linguistic communication. Communication worth making generally requires a series of actions extending, at the time of deliberation, at least some way into the future, for example the successive utterances of a number of words or sentences. So the communication of a given content generally requires the execution, over time, of a communicative plan. Agents who deliberate about what content to communicate, about what to say, ipso facto deliberate between the rival communicative plans — the rival series of actions extending into the future - that are required to communicate rival contents. There are indeed some actions on which, as we perform them, we place a value which is largely independent of assumptions about our action at other times: consider braking to avoid hitting a 1

The notion of a plan is deployed in Rawls' A Theory of Justice, Harman's 'Practical reasoning', and then most recently in Bratman's Intention, Plans and Practical Reason and Gauthier's 'In the neighbourhood of the Newcomb predictor'. Everyone agrees that plans are sequences of agency extending through time. But otherwise opinions differ. Rawls' plans are life-plans - plans for a whole existence. But I do not think that plans so extended are what practical deliberation is characteristically about. Plans as I understand them can be as short as seconds or minutes. Gauthier's plans include second-order as well as first-order agency; whereas mine contain first-order actions alone. This difference between Gauthier and me reflects rather deep differences in our conception of the will. Gauthier's plans are discussed in chapter 6.

Decision-making and freedom

69

pedestrian, or jumping back to avoid a falling object. But actions such as these are hardly characteristic of our agency. For we conceive of ourselves as agent continuants persisting into the future. And so deeply a part of our ordinary thinking is this conception, that the range of serious candidates for action with which we start our practical deliberation - the actions between which we actually deliberate - almost invariably consists in rival extended, future action-involving plans. We have almost no underived, plan-independent practical concern for the present. Bradley claimed that there is a certain holism of practical reason: And, if we turn to life, we see that no man has disconnected particular ends; he looks beyond the moment, beyond this or that circumstance or position; his ends are subordinated to wider ends; each situation is seen (consciously or unconsciously) as part of a broader situation, and in this or that act he is aiming at and realizing some larger whole, which is not real in any particular act as such, and yet is realized in the body of acts which carry it out. ('Why should I be moral?', Ethical Studies, p. 69) Bradley was surely right in his claim that a single action at a given time characteristically has value as part of a larger whole - as part of a series of actions extending through time. Practical reason is characteristically concerned, not with the single action in the present but the plan - the series of actions extending, at the time of deliberation, at least some distance into or within the future, and depending on the agent's future motivation. So if our present deliberations do recommend that we now perform a given action and cause us now to perform it, they will be recommending that action as part of some plan - a plan which, if we are rational, those deliberations will cause us to perform in its entirety. At the same time as they recommend the present action, our deliberations are also going to recommend the future actions which form part of the same plan, and, through the will - through our decision and intention in advance of action — cause us to be motivated to perform those actions too. I have said that our deliberations about our future action count as fully practical because they can and do have effects on how we shall act. But it is also true that, as agent continuants, we count as practically rational at all only because, through the will, our deliberations - which primarily concern plans, and so concern our future actions as much as our present ones — affect our future actions as well as our present ones.

70

The psychology of freedom

Our practical deliberation must do more than have effects into the future. We must be left knowing what these effects will be. For at any time we can only put a part of any plan into action - that part containing the actions which are to occur at that time. Now, as we have seen, the value of a part of a plan generally depends on our executing the rest: it is useless to study for a qualification without taking the exam, to begin searching for food without continuing to search, to start attending to an argument without reading it to the end. So if we are to put part of a plan into action, we must be in a position to know that the rest of the plan will be executed; otherwise we shall not have reason even to execute that part. The same decision-making which ensures that we shall execute the future components of a plan must tell us that we shall indeed be executing those components, rather than components of some other plan. And, as I have already suggested, our decision-making does often let us know how we shall act. My decisions to act can inform me of how I shall be acting by leading me to believe that I shall act as decided. And this decision-based belief counts as knowledge because it is held under conditions - my having taken a decision so to act — which make that belief at least very probably or even certainly true. By giving me advance knowledge of how I shall act, my decision-making allows me to use that knowledge as input to further deliberation and decision leading to action. Having decided to holiday in France rather than Germany, my knowledge that I shall be holidaying in France feeds into my further deliberations, and so through to my actions; I am led to prepare for a holiday in France rather than elsewhere. Thus, as Harman and Bratman have both claimed already, decision-making in advance of action has an action co-ordinatory function. It helps ensure, or at least makes it more likely, that the actions which an agent performs at any given time, are those which he is justified in then performing-justified, in particular, given the actions which he performs at other times. Decision-making in advance of action allows the practical rationality of an agent continuant, a rationality which is concerned with the agent's interests and actions over time, to be made effective - to be applied in a consistent way in that agent's activities over time. Our capacity to deliberate about, and decide on, our actions in advance has further benefits. For example, given the limits to

Decision-making and freedom

71

our intelligence, there is a maximum complexity to the practical problems which we can solve at a given time. Deliberating and deciding on action in advance allows us - perhaps having first in a very general way deliberated about the structure of the entire problem, about what decisions are dependent on what - to split a very complicated practical problem up. We can deliberate and decide about matters of priority, and use our foreknowledge of the actions decided upon as input to subsequent deliberation and decision about the remaining parts of the problem. Again, the time for action may not be the optimum time for deliberation. It maybe that we shall be subject to distractions at the time of action which would make efficient deliberation then very difficult. For example, it is easy when deliberating under battle conditions for a tactician to omit to consider relevant tactical considerations, even if they are considerations of which he is aware — which is one reason why, where they can, tacticians deliberate and decide about tactics in advance of battles and not during them. Bratman, in particular, has argued that a capacity to take decisions and form intentions in advance of action is vital to any agent of bounded rationality and intellectual resource. Such agents need to subdivide practical problems and to deliberate and resolve them in stages over time. Such agents also need to avoid deliberation at times when the effectiveness of their deliberations would be most likely to be impaired. But we should not forget the underlying and fundamental function of the will, a function to fulfilling which these subsidiary applications of the will help contribute. Our practical deliberations are the deliberations of agent continuants - of agents who persist through time. As rational agent continuants, we are concerned with our actions both in the present and the future. The characteristic unit of practical reason is not the single action in the present, but the plan extending into and within the future. Hence an agent's practical deliberation should ideally influence that agent's action in the future exactly as it influences his action in the present. The actions which the agent performs in the future should be those he is justified in performing given those he performs in the present, and vice versa. To the identity of the agent over time there must correspond a rational unity in his agency over time. And it is this unity which the will facilitates.

72

The psychology of freedom DECISIONS AS DETERMINANTS OF ACTION

If the unity of our action over time is to be thus assured, then our present decisions should not merely increase the chance of our acting as decided. They should make that chance a certainty. Our decisions should ensure, or causally determine that we act as decided. The unit of practical reason is the plan extending into the future, not the action one performs now. The function of decisionmaking is to ensure that we do consistently apply reason in that plan-centred form - to ensure that our deliberations have the same influence on our future actions as they have on our present ones. And that requires a decision-making capacity which is actiondetermining - which serves to ensure that if we perform part of a plan now, we shall go on performing the rest of that plan in the future. The fact that, at the time we embark on a plan, only part will yet be being executed should not in itself reduce the chances of the plan - which we have decided upon in its entirety - being executed in its entirety. Of course, our decisions should not determine that we act as decided whatever happens. The function of decision-making is to apply the beliefs which motivated the decision - beliefs which based and formed a view of how it was desirable to act. And those beliefs may in time prove to have been false or at least not clearly true, or incomplete — or so at least it may eventually appear to the decision maker. As a decision maker, I may lose some of the beliefs which originally motivated my decision - or acquire new ones which, had I held them at the outset, would have motivated me not to take that decision. For example, I may have decided to leave for the station at 7 o'clock in order to catch the 7.30 train to London, only to discover before I can leave that the 7.30 has been cancelled, or that owing to a go slow on the railway, it would be quicker to drive. Or I might simply change my view of what is desirable. I might simply come to judge it more desirable to stay at home and rest. In other words I may come to revise the assumptions on which my decision was originally based. I may come to hold beliefs which, had I held them at the outset, would have moved me not to take the decision. My decision should not ensure that, regardless of my change in attitude, I still act as decided - that I still leave for the station at 7 o'clock. The causal influence of our decision-making is, therefore, a causal influence which can be counteracted. It can be

Decision-making and freedom

73

counteracted by decision-relevant belief change - by one's coming to hold beliefs which imply the incompleteness, or possible or actual falsehood of assumptions on which one's decision was based, and which, had one possessed them at the outset, would have moved one not to take that decision. The function of decisions to determine action in advance is, therefore, conditional on such beliefs not being acquired. But, provided they are not acquired, deciding to perform an action should indeed determine that, thereafter, we do act as decided. What if, before we act, we deliberate anew? Is not the actiondetermining function of decisions conditional on such deliberation not occurring? Once we deliberate anew, should not it be the outcome of that deliberation, and not any prior decision, which then determines how we act? As 'deliberation' is often used, one's continuing to deliberate about whether to do something does imply continued indecision about whether to do it. Someone cannot still be genuinely wondering what to do - still genuinely deliberating in this sense - if he has already made his mind up. And, of course, deliberation in this sense, implying indecision as it does, ipso facto implies that no previous decision is determining how the agent will act. It can only be the outcome of this deliberation, rather than any previous decision, which determines the agent's subsequent action. But then in so far as it is the function of a decision to determine action in advance at all, it must also be the function of a decision to prevent further indecision, and so also further indecision-involving deliberation, from occurring - unless, of course, there has been decision-relevant belief change first. The point of arriving at a decision, then, is to prevent ourselves from continuing to wonder what to do. But, even if it is the function of a decision to prevent further indecision, that does not mean that a decision need, or should, prevent all further deliberation about what has been decided. For not all such deliberation involves indecision. Instead of genuinely wondering what to do, we can simply be reviewing or defending a decision we have already taken. A decision need not in every case prevent us considering and debating further the grounds for taking it — as might occur in a discussion of whether and why the decision we have taken really was right. For such continued consideration of the grounds for a decision is perfectly consistent with our

74

The psychology of freedom

remaining decided - with that decision's continuing to determine how we will act. Suppose I am going through the grounds for my decision to do A with you, hoping thereby to convince you that my decision was at least no worse than any other I might have taken. That should not in itself preclude the decision's continuing to ensure that I do eventually do A. How I shall be acting should not suddenly become unsettled again. Just checking that my decision was the or a right one to take should not undo it. Nor, usually, will my decision be undone - unless of course I am led by my reflection to revise crucial assumptions on which that decision was originally based. It follows that the action-determining function of decisions is not conditional on no further deliberation occurring. It is conditional merely on there being no subsequent decision-relevant belief change. It is the function of decisions to settle action in advance. So decisions should prevent further deliberation of a kind that would leave how we shall act unsettled. Decisions should prevent us from continuing to wonder about what to do - unless , of course, decisionrelevant belief change has occurred first. On the other hand, it is perfectly true that decisions need not prevent us from reviewing them. But then the deliberation involved in merely reviewing a decision one has made is quite consistent with that decision's continuing to determine action - unless, again, decision-relevant belief change arises from the review. It is important that the point of deciding to act is to determine that one does act as decided. Deciding to act is not just about raising the chance somewhat of the action decided upon. For just raising the chance would not ensure the consistent application of a plan-centred practical reason. And it is the consistent application of reason through time which our decision-making is supposed to guarantee. One way of illustrating the point further is through an analogy between decision-making and betting — an analogy which is very important, and to which we shall be returning. By deciding on my actions in advance, I effectively increase the stake on a bet that I shall perform the act decided on. For my decision will lead me to perform further actions which, because they are co-ordinated with the action decided upon, increase what I gain from performing that action. Indeed, without these other actions, the action decided upon may not be worth performing at

Decision-making and freedom

75

all. On the other hand, such actions typically increase what I shall lose if I do not in fact act as decided. For example, by deciding to holiday in France well in advance of actually going, I am led to put deposits down on hotel rooms in France, and generally to devote time and other resources to preparations for holidaying in France. These actions increase what I gain from going to France on holiday. I have a better prepared and so more enjoyable holiday. But, at the same time, these same actions increase what I lose if I do not eventually act as decided. If I do not actually holiday in France, the actions will cost me the waste of my time and other resources — such as deposit money — all of which were spent on the assumption that I would act as decided. It is much better not to go to France having never decided to go, than not to go despite having earlier decided to go. My position as a decision maker, then, is like that of a bettor who, as he seeks to affect the outcome of a bet - as he seeks to cause the outcome by which he would win - also increases his stake. So, naturally, the function of my decision-making so understood must be to ensure that the outcome of the bet is in my favour. The function of taking a decision to act must be to determine outright that I win - that I do act as decided. Deciding to perform our actions in advance of actually performing them is supposed to increase what we gain from their performance. Action-determining decisions offer this increased gain — but without any risk of increased loss from the action's non-performance. Any benefit there is to raising the chance of the action decided upon by taking a prior decision to perform it, would be enhanced and not lessened by a decision which raised that chance to a certainty. The moral remains. The function of taking decisions in advance of action is to determine action in advance. Not only is it the function of decision-making to determine action in advance. Many of our decisions seem actually to fulfil their function. And that they should do so is, we ordinarily suppose, part of what it is to be fully in control of how we act. At the moment I have the option of finishing the paragraph which I am now writing, or of leaving my desk for a cup of coffee. It would definitely impair my capacity to control how I act if by taking a decision I could not determine which of these I shall do. That this is so is an extremely important fact about our freedom — a fact about freedom which any adequate theory of it must explain.

76

The psychology of freedom

And it is to the nature of freedom that we shall now turn. Given that decision-making has an action-determining function, what is it about freedom that might generate a dependence of action control on a prior decision control? INGOMPATIBILISM

We often take decisions to act well in advance of the actions decided upon. And the point of so doing is to settle definitely and in advance how we shall be acting - to ensure that we subsequently perform the right action. The function of a decision to do A, then, is to be a determining cause, given or along with the conditions under which it is taken, of the agent's subsequently doing A. Of course, as we have seen, this function is conditional: the agent should not acquire beliefs which, if he had held them at the outset, would have moved him not to take that decision. But, assuming that this condition is met - and I shall for the moment be considering only cases where this condition is met - the function of decisions is actiondetermining. That decisions have this action-determining function suggests that there may be a connexion between our problem about the psychology of freedom - the issue of whether Dependence is true - and a well-known and traditional dispute about freedom's metaphysics. This is the dispute about the compatibility of freedom of action with causal determinism - the dispute between Incompatibilists and Compatibilists. Compatibilists assert that our freedom to act otherwise than as we actually do is compatible with the causal determination of our actions by prior conditions which lie outside our control. Incompatibilists deny this compatibility. Suppose that we take a properly functioning decision to do A. With the conditions under which it is taken, that decision must then determine that we subsequently do A. Suppose, further, that we lack control of that decision - that the decision is a passive passion which simply happens to us. Then - at any rate in the absence of our having control of any other of the action's determining conditions it follows that our performance of the action is determined by a set of prior conditions which are outside our control. Incompatibilism says that agency-determining conditions which are outside our control are control-depriving causes of that agency: that if how we act is causally determined by conditions which are

Decision-making and freedom

77

outside our control, then how we act is outside our control also. Hence, Incompatibilism predicts that where a decision to act is functioning properly - where the decision determines that we shall act as decided - and we lack control over whether we take the decision, so too we must lack control of the action thereby determined. We cannot be free not to perform that action. The connexion between Incompatibilism and our belief in Dependence seems obvious. Incompatibilism can explain why there might indeed be a dependence of freedom of action on freedom of will. Where our decisions are properly functioning - where they do determine that we shall act as we have decided - then, Incompatibilism predicts, our action control does depend on our also having decision control. So, to the extent that our decisions often appear to us to be functioning properly - to the extent that they render the action decided upon at least highly probable and may, for all we know, actually determine its occurrence - then, given belief in Incompatibilism, our action control will appear to us, as it certainly does appear to us, to depend on our prior decision control. Incompatibilism thus provides some explanation for our intuition that Dependence is true - an explanation which, an Incompatibilist may claim, is non-regressive. For Incompatibilism, as I have understood it, predicts the dependence of our control over a given activity on our control of prior events only for those cases where the activity is causally determined by those further events. So, provided our decisions are not causally determined by prior events in turn - such as by our prior perceptions, beliefs, and desires - then our control of what happens can perfectly well start with our decisions. But, the Incompatibilist alleges, our control of what happens cannot so plausibly start just with our action. The point of deciding to act in advance of the action decided upon is, after all, precisely to determine in advance how, at least in the absence of belief revision, we shall eventually be acting. Compatibilists, of course, will take a different view. The relation between actions and decisions which the Incompatibilist explanation exploits — the relation of causal determination — does not provide, for Compatibilists, any justification at all for a dependence of action control on decision control. Moreover, Compatibilists have often claimed that this relation of causal determination holds, at least in rational agents, between decisions and their causes also. Where conditions outside our control render one action more

78

The psychology of freedom

justified than another, those conditions should determine, and perhaps very often do determine - through our recognition of them in our beliefs - that we decide to perform the more justified action. So Compatibilists will want to deny that the explanation really is non-regressive - that it really does appeal to a relation which is peculiar to actions and prior decisions to act. Incompatibilists appear then to have some explanation of why Dependence should generally at least appear to hold - indeed of why in many cases there may well actually be a dependence of our action control on a prior decision control. The basis for their explanation is a relation - that of causal determination - which, given that it is the function of decision-making to settle how we shall act, should commonly hold between actions and the prior decisions to perform them; but which, though Compatibilists might dispute the point, does not quite so plausibly hold between actions and other prior events. And, of course, given the availability of this Incompatibilist explanation of why action control should depend on decision control, the intuitiveness of that Dependence seems in turn to support Incompatibilism. An explanandum, provided it has some independent credibility, gives support for whatever seems to provide its best or only explanans. And so it seems to be Incompatibilism which does justice to the importance that we naturally accord to our freedom of will. It seems to be Incompatibilism which explains why having control over which actions we decide to perform should matter to our having control of our actions - to our being free at all. Now not all Incompatibilists have taught a dependence of freedom of action on a prior freedom of will — and we shall shortly see one possible reason why. But one who does, Robert Kane, has appealed precisely to this idea that freedom of will matters just because many, perhaps most, of our first-order actions are determined in advance through the will: [Incompatibilism] does not require that all acts for which we are ultimately responsible (and hence which are done 'of our own free will') must be undetermined. Often we act freely and responsibly (even in an incompatibilist sense) out of a will already formed. What [Incompatibilism] requires is that we ourselves formed that will to some degree in the past by acts that were undetermined. This is what allows us to say we are acting out of our own free will-a will of our own making-even when our characters and motives determine our actions. (It is also why, in the

Decision-making and freedom

79

final analysis, the free will issue is about the will.) (Tree will: the elusive ideal', p. 55) THE PSYGHOLOGISING CONCEPTION OF FREEDOM

Is the problem of the psychology of our freedom then just a part of the traditional problem of its metaphysics - the problem of whether freedom of action is consistent with causal determinism? Is Dependence a specifically Incompatibilist claim? So the allegiances of some past philosophers might suggest. For example, just as Hobbes combined a denial of Dependence with the affirmation of Compatibilism, so too his opponent Bramhall combined belief in Dependence with Incompatibilism. In BramhalFs view, freedom of action required both that our actions be undetermined by prior conditions outside our control, and that we possess a free will - a control over which actions we decide to perform. But I want to detach Dependence from Incompatibilism - and from Compatibilism too for that matter. The debate about the true psychology of our freedom is, I want to argue, largely independent of disputes between Incompatibilists and Compatibilists. For Dependence, in the exact sense in which we believe it, is anyway not fully explained by any purely Incompatibilist condition on freedom. Dependence, in the sense we believe it, needs to be established on other grounds. It needs to be established by appeal to the idea that freedom involves a capacity for rational self-determination. And this is an idea of freedom which is not specifically Incompatibilist though, at least in the form I use it, I hope that it can be made acceptable to Incompatibilists and Compatibilists alike. It is easy enough to show that Incompatibilism does not do justice to our intuition that Dependence is true. Incompatibilism, as I am understanding it, is the doctrine that agency-determining conditions which are outside our control are control depriving causes of that agency: that if how we act is causally determined by conditions which are outside our control, then how we act is outside our control also. So Incompatibilism does predict a dependence of action control on a prior decision control — but only in those cases where an action actually is causally determined by a prior decision to perform it. Yet our intuition that Dependence is true seems far more general. We have control over which actions we perform, we

80

The psychology of freedom

naturally think, only because we have a general capacity to exercise control over how we act through prior decision-making. It is up to us which actions we perform only because we can decide which actions we shall perform, and it is up to us which such decisions we take. But if that is our intuition - and is not this a claim that we quite naturally assent to? - then Incompatibilism cannot explain it. For the intuition, as stated, applies to actions generally, whether or not their performance is actually preceded, let alone causally determined, by a prior decision. The intuition implies, for example, that agents, such as the lower animals, who altogether lack a decisionmaking capacity must also, and simply for that reason, lack freedom of action. But Incompatibilism does not predict that at all. Since none of these agents' actions would be causally determined by prior decisions to perform them, Incompatibilism has nothing to say about why there should be a dependence of action control on a prior decision control in their case. Our intuition that Dependence is true is based, not on any specifically Incompatibilist condition on freedom of action, but rather on a quite different, Psychologising conception of freedom. By a Psychologising conception of freedom I mean the claim - which we seem to believe - that our control over our actions depends on our having a capacity to control how we act through free decisionmaking. The rationale for the term is obvious. Freedom of action is made out to depend on the possession of a psychological analogue of that freedom - on the capacity for a free second-order agency of the mind. Correspondingly, I shall term Depsychologising any theory of freedom of action which denies the Psychologising claim. Incompatibilism makes a condition of our freedom of action the absence of uncontrolled determining causes of action. Whereas a Psychologising conception of our freedom of action is quite different. It makes a condition of that freedom, not the absence of uncontrolled determining causes of action, but the presence of a capacity for a further kind of action-generating action. A Psychologising conception of our action control ties our action control to a capacity to exercise it in advance, through freely taken, actiongenerating second-order actions. The basis of this Psychologising conception of our freedom, and so of our belief in Dependence, lies then not in Incompatibilism, but rather in some theory of an altogether different kind.

Decision-making and freedom

81

SURVIVAL

Our conception of our freedom is Psychologising. We think that free agents must have the capacity to control how they act through taking free decisions. But that is not all that we believe. We also think that the use of free decision-making to control our future action is always consistent with our continued freedom of action. As well as Dependence, we also believe Survival. This is the claim that our freedom of action must survive the exercise of our freedom of will - that where we have had control of whether or not we took the decision, deciding to do A cannot of itself remove our continuing freedom to change our minds, and not do A. So, for example, just deciding not to catch the 7.30 to London cannot, on its own, deprive me of the freedom to change my mind and catch it. What deprives me of that freedom is the action, or rather inaction, which eventually results from that decision. I lose that freedom by sitting at home until it is too late to catch the 7.30 train. We retain an undiminished freedom to act otherwise than as we have decided until, commonly through our own subsequent actions or inactions, the alternative options are removed or reduced. It is important that we assent to both of Survival and Dependence. Where a decision to do A causes us to do A and forewarns us of how we shall act, our intuition is both that our having control of whether we do A depends on our also being in control of the decision that will get us to do it - the decision cannot be some passive passion which simply happens to us - but also that our taking that decision leaves us perfectly free thereafter to change our mind and not do A. We treat Dependence and Survival as perfectly consistent. Our conviction that we control what actions we perform only because we control what actions we first decide to perform may be strong; but that conviction never weakens our further belief in our continuing freedom to act otherwise than as initially decided. Not only are the two convictions consistent. The Psychologising conception of our freedom which leads us to believe Dependence, makes it hard to avoid believing Survival too. According to the Psychologising conception, it is our capacity for free decisionmaking which makes us free agents. But how can exercising the very capacity which gives us our freedom of action, take that freedom away? Survival must be true as well.

82

The psychology of freedom

That freedom of action depends on freedom of will is something which we commonly suppose, and has often been claimed by philosophers. That freedom of action is sure to survive the exercise of freedom of will is not so commonly noted - but it is an equally prevalent intuition, and equally in need of explanation. Any explanation of why Dependence is true needs to be consistent with Survival. Indeed, if as I suspect Dependence cannot be true without Survival being true as well, any decent explanation of Dependence should explain Survival too. We have seen that Incompatibilism on its own provides, at best, a rather partial explanation for why Dependence should be true. So how does Incompatibilism stand to our belief in Survival? And the answer to that question depends on what sort of Incompatibilism we have in mind.

STRONG INCOMPATIBILISM

Incompatibilism is standardly expressed as the doctrine that if at any time our actions are causally determined by occurrences outside our then control, our actions are outside our then control also. Thus stated, in what I shall term its Strong form, Incompatibilism has a distinctive metaphysics of control - a metaphysics that many philosophers, at any rate, clearly find very appealing. This metaphysics of control says that, given that the past is outside one's control, one's continued freedom of action is incompatible with the causal determination of one's action by past events - by history to date. Having control at t over whether one does A - having freedom both to do A and not to do A - requires that history to t leave it still causally undetermined whether one does A. Action control requires that it remain causally open how one acts. As Kant put it, in a famous statement of Incompatibilism: what we wish to understand, and never shall understand, is how predeterminism, according to which voluntary actions, as events, have their determining grounds in antecedent time (which, with what happened in it, is no longer within our power), can be consistent with freedom, according to which the act as well as its opposite must be within the power of the subject at the moment of its taking place. (Religion within the Limits of Reason Alone, p. 45)

Decision-making and freedom

83

Incompatibilism, we saw, does imply some dependence of our action control on a prior decision control. It implies the dependence in just those cases where the action is preceded by a prior decision which, along with the conditions under which it has been taken, determines the action's performance. But those are just the cases where Incompatibilism, in its Strong form, also implies the falsehood of Survival. Once an action-determining decision has been taken - once that decision is past - it becomes an action-determining occurrence which has ceased to be within our control. So, once the decision has been taken, how we shall act is causally determined by a set of conditions which are by now outside our control. But then, Strong Incompatibilism predicts, the action determined must by now be outside our control also. The same Incompatibilism which does something to conform to our intuitions by implying that, in certain cases at least - cases where our actions actually are causally determined by prior decisions - action control does indeed depend on decision control, then implies that in those same cases, our action control must also end with the decision taken. So whatever does explain our belief in Dependence, it certainly could not be belief in Strong Incompatibilism. For then our belief in Dependence would weaken our belief in Survival -which it does not. Once we have taken an action-determining decision, Strong Incompatibilism implies, the action determined thereafter ceases to be within our control. We are no longer free to change our minds and act otherwise. This is something of a pill to swallow. The Strong Incompatibilist may sweeten it somewhat. Like Kane, 2 he might remind us that, in such cases, we can still be said to act 'freely' or 'of our own free will' when we eventually perform the action. The action is Tree' in the sense that it is the intended consequence of prior agency - the agency of our decision-making - which, at the time we performed it, we were free not to perform. The action is therefore, as Kane puts it, 'free' in the sense that it is something for which we are 'ultimately responsible'. But, of course, where Survival is concerned, it is not our ultimate responsibility for the actions on which we have decided that is at issue, but our continued freedom to change our minds and not perform them. As ever, the freedom with which this book is concerned is always a freedom both 2

See his 'Free will: the elusive ideal', p. 55.

84

The psychology of freedom

to do A and to refrain - a freedom of alternative possibilities. And our intuition is very definitely that deciding to do A will not of itself remove this freedom. Given Strong Incompatibilism, then only if our decisions leave some actual chance that we will not act as we have decided, do we retain a continuing freedom not so to act. Combine this Incompatibilism with the idea that Survival is true - that the exercise of free will can never deprive us of continued freedom of action - and we are left with the conclusion that, at least when freely taken, our decisions can never be action-determining. Some philosophers, assuming both Strong Incompatibilism and Survival, have arrived at just this conclusion. Rather than, like Robert Kane, attempting to explain Dependence by emphasising the action-determining nature of decision-making, these Incompatibilists have tried to protect Survival by denying that decisions ever fulfil such an action-determining role. (Which may be precisely why some Incompatibilists might not wish to follow Kane in teaching or emphasising Dependence. The only specifically Incompatibilist rationale for Dependence assumes that decisions are action-determining. But, by standard Incompatibilist lights, that assumption threatens our no less firmly held belief in Survival.) Thus Henry Sidgwick: It is sometimes vaguely thought that a belief in Free Will requires us to maintain that at any moment we can alter our [future actions] by a sufficiently strong exertion. And no doubt most commonly when we make such efforts, we believe at the moment that they will be completely effectual: we will to do something hours or days hence with the same confidence with which we will to do something immediately. But on reflection, no one, I think, will maintain that in such cases the future act appears to be in his power in the same sense as a choice of alternatives that takes effect immediately. Not only does continual experience show us that such resolutions as to the future have a limited and too frequently an inadequate effect: but the common belief is really inconsistent with the very doctrine of Free Will that is thought to justify it: for if by a present volition I can fully determine an action that is to take place some hours hence, when the time comes to do that act I shall find myself no longer free. We must therefore accept the conclusion that each such resolve has only a limited effect: and that we cannot know when making it how far this effect will exhibit itself in the performance of the act resolved upon. (Methods ofEthics, pp. 74-5)

Decision-making and freedom

85

Sidgwick's denial that decisions can ever be action-determining, is hardly very compelling. It is indeed perfectly true, as he observed, that not all decisions do determine that we act as decided. And there are perfectly good explanations why. The agent's incapacity to act as decided aside, the decisions may concern actions too far into the future; or there may be a significant chance of the agent's revising the beliefs on which the decision was based; or the decision maker may simply be weak-willed in this case. But such explanations need not always apply. And where they do not, decisions may perfectly well be action-determining in fact as well as in function. Consider decisions to execute basic short-term plans — plans for obtaining food, for example - which extend no more than seconds, minutes, or at most hours ahead. Our survival depends on our ability consistently to adhere to such plans. It would be strange had we developed no psychological capacity which ensures that we do so. Moreover, it is natural to suppose that someone fully in control of how they act should manage to take action-determining decisions. Remember the options I mentioned earlier, and which are available to me again. I have the option of finishing the paragraph which I am now writing, or of leaving my desk for a cup of coffee. As we noted before, my action control is impaired if by taking a decision I cannot now determine which of these I shall do. And, despite his condescending tone, Sidgwick showed himself to be aware of this element to our conception of our freedom when he reported the currency of the idea that: a belief in Free Will requires us to maintain that at any moment we can alter our [future actions] by a sufficiently strong exertion. It is easy to see why we associate action control in its unimpaired, complete form with a capacity to take action-determining decisions. That is simply a reflection of the Psychologising conception which we have of our freedom. An agent in complete, unimpaired control of his actions is just an agent who is able to make effective exercise of the capacities which give us our action control. Now, the Psychologising conception says that our action control comes from our possession of a decision-based capacity for future action control — from a free will. What is it for this capacity for future action control to be exercised effectively? We exercise the capacity effectively by taking decisions which determine which future actions

86

The psychology of freedom

we perform. So, as natural Psychologisers, our paradigm of a free agent is an agent who manages to determine how he will act by his own free decision-making. Our paradigm of freedom involves, then, just the kinds of decision which, on a Strong Incompatibilist conception of freedom, should remove any freedom thereafter not to act as we have decided. But even so, we never ordinarily suspect or suggest that our own decision-making poses any such threat to our subsequent freedom. Our belief in Survival, then, is inconsistent with Incompatibilism in its standard Strong form given one highly plausible premiss — that freely taken decisions can be action-determining. And it does seem to be Survival, rather than standard Incompatibilism, which we ordinarily believe. No matter how great our confidence that we shall not actually abandon a decision we have made - no matter how far we are from assuming any chance that we will not act as we have decided - that confidence alone never diminishes our belief that we remain perfectly free to abandon our decision. We seem, therefore, to treat our own decisions as guaranteeing us a freedom to change our minds and abandon them. But if decisions can be actiondetermining, Strong Incompatibilism allows us no such guarantee. But at this point an Incompatibilist might very well object. Incompatibilism, he maintains, need not after all imply the price of Survival to be decision-making that fails to determine subsequent action. For there may be more than one concept of a freedom to act otherwise which we deploy - more than one sense in which we can be said to be 'free to act otherwise'. Survival certainly does seem to be something which we ordinarily believe; and we are happy to believe it despite the fact that many of our decisions are or may well be action-determining. But that just shows that in believing or asserting Survival we are simply understanding 'freedom to act otherwise' in a special and weak sense — a sense which can continue to apply despite the determination of our actions by our past decisions. The Incompatibilist maintains that he is concerned to analyse a far stronger sense of'freedom to act otherwise' - a sense which simply is not employed in stating Survival. But if the Incompatibilist takes this tack, it is fair to ask him what sense of'freedom to act otherwise' it might be that he is concerned with. For the sense of'free to act otherwise' which we employ when we state Survival certainly seems no different from that which we

Decision-making and freedom

87

employ when we state Dependence - and it seems to be a sense in which 'freedom to act otherwise' is standardly understood. We ordinarily believe, I have claimed, that we are free to act otherwise than as we actually do; that our freedom to act otherwise depends on our having a corresponding freedom to decide otherwise; and that we do not lose or reduce this freedom to act otherwise simply through taking decisions about which action we shall perform. Not only are these beliefs about our freedom which we ordinarily hold. They seem also to be beliefs about precisely the same kind of freedom throughout. And the same point could be made using 'control5 as an alternative to 'freedom'. We possess control over which actions we perform — a control which depends on our also having a corresponding control over which actions we decide to perform, but which is also sure to survive the mere taking of a given decision to act. All these claims seem to be true of one and the same kind of control over what we do. If our belief in Dependence and Survival is about the same freedom throughout, the Incompatibilist must be correspondingly consistent in his treatment of it. Either our ordinary belief is about the very same kind of freedom which Incompatibilism seeks to explain. In which case Incompatibilism can justify our belief in Dependence, but, as we have seen, only under certain conditions - conditions where our actions actually are determined by prior decisions to perform them; and then, in its Strong form, only at the price of claiming that our belief in Survival is mistaken. Or our ordinary belief is about some quite different kind of freedom throughout - a kind of freedom with which Incompatibilism is not concerned at all. In which case, in so far as the Incompatibilist cannot contradict our belief in Survival, so too he cannot justify our belief in Dependence. And that would make Incompatibilism a doctrine of very little interest to us indeed. For it would then have absolutely nothing to do with a kind of freedom of action in which we very clearly believe, and which I am seeking to explain - a freedom of action of which both Dependence and Survival are true. In fact I doubt that, in its standard Strong form, Incompatibilism provides a true account of any variety of freedom of action which we ordinarily believe ourselves to possess. If it did, then there would have to be an ordinary sense of'freedom to act otherwise' in which it was indeed true that, just by our own free decision-making, we could deprive ourselves of the freedom to change our minds and act

88

The psychology of freedom

otherwise than as we had decided. But I reckon there simply is no such ordinary sense of'freedom'. If the Strong Incompatibilist ever speaks truly, it is with some invented term of art. WEAK INGOMPATIBILISM

Belief in Dependence is part, then, of a more general Psychologising conception of our freedom - a conception which seems on examination not only to owe little to Incompatibilism, but, if anything, to be at odds with it. Not only does Incompatibilism not explain why our freedom of action should require what, as believers in Dependence, we ordinarily suppose it to require - a capacity for action-determining, second-order action. In its standard form, Incompatibilism is committed to denying our belief that the exercise of this capacity for second-order action is essentially freedom-preserving. Does that mean that this Psychologising conception of our freedom, of which Dependence and Survival both form a part, is an essentially Compatibilist conception? Philosophers nowadays would generally suppose so. For modern philosophers generally, the dispute between Incompatibilists and Compatibilists is simply a dispute about what I have called Strong Incompatibilism — the Incompatibilism which claims that our present freedom of action is inconsistent with our action's causal determination by past events or history to date. Thus David Lewis has defined Compatibilism as the doctrine that it may be true that: one freely does what one is predetermined to do . . . that in such a case one is able to act otherwise though past history and the laws of nature determine that one will not act otherwise. ('Are we free to break the laws?' P- 29O For Lewis the dispute between Compatibilism and Incompatibilism is nothing more than a dispute between Compatibilism and Strong Incompatibilism. But perhaps this very widespread assumption is false. Many of us, when we first consider the metaphysics of our freedom to act otherwise, begin as natural Incompatibilists. We are naturally swayed by various examples of action-determining conditions which, supposedly, take away the freedom not to act as they determine. Consider someone who has been determined to decide on and perform an action by the manipulations of some

Decision-making and freedom

89

scientist. Or consider someone who is determined to perform a given action under given circumstances by an element of his genetic make-up, fixed before his birth. We find it very natural to doubt whether the people in such examples do have the freedom not to act as they have been determined to. But we are not normally asked to consider people who, supposedly, lack the freedom to act otherwise simply because through their own decision-making they have already managed to determine exactly how they will act. For such examples would not be nearly so convincing. As we have seen, it is not an obvious threat to our continuing freedom of action, that we may have managed by our own decision-making to determine how we'll act. It is certainly not like supposing that some manipulative scientist, or some gene, has determined how we will act. Our natural Incompatibilism, which the stock examples evoke and exploit, can easily be accompanied by a pretty robust belief in Survival. Does combining Incompatibilism with Survival mean that we are confused? - or that, when thinking about our freedom to act otherwise in relation to our own past decisions, we suddenly start using a different, 'Compatibilist' concept of freedom? Neither need be the case. It may simply be that our Incompatibilism is Weak, rather than Strong. Weak Incompatibilism is recognisably a form of Incompatibilism. It agrees with its Strong competitor, that action-determining conditions which are at all times outside our control remove our freedom not to act as they determine. So, for example, it agrees that if how I now act is causally determined by events before or at my birth, then how I now act is outside my control. But Weak Incompatibilism disagrees with its Strong competitor about actiondetermining conditions which, though now outside our control, were once within our control. Some at least of these actiondetermining conditions - such as our own past decisions - do not deprive us of our freedom of action. And the reason for this is that Weak Incompatibilism is autonomy-centred rather than timecentred.

For the Weak Incompatibilist, the mere pastness of an action determinant does not remove our freedom not to act as it determines us to. What matters is whether, besides being past, the operation of the action determinant is a threat to our autonomy - to our status as a properly self-determining agent. Being

go

The psychology of freedom

determined to act by the manipulations of others, or by events before our birth - by occurrences that have never been within our control and are not of our own free doing - does threaten our capacity for self-determination. Or so, not implausibly, the Weak Incompatibilist supposes. But there are other past action-determining occurrences - our own decisions - which are of our own free doing, and to which autonomy-centred Weak Incompatibilism takes a very different attitude. And this discrimination is based precisely on the importance of these decisions to our autonomy- to our capacity for self-determination; an importance, moreover, which we can all recognise, whether or not we are Incompatibilists as well. For in so far as we all naturally conceive our freedom in Psychologising terms, we all recognise that our ability to control how we act through our own free decision-making is necessary to our being free - truly self-determining - at all. The Strong Incompatibilist does not discriminate between the various kinds of action-determining past history — between the manipulations of other agents and events before our birth on the one hand, and our own past decisions on the other. But to the extent that we have Incompatibilist intuitions, many of us naturally do. We are adherents of a Psychologising conception of our freedom who, in so far as we are Incompatibilists as well, are naturally Weak, autonomy-centred Incompatibilists. Strong Incompatibilism is based on the belief that continued action control requires continued causal openness. This is its distinctive metaphysics of control. But it is not a metaphysics which the Weak Incompatibilist shares. Weak Incompatibilism is based instead on a conception of what it is to be a genuinely selfdetermining agent - a demanding conception of self-determination which rules out the determination of one's actions by occurrences that have been at all times outside one's control, and so are not of one's own doing; but which, like any plausible conception of what self-determining freedom comes to, very much rules in the determination of one's actions by one's own decisions. Decisions are actions by which we freely determine our subsequent actions. And their action-determining influence does not remove our capacity for self-determination, but helps constitute it. Dependence and Survival are both true. But of course, there are other free actions by which we could also determine our subsequent

Decision-making and freedom

91

actions - though in a way that would remove our freedom and not preserve it. Taking addictive drugs, for example - that is an action which, initially at least, we can perform freely, but whose effects notoriously tend to remove our freedom rather than preserve it. An autonomy-centred Weak Incompatibilism must explain what makes decision-making so special - why the action-determining influence of decisions on subsequent action is freedom-preserving, when the action-determining influence of other, equally freely performed actions, such as drug-taking, would not be. But then that is a job for all of us anyway. For, Incompatibilists or not, as natural Psychologisers we all share the intuition that decision-making is special. We all believe that it is our capacity for free decision-making in particular that makes us truly selfdetermining and so free - that therefore the influence of our own decisions must preserve our freedom, as the influence of other actions, such as drug-taking, need not. So we shall all need some account of what makes decision-making special — of why both Dependence and Survival are true. Once we have developed the required theory of decision-making — as we shall shortly begin to do — not only will we have defended a Psychologising conception of our freedom. We shall also, and by that very fact, have provided a theory which the Weak Incompatibilist can use too: a theory of how, unlike other determinants of action, action-determining decisions constitute our capacity for self-determination, and so cannot remove it. There certainly are arguments for Incompatibilism in its Strong form. There is, in particular, the Consequence argument — an argument or family of arguments which, in its various forms, seeks to move from our lack of control over history to date to our lack of control over any action which is determined by - a consequence of- history to date. It would be a major distraction to go into all possible forms of the Consequence argument here - let alone into every variety of objection which the Consequence argument in its various forms has met. But as Kane has observed, (Tree will: the elusive ideal', p. 32) such arguments have convinced no one on formal grounds alone. Whether the arguments are sound depends crucially on how one should understand terms such as Tree to act otherwise'. And yet it is precisely how such terms should be understood that is at issue. So far, no version of the Consequence argument for Strong Incompatibilism has managed to compel the

92

The psychology of freedom

belief of the generality of the logically competent but not already converted.3 In the absence of such a decisive argument in its favour, why espouse Incompatibilism in its Strong form? Such an Incompatibilism is so clearly inconsistent with two claims about the will which, we have seen, are firmly entrenched in our conception of our freedom. These are the claims that the point of taking a decision about what to do is to settle definitely what thereafter we shall be doing - that the function of decision-making in advance of action, is to determine which actions in the future we shall be performing; and the claim that Survival is true - that our own decision-making is sure to leave us free to change our minds and act otherwise than as decided. It might still be argued that once we give up Strong Incompatibilism, there is no rationale left for Incompatibilism in any form. But this is just to assume that the only available Incompatibilist metaphysics of control is the Strong Incompatibilist's time-centred variety, which ties continued control to continued causal openness. But this is false. There is also the quite distinct autonomy-centred metaphysics of Weak Incompatibilism, which, by tying freedom to a capacity for true self-determination, rules out the determination of free actions by occurrences that are not of our own free doing - but not the determination of free actions by our own free decisions. This metaphysics is at least as attractive. Indeed, given the great plausibility of Survival, it has a claim to provide the most attractive Incompatibilism around. Weak Incompatibilism takes threats to our freedom to act otherwise to be, not past determinants of action in general, but only those which - like the manipulations of other agents but unlike our own decisions - specifically threaten our capacity for selfdetermination. Until there is some decisive argument for why this autonomy-centred discrimination is incoherent - and in the absence of an uncontroversially sound argument for Strong Incompatibilism, I know of none - why should not Incompatibilists discriminate? If they do, they will end up with a version of 3

For versions of the Consequence argument, see for example D. Wiggins, 'Towards a reasonable libertarianism', Van Inwagen, An Essay on Free Will, C. Ginet On Action. Objections can be found in Slote, 'Selective necessity and free will', and Lewis, 'Are we free to break the laws?'

Decision-making and freedom

93

Incompatibilism that is perfectly consistent with Survival, and so with the Psychologising conception of our freedom which we so clearly also deploy. And that must be a point in favour of this version of Incompatibilism. DECISION-MAKING AND RATIONALITY

Decision-making seems to be a very special kind of actiongenerating action. An agent who determines how he acts by his own free decision-making is our paradigm of a free, truly selfdetermining agent. Indeed, our capacity deliberately and freely to decide to perform this action rather than that, is the capacity which makes us free at all. It is a capacity by exercising which we always preserve our freedom and never directly remove it. What makes decision-making so special? What makes taking a decision such a very special kind of actiongenerating action, I shall now suggest, is the connexion between decision-making and our rationality. Taking a decision constitutes a distinctively rational form of self-determination. The first point to be made about decisions, at least as we ordinarily conceive them, is that rational decisions to act dispose us to perform rational actions: that if a decision to act is taken rationally, then, in the absence of new information which warrants revising the assumptions on which the decision was based, the action which executes the decision will be rational too. The rationality of deciding to do A guarantees the rationality of doing A thereafter. We shall be returning to explain this characteristic of our decisions - the rationality-preserving nature of their influence on action - in later chapters. But it is a characteristic which we constantly assume decisions to have. If we grant the sense of deciding on a particular course of action, we never deny that performing the course of action decided upon is sensible too. Not only is a past decision a rationality-preserving cause of the action decided upon. That past decision is something we have ourselves done. A decision, in fact, is an exercise in self-governance - a form of self control. Not only that, but the decision is a form of self-governance in the service of our own continuing rationality. The point of settling how we shall act by taking decisions, after all, is to ensure that we act rationally thereafter - that we perform the right actions. And decision-making ensures that we act rationally

94

The psychology of freedom

thereafter precisely because decision-making is rationalitypreserving in its effects - because rational decisions to act give rise to actions that are no less rational in turn. When we take a decision, and do so rationally, our decision ensures that we will be acting no less rationally in the future. Now when decisions have this rationality-preserving effect, this is precisely not a case of our rationality being imposed on us from without. We have here a continuing rationality being imposed on us from within, through our own free agency - through what we ourselves have freely done. And it is a familiar and deeply appealing thought - a thought which we find, for example, at the heart of Kant's theory of freedom — that the free self-imposition of a law of reason must be consistent with our freedom as subjects on whom that law has been imposed.4 It is very natural, then, to explain the peculiar plausibility of Survival - a plausibility which it has even for those of us naturally inclined to Incompatibilism - in terms of the fact that, as we ordinarily conceive them, decisions are a very special kind of reason-applying action. In rationally taking a free decision, we ipso facto freely impose reason on our subsequent action. The exercise of such a capacity for free and rational self-determination can never be inconsistent with our subsequent freedom. And if that explanation for Survival is right, we also begin to see what it might be that makes Dependence true, and so what might underly the Psychologising conception of our freedom. Free agents need to be able to control their actions through free decisionmaking, it may well be, because freedom requires some capacity for rational self-determination - a capacity which only the free reason-imposing agency of the will provides. The next chapter is in fact going to provide just this explanation of both Dependence and Survival. It will argue that decisionmaking is indeed a special reason-applying activity — an activity with a special executive function. Our decisions to act are second4

For the Kantian, the active self-imposition of a law of reason through the will is a legislative act - an act which (somehow) makes the law. As Kant put it: 'A rational being must always regard himself as making laws in a kingdom of ends which is possible through freedom of the will' (Groundwork, ed. H. Paton, p. 101). But the intuition that such self-imposition is freedom-preserving still appeals, I think, if we conceive of the will's act of self-imposition as merely executive - as the agent's own imposition on himself of a law that comes from outside his will. In the next chapter I argue that we should conceive of the will as the locus of an executive rather than a legislative agency.

Decision-making and freedom

95

order actions by which we impose reason as it governs our first-order agency. The exercise of such a capacity for rational selfdetermination can never take away our freedom; and possession of the capacity is necessary to our being free at all. This Psychologising theory of our freedom will not be based on any Incompatibilist doctrine. But neither is it obviously inconsistent with Incompatibilism in all its forms. There is one form of Incompatibilism, in particular, with which the theory will be wholly consistent. This is the Weak Incompatibilism which claims that freedom of action requires, not continued causal openness, but genuine autonomy. Rational self-determination through one's own free decision-making may prevent its remaining causally open how one will act. But it can surely never preclude one remaining a truly autonomous, self-determining being. Nothing in the remainder of this book, then, will presume that this autonomy-centred Incompatibilism is false. Indeed, a combination of a Psychologising conception of freedom with an autonomy-centred Incompatibilism is probably the view of our freedom to which many of us initially incline. What else is a free agent but an agent who, while remaining perfectly free to change his mind thereafter, can determine exactly how he will act by his own free decisions — decisions which nothing outside his own doing determines him to take? But it is not the purpose of this book to prove any particular form of Incompatibilism. The purpose of this book is to prove a Psychologising conception of our freedom, and in a way that I believe it can be proved - in terms that are independent of Incompatibilism, though not I hope in conflict with it. It is one issue whether freedom of action requires, as Incompatibilists all suppose, the absence of determining causes of action that are not of our free doing. It is another and quite distinct issue whether, as Psychologisers all suppose, freedom of action requires the presence of a second-order, action-generating agency of the will. And it is this second issue which we shall now be pursuing. From this point on we leave the battle between Compatibilists and Incompatibilists behind us. DECISIONS AND DECISION DRUGS

Notice that even if Survival is true, we do not yet have a full explanation why. We have not yet explained what makes decision-

96

The psychology of freedom

making quite so special. There must be even more to decisionmaking than we have so far discovered. For besides simply taking a decision, there are other free activities by which, at least in principle, I could exercise self-control - by which I might determine how in the future I shall act. These activities might resemble decisions in their effects; but, unlike simple decision-making, even when performed freely they can much more plausibly deprive us of our subsequent freedom of action. For these decision-like activities, there is no equally plausible analogue of Survival. For example, I might take a drug that was designed to cause me to do A by ensuring that, having taken it, I would thereafter desire to do A. Call such a drug a decision drug for doing A. Taking that drug would produce an effect on my subsequent desires. The drug would ensure that, though still disposed to refrain from doing A if I so decided, owing to a strong desire to do A I would not so decide. Now it is far from obvious that taking such a drug would leave me with a continuing freedom not to do A. Normally - in the absence of an antidote - one's control over whether one suffers the effects of a drug ceases once the drug has been taken. Unless decision drugs are to be an exception to this rule, taking them, in the absence of an antidote, must bring one's control over whether or not one does A to an end. Let me note one further similarity between taking a decision and taking a decision drug. The effect of a decision to do A, I have claimed, is to determine that we subsequently do A. Now this effect is subject to there being no decision-relevant belief revision. It must not come to appear to us that assumptions on which our decision depended were crucially false or incomplete. So let me stipulate that the causal influence on action of decision drugs be similarly restricted. Let the antidote to a decision drug - the event which counteracts the causal influence of the drug - consist in a comparable revision of the beliefs which originally moved the drug taker to take the drug. Now, unless an antidote is available - unless we actually take an antidote, or at least have control over whether we take it - our control over the effects of a drug ceases once the drug has been taken. The antidote in the case of a decision drug is, as with decisions, a certain sort of belief revision. Assume that such belief change does not occur, and that whether it does is anyway outside

Decision-making and freedom

97

our control - then, it seems, once we take a decision drug, our control over whether we suffer its effects must come to an end. But, as we have seen, that is not what we believe of decisions. Even when there is no change in the beliefs which have led us to take a decision, and even when whether such belief change occurs lies outside our control - as when the relevant beliefs are about matters outside our control - we do not therefore suppose that our action control ceases. We suppose that Survival is true, and that we retain a continued freedom to act otherwise than as we have decided. It might be objected that the decision drug deprives us of continued freedom of action because taking it affects our subsequent desires. Whereas, this objector observes, taking a decision does not similarly affect our subsequent desires. The job of our decisions is to help fulfil our future desires - not to determine what those desires will be. However, this objection is not decisive. I have been using 'desire' in a natural, and commonly employed sense - the sense in which our motivation for action depends on which actions we desire to perform. At any time we can only be motivated to perform one of those actions which we then most strongly desire to perform. In this sense it is true that, in the absence of any incapacity, how at any time we act depends on how we then desire to act. And it is by employing this sense of 'desire' that I claim that decision drugs affect action byway of affecting our subsequent desires to act. But of course, on this same conception of desire, if our decisions are to affect our subsequent actions then they too must affect our subsequent desires. I have already characterised decision-making as a motivation-influencing activity. Our decisions can only determine how we act in the future by determining how, in the future, we shall be motivated and so disposed to act. In affecting future desires, then, decision drugs are producing no effects which decisions do not have to produce as well. Decisions and decision drugs seem very alike. In each case, taking them is a form of agency which causes a future action by influencing future motivation. And in each case the effects of taking them are belief-dependent too. Discrediting of the beliefs which motivated us to take them would counteract those effects. Yet taking a decision does seem, nevertheless, to be rather a different matter from taking a decision drug. Taking a decision

98

The psychology of freedom

always leaves us free to act otherwise than as that decision causes us to act. Whereas decision drugs do not nearly as obviously leave us free to act other than as those drugs cause us to act. There is a further difference between decisions and decision drugs. We have seen that, very plausibly, decisions are rationalitypreserving causes of action. The rationality of deciding to do A, guarantees the rationality of doing A as a result thereafter. Rational decisions lead to rational actions. Now, just as there is no equally plausible analogue of Survival for decision drugs, so too the effect of decision drugs on action is not as obviously rationalitypreserving either. Consider the decision drug for performing some action A - a drug taking which would cause me to do A thereafter. Even supposing I did take that drug rationally, that certainly would not guarantee the rationality of my doing A thereafter. For example, what made taking the drug rational could be some huge reward simply for taking it - a huge reward I should obtain as soon as I took the drug, and whether or not I did A thereafter. Now such a reward might be big enough to compensate me for any ill effects from doing A through the drug's influence. But even so, the reward obviously would not do anything to justify doing A. It is, after all, taking the decision drug, not doing A thereafter, which actually wins the reward. Nor need there be any other good reason for me to do A once I had taken the drug, and there could perfectly well be some considerations against. So no matter how rational it was for me to take the decision drug, by leaving me motivated to do A the drug could perfectly well be diminishing my rationality, and not preserving it. The effect of a decision drug on subsequent action, it seems, need not be rationality-preserving. The rationality of deciding to perform an action, guarantees the rationality thereafter of acting as decided. But it is much less obvious that the rationality of taking a decision drug, guarantees the rationality thereafter of whatever action it might be which that drug motivates and causes. We shall be returning to consider further this point of contrast between decisions and decision drugs. That Survival is true, and that decisions are rationalitypreserving causes of action, are surely connected. That is why I could, with some plausibility, appeal to the rationality-preserving nature of decision-making to back up our intuition that Survival

Decision-making and freedom

99

must be true. But to explain Survival, we cannot afford to rely simply on the claim that decisions are rationality-preserving in their effects on action. For the question then arises why decision taking should be rationality-preserving in its effects, when there is the apparently similar activity of decision drug taking which - it seems -would not be. In both cases, we have a form of agency which causes us to perform an action by ensuring that, thereafter, we are motivated to perform it. Yet there seems to be something very special about decision-making in particular - something which prevents the activity of decision-making from directly impairing our freedom or rationality. It is clearly not a trivial matter to explain why Survival is true - why the second-order agency of deciding is guaranteed to be freedom-preserving when other, otherwise similar action-causing first-order actions, such as decision drug taking, might not be. So we need to give further account of the links between our decisionmaking and our freedom. We shall need this account not only to explain Survival, but to provide an explanation of Dependence too. CONCLUSION

This chapter has raised the question of how a freedom and agency of the will might matter to our freedom to act otherwise. It seems we believe in Dependence — a dependence of our freedom to act otherwise on a prior freedom to decide otherwise. If there is such a dependence, this is going to be explained by a combination of facts (a) about the relation between decisions and actions and (b) about the nature of freedom itself. As regards (a), I have argued that decisions function as actiondeterminants. By determining that we act as decided, our decisions ensure that we consistently apply practical reason through time. If decisions do function to determine subsequent action, then that fact can help explain a dependence of action control on decision control. But as regards (b), the precise explanation given varies depending on which one of two very different theories of freedom we employ. Incompatibilism ties our action control to the absence of uncontrolled determining causes of action. So Incompatibilism does imply a dependence of action control on prior decision control — in those cases, but only those cases, where actions actually are

i oo

The psychology of freedom

preceded and determined by prior decisions. Incompatibilism says nothing about whether and why free agents need a decision-making capacity in the first place. A Psychologising conception of our freedom, by contrast, does actually tie our freedom of action to our possession of a decisionmaking capacity - a decision-making capacity which must be free in turn, and through which we can freely determine that we act rationally thereafter. And it seems to be the Psychologising conception which best explains our belief in a dependence of action control on decision control. We are natural Psychologisers in the way we think about freedom. We believe that it is up to us which actions we perform only because we can decide which actions we shall perform, and it is up to us which such decisions we take. It is our capacity for free decision-making, we believe, which makes us free at all - a capacity exercising which can never of itself remove our freedom. Our Psychologising conception of our freedom leaves decisionmaking a very special sort of action. And the conception is based, I have conjectured, on a theory of freedom as requiring a capacity for rational self-determination. For what makes decision-making special among all forms of action-determining agency, is precisely the way that decisions impose reason on the actions which they determine. Rational decisions to act leave us disposed to perform rational actions. But quite why this should be, we have not yet explained. There could be other forms of agency, such as decision drug taking, which mimicked decisions closely - but without being either freedom- or rationality-preserving in their effects. So let us now examine further what more precisely the Psychologising conception of our freedom comes to - and what sort of conception of the will it assumes. If decisions to act are such very special actions themselves, what serves to make them so very different from the actions which they explain?

CHAPTER 4

The Psychologising conception of freedom

THE PSYCHOLOGISING CONCEPTION AND PRACTICAL RATIONALITY

We have a Psychologising conception of our freedom. Freedom of action, as we conceive it, depends on the possession of a psychological analogue of that freedom - on the capacity for a free, actiondetermining second-order agency of the mind. It is up to us which actions we perform only because we have the capacity to decide which actions we shall perform, and it is up to us which such decisions we take. The Psychologising conception of our freedom lies open to obvious objections. Why should our freedom of action presuppose any further kind of freedom? How can yet more of the same phenomenon be what makes the very occurrence of that phenomenon possible in the first place? We need to find some explanation if we are to avoid a vicious regress. For suppose free actions do have to be explicable by the exercise of a free will. What of this further freedom of will? Will not it follow that its operations must, in turn, be explicable in the same terms - by free second-order willings? The regress threatens to be vicious first because, as we have already seen, the will is not subject to the will: our willings - our decisions and intention formations do not appear to be explicable in terms of higher-order willings. And there must in any case be a finite limit to the orders of willing of which we are capable. To stop the regress, we need the right kind of explanation for why the Psychologising conception of freedom is true. The explanation must show that the dependence of freedom of action on a secondorder freedom of will is generated by a feature of action and action's relation to the will which is unique - which is not going to reappear IOI

102

The psychology of freedom

at the level of the will, and generate a further, higher-order version of the dependence. Once there did seem to be the required kind of explanation. It lay in the theory of rationality. From the Stoicism of late antiquity to the Scholasticism of mediaeval and early modern Europe, it was widely assumed that we are free agents only because we have a capacity for practical rationality - only because we have a capacity to recognise and apply rational requirements on action. And this capacity for practical rationality was supposed to involve a free will - a decision-making capacity the exercise of which lay within our control. Animals who lack such a decision-making capacity were held to lack the capacity for rational action - and, for that reason, they were held to lack freedom of action as well. It was assumed that there was a distinctive psychology of freedom. And this psychology was the psychology required for rational agency. The idea, then, is that freedom involves a capacity for rational selfdetermination. Freedom presupposes the capacity to apply reason in its exercise. And this idea is very plausible. Amongst other things, the idea explains why we do not readily think of many of the lower animals as free agents. Pace Hobbes, we do not naturally think of sharks and mice as free to act otherwise than as they actually do. But this might well be because we do not naturally think of sharks and mice as having a capacity for practical rationality either. There is a distinction between agents who can act rationally and mere agents. And this distinction may also be the distinction between agents who can act freely, and agents for whom freedom of action is not a possibility. But why should freedom involve a capacity for practical rationality — and how developed a rationality does freedom require? The answer to both questions is simple enough. Freedom involves, on any view, some capacity for self-determination. And selfdetermination worth the name surely presupposes, at the very least, that one can develop a conception of one's own good - of what is desirable, at least from one's own point of view - and apply that conception in what one does. The development and application of such a conception clearly involves a degree of practical rationality. And it is that degree of rationality, and no more, which I take to be a condition of our freedom. But, of course, practical rationality in its most developed form may come to much more than that. Perhaps, for example, Kant

The Psychologising conception of freedom

103

was right, and a capacity for full practical rationality requires a capacity for a rationality which is pure. One must be capable of being moved to act by practical laws that are formal — laws which can motivate one to, amongst other things, altruism. But, even if Kant were correct about what full practical rationality came to, we need not follow him in assuming that rationality thus understood is a condition of freedom. For it is plausible that such a capacity far exceeds anything needed for self-determination as I have understood it. DELIBERATION AND EXECUTION

We may be able, then, to defend a Psychologising conception of freedom precisely as past Psychologisers have sought to defend it - in terms of a theory of practical rationality. Freedom of will may turn out to be necessary to a capacity for rational selfdetermination. But in what way? Practical rationality, as we shall now see, involves two kinds of capacity. And the need for second-order freedom could arise in respect of either kind of capacity. First practical rationality requires a deliberative capacity - a capacity by which the agent can recognise rational requirements on action. This first kind of capacity is exercised in the formation of judgments about which actions are desirable - the practical judgments whereby one recognises justifications for action. Perhaps it is this deliberative capacity which involves second-order freedom. It might be that an agent who can form practical judgments ipsofacto has control over which practical judgments she makes. In which case, in so far as a free agent must be a rational agent, she must possess this deliberative freedom also. Secondly practical rationality requires an executive capacity - a capacity by which the agent can apply rational requirements on action. Without such an executive capacity, the agent's rationality would hardly be practical. How is this capacity exercised? There is one obvious way, at any rate. It is exercised, at the very least, in our first-order agency - in the very performance of the actions which practical reason requires. If freedom presupposes the capacity to apply reason in its exercise, then a free agent must possess an executive control - a capacity to apply her deliberations through the exercise of

104

The psychology of freedom

her action control. Perhaps, then, this executive control over her actions has to be exercisable through second-order agency; in which case second-order freedom is going to be necessary to a free agent's executive capacity. So we have two ways of conceiving of second-order freedom in relation to practical rationality. We can conceive of it either as a deliberative or as an executive freedom. Of the two ways, the second may be the less immediately attractive. Practical reason's requirements, we have so far assumed, are requirements on action - on our first-order agency. But that means that we already have in our action a perfectly good capacity for reason-executive agency. Why then should a free agent's executive control involve anything more than, simply, control of her actions? We shall not easily defuse the threat of regress which all Psychologisers face, by claiming that to execute a requirement on you to do A, you have to be able to do something else first, which then causes you to do A. It is less obviously regressive to conceive of second-order freedom as a deliberative freedom. And this is what many past Psychologisers have done. The Stoic will was, as we have seen, a capacity for practical judgment - for sunkatathesis. And Aquinas, as we have also seen, conceived freedom of will as a deliberative and judgmental freedom — as based on a freedom of the practical intellect. In so far as we are practically rational, Aquinas thought, we must have a capacity to deliberate and arrive at practical judgments. Where we need to deliberate - where it is not obvious which action is most desirable each of assent or dissent lies within our power, and is subject to our command. (Summa Theologiae, ia2ae, question 17, article 6, p. 196) And this freedom of practical judgment is a condition of freedom of action - a condition which non-rational animals fail to meet: That man is master of his actions is because of his ability to deliberate about them; because his deliberating reason is balanced between opposites, the will can choose either. But, as we have noticed, voluntariness in this sense is not present in dumb animals. (Summa Theologiae, ia2ae, question 6, article 2, p. 12)

In the Grundlegung, Kant also suggested that practical rationality involves a freedom of practical judgment. In that work he identified the will with an agent's practical reason:

The Psychologising conception of freedom

105

Only a rational being has the power to act in accordance with his idea of laws - that is, in accordance with principles - and only so has he a will. Since reason is required in order to derive actions from laws, the will is nothing but practical reason. {Groundwork, p. 80) Kant also supposed that a rational agent's will must itself be free. This was because rational agents must be the legislators of, as well as the subjects governed by, practical reason's requirements: A rational being must always regard himself as making laws in a kingdom of ends which is possible through freedom of the will. {Groundwork, p. 101)

It has been natural, then, for many philosophers to conceive of freedom of will as a deliberative freedom - a deliberative freedom which any practically rational agent will possess. Freedom of action depends on freedom of will, because freedom depends on practical rationality - and practical rationality involves a deliberative capacity which is free in turn. But such an explanation of Dependence faces problems. First, as I noted in chapter 1, our ordinary conception of our second-order freedom of will, seems to be a conception of an executive, rather than a deliberative freedom. Taking a decision seems to be something which we do on the basis of - or even, in some cases of acrasia, despite - our prior deliberations, and not as part of those deliberations. Taking a decision does not itself seem to be a deliberative move. Secondly, as we noted in chapter 2, practical judgments seem to be pro attitudes. Practical judgments seem to be justified purely and simply by the likely desirability of the actions which form their objects. A practical judgment that doing A is desirable is justified purely and simply by the likely desirability of doing A. So practical judgments are not means-end justifiable occurrences. They are not justified in terms of desirable ends which those judgments themselves are likely to further. They are not justified in the same way as the actions which they explain. But in chapter 2 I suggested that means-end justifiability looks a very promising candidate for constituting rationality in its distinctively practical form. And if means-end justifiability is reason in its practical form, there can be no agency and so no freedom of practical judgment. My own preferred strategy, then, is to show that free agents need a second-order freedom of will as part of their executive control.

106

The psychology of freedom

Free agents need freedom of will as an executive freedom. But how might one show this? One possible way of showing this has already been mentioned. It is provided by the theory that practical judgments are second-order desires. A judgment that it is desirable to do A is, on this view, a desire to desire to do A. Now, to be applied in one's action, such a judgment would have to move one to do A. How would a practical judgment qua second-order desire do that? Plausibly, as a secondorder desire, the judgment would motivate action by first getting one to form the desire desired - by getting one to desire to do A. So, in the course of applying practical judgments qua second-order desires, one would also be forming desires to act on the basis of desires to hold those desires. To apply practical reason, therefore, one would have to be able to desire as one desires to desire. The execution of practical judgments qua second-order desires would therefore require a Frankfurtian free will. If practical judgments were second-order desires, then a free agent's executive control would indeed take a second-order form. But there is still our old complaint. Why should an agent deliberating about how to act be interested in anything other than how to act? Why should a concern to perform actions for which there is justification be a second-order, rather than a first-order concern? The proposal that practical judgments are secondorder desires gives us the basis for a Psychologising conception of freedom. But the proposal is too ad hoc to satisfy. Our problem remains. If rational agents need be concerned with nothing more than how to act, why should they require any more by way of executive control than their action control? PRACTICAL REASON AND PLAN CONTROL

The answer lies in the structure of practical reason itself. It is true that rational agents need be concerned only with how they act. But there is a general structure to that concern - a way in which actions matter to rational agents. It is that structure which demands that executive freedom take a second-order form. In chapter 31 argued that our concern for our action extends over time. We care, and so reason practically, not just about the present, but also about the future. Now in reasoning practically about the future, we do more than reason just about the future consequences

The Psychologising conception of freedom of our present actions. We also reason about genuinely future actions: actions which depend on our future motivation - which we will perform only if, at some future time, we are then motivated to perform them. It is this fact that we reason practically about how to act in the future as well as about how to act now which is key to any explanation of Dependence. As I have already argued, the actions about which we deliberate have value for us only as parts of plans: as parts of temporally extended series of actions stretching, at the time of our deliberation, into or within the future - series of actions which, at the time of deliberation, depend on our future motivation. The value which we place on performing a particular action as we perform it typically depends on its being part of a plan; and we favour the action because we favour the plan of which it is a component as compared to other, rival plans. This is so whether we consider action designed to satisfy needs — such as hunting for food — or action designed to communicate information - such as everyday speech. In each case needs are satisfied and content communicated only if a whole plan is executed - only if a whole sequence of actions occurs over time. So practical reasoning is not characteristically between alternative actions in the present, but between alternative sequences of action extending into the future. We have almost no plan-independent concern for the present. Freedom requires, we have supposed, a capacity to apply our practical reason in its exercise. And I have claimed that the unit of practical reason is the plan. A capacity to exercise action control at the behest of our practical reason must, therefore, be a capacity to exercise control over plans. Since it is plans which at any time concern us — since we hardly ever have any plan-independent concern for how we act now — it is clearly over plans that our control should at any time extend. How can agents be free if at no time does their control extend to what really, and quite rationally, then concerns them? Without plan control, we should at any given time lack control even of how we satisfy basic needs, or of what we communicate. Freedom, therefore, requires a capacity to exercise control over plans. But plans are sequences of action extending into or within the future. A capacity for exercising control over plans requires, therefore, a capacity to exercise control over future action. How do we exercise such control? As we saw in chapter 1, our control of what

107

108

The psychology of freedom

happens comes from what we do - from our agency. If there is nothing I can deliberately do now which would affect whether or not 0 occurs, then, plainly, I must lack any present control over the occurrence of O. So if I do now have control over the occurrence of O, there must be some present agency which I am free to perform which gives me that control. Either that agency would itself constitute O's occurrence, or it would cause O's occurrence. Where present control over future action is concerned, the agency which gives me that control obviously cannot already constitute the action, but must cause it. Plan control requires that 1 control some present agency which causes the future action. And since future action causally depends on future motivation, if it is to cause my future actions that present agency must also cause the required future motivation. Decision-making — if we have control over it - is clearly a form of agency that meets this need. Given our control of our present decisions, those decisions' causal influence on our future motivation, and thereby on our future actions, is just what plan control requires. So what I call Dependence - the dependence of action control on decision control - is true partly because of a more general dependence: that of action control on plan control. Nor should it surprise us that the dependence of our action control on our decision control is explained by this further dependence. If free agents' action control need not extend into the future, over whole plans, why should it depend on their control of an event preceding action - their deciding to act - which, when controlled, precisely allows them to control action in advance? CONTROLLING FUTURE ACTIONS

But we still have not fully explained Dependence. For there are ways of controlling plans other than by mere decision-making. Rather than rely on mere decision-making to control our future action, we can surely make use of present action. Why must free agents possess a capacity to exercise plan control through present decision control in particular? Suppose deciding to act were not a form of agency over which we had control. Suppose, as Hobbes claimed, our control of what happens began with our actions and not with the prior decisions to perform those actions. Could not plan control still be provided by present actions - by cases of

The Psychologising conception of freedom present first-order agency - which had future actions amongst their effects? After all, there are plenty of actions which we do actually perform in order to control our future actions. For there are cases - albeit perhaps not typical or normal cases - where we think that simply taking a decision now to do A in the future might not be enough to ensure that we actually shall remain motivated to do A. These are cases where there is some doubt about the power of our present decision-making to determine our future motivation. If there is a risk that decision-making alone will not get us to perform an action, or refrain from its performance, we may supplement or replace our decision by making use of various kinds of action-determining action.1 The first method we can use is to perform actions which, rather than affecting our future action motivation, simply bypass it. We prevent ourselves from performing certain actions in the future by performing actions now which bind us - which put obstacles in the way of what our future selves can do. Ulysses famously did just this so that he could safely hear the Sirens' song. He knew that, once he heard the Sirens singing, his having decided beforehand not to leave his ship would not be enough to ensure that he remained on board. So he ordered his crew to lash him to the mast, preventing him from thereafter leaving the ship even if he wanted to. Similarly, forming a plan to give up smoking, and not trusting to any present decision not to smoke to motivate me to refrain in future, I might take care that I had no spare cash when passing any tobacconist. In so far as it bypasses future motivation, this first method for controlling future action is radically unlike decision-making. For decisions influence future actions precisely by influencing future action motivation. But there are other kinds of action-generating actions which are a little less unlike decisions - which are motivation-influencing too. Interesting discussions of the use of present first-order actions in future action control are to be found in Jon Elster, Ulysses and the Sirens, and Thomas Schelling, Choice and Consequence,

especially the latter's essays The intimate contest for self-command' and 'Ethics and the law of self-command'. Neither author considers the fact that, besides occasionally using first-order actions, we also, and more characteristically, exercise future action control through the second-order agency of decision-making. And so neither author compares the two rather different kinds of future action control.

109

11 o

The psychology of freedom

There are actions which affect our future motivation for action by changing the expected pay-offs on options. By performing these actions we alter the likely outcomes of our future acts. If I cannot do anything to guarantee that obstacles will not always be there to prevent my smoking if I so desire, I might instead do something to change the expected consequences of my smoking, so that these consequences become ones which will be markedly less attractive to me. Noting my strong concern for my reputation, I might warn my friends of my plan to give up smoking. By so acting I could count on my future fear of losing face to provide a further motive for not smoking, and thus prevent my ever abandoning my plan. Lastly, we can perform actions which, without affecting our expectations as to the likely outcomes of what we do, influence our future action motivation directly. We can take drugs or cold showers - actions which directly affect the strengths of various desires, and that way affect which future actions we are motivated to perform. The decision drug which I discussed at the end of the last chapter was an example of just such a means of future action control - an imaginary example which differs from the many drugs actually available only in the degree to which it mimics a real decision. Notice that these action-based methods of future action control are not exclusively first personal. They are the methods which we also use to influence the actions of other people. Indeed, they are the methods which we have to use to influence the actions of others. For the use of decision-making on its own is, by contrast, an exclusively first personal means of future action control. I cannot expect to determine, or even influence, which actions you perform simply by taking decisions about how you shall act. I cannot take decisions about how you will act, and expect that, without my having to do anything more, you will end up acting exactly as I have decided. To influence what actions you perform, I have to perform some action myself— an action which then influences your action in one of the ways already described. I might lock you up or otherwise bind you and put obstacles in your way. Or I might change your expectations about what will happen if you perform a given action. I might persuade you of what I claim to be the facts, or issue promises or threats. Finally I might simply drug you, or otherwise subject you to treatments which, other than by way of affecting your beliefs,

The Psychologising conception of freedom directly affect which actions you are motivated to perform. I might put something in your tea. To exercise control over how you act, I am dependent on using first-order agency. I have to perform some action myself. Why could not there be a free agent who did exercise control over his own future actions - as we have agreed free agents must - but similarly depended on using actions for so doing? Why cannot there be free agents who can only control their own future actions as they control the actions of others - through performing actions? First-order agency can give us future action and so plan control. But if Dependence is true, our freedom of action depends not just on our having plan control, but on our being able to exercise that plan control through the exercise of free will - through performing second-order agency. There had better be some story why. It cannot simply be that, as a matter of fact, decision-making provides our principal means of controlling our own future actions. That may be true. In the first personal case, we may indeed rely far more on merely taking decisions than we do on action. But this difference is still one of degree. The simple extent of our reliance on mere decision-making for future action control will not on its own explain why our action control should depend on our having a prior decision control in particular. To explain Dependence, there must be something distinctive about the future action control provided by decision-making as compared to that provided by present action. Decision-making must give us a control over our own future actions which is special - a plan control which free agents really need. And so we arrive at what fundamentally distinguishes the use of decisions as means to future action control. What is distinctive has again to do with the fact that practical reason governs the actions of agent continuants, and is structured around plans. WHY FREE AGENTS NEED PLAN CONTROL FROM FREEDOM OF THE WILL

The availability of plan control

To apply practical reason through our action control, we need a method of controlling future actions which is available to us when, as practical reasoners, we need it. We need a method which we can

111

112

The psychology of freedom

in general use whenever the planning and co-ordination of our action demands. It may be that I need to know now where I shall be spending my holiday, because now is the best time for planning how to get there. It is now that I have the information about transport about routes, times and prices - at my disposal. So I should be in a position, simply on the basis of my judgment about what the most desirable holiday location would be, to do something now which settles and so can inform me where I shall be going. Decision-making provides the necessary method. It is a method of plan control which I can use, and in general use effectively, simply on the basis of forming judgments about how it would be desirable to act. The method does not depend on special resources which, as a deliberator, I might lack. Unlike the use of present actions, the use of decisions to control future actions does not depend on the availability of ropes and gags, willing accomplices, drugs, cold showers, or friends before whom I could humiliate myself. To exercise future action control through decision-making, I do not even need to have formed beliefs about whether or how my decisions will cause the future actions upon which I now decide. For, as we have seen, the second-order agency of the will can be motivated by nothing more than the very same attitudes as would motivate the action decided upon. I can take a decision to go to the dentist tomorrow on the basis of nothing more than beliefs about what desirable ends might be furthered by my going to the dentist tomorrow. I do not need any beliefs about my decision and what effects taking it would have. Simply on the basis of judging it desirable that I go to the dentist, I can take the decision to go — and so, eventually, I act. Whereas, if I am to use actions to control my future actions, means must not only be available. I must also have worked out that the means are available - that performing certain actions now will cause me to perform certain actions in the future. I must possess the right beliefs - beliefs which need no more be available to me simply as a deliberator, than are the action-based means which they concern. Plan control and plan execution

But decision-making possesses a further and quite crucial advantage. Decision control not only allows us to exercise plan control,

The Psychologising conception of freedom thereby applying practical requirements which are plan-centred. It allows us to do so in a way consistent with the rationality of the actions which, thanks to the decision, will later execute the plans. Decision-making not only gives us a capacity for rational plan control; it also allows us an equal capacity for rational plan execution. This is a capacity which we also need if we are to be capable of applying reason in what we do; and so this capacity is also a condition of our freedom. I have claimed that we count as free agents only because of our capacity to apply reason in what we do. Now, as we have seen, practical reason characteristically recommends plans of action extended over time, and action at a given time derivatively, through such plan-centred recommendations. But that means that applying practical reason involves two kinds of agency. First, as we have just been discussing, there is the initial agency through which we ensure our performance of whole plans - the agency of the initial decision to perform all the actions planned. This is the agency by which, as free agents, we exercise plan control. But secondly, there is also the subsequent agency of plan execution - of actually performing each of the actions making up the plan. The same practical reason which recommends plans to us, ipso facto recommends the actions which compose the plans. Each of the actions making up a plan is justified by the same considerations that justify the plan which they compose. Take the action of turning the corner to the shops — an action that forms part of a plan for getting food. What justifies that action is what justifies the plan of which it forms a part. It is the consideration that thereby I can get and eat food. So the practical reason which recommends plans of action to us must be applied, not only in the prior agency which causes the whole plan - in our exercise of plan control - but also in the agency which executes each stage of the plan - in the eventual performance of the actions planned themselves. Only then will our practical reason be applied throughout the agency which it governs and recommends. Rationality is a characteristic intimately linked to our identity through time. Rationality is a characteristic which we have as agent continuants. It is because we are agents who persist through time, that reason recommends actions to us in the form of plans or sequences extending through time. So too it is because we are agents who persist through time, that our rationality involves being

113

114

The psychology of freedom

able to apply those recommendations consistently through time, in the performance of the actions planned themselves. Our decisionmaking capacity gives us that ability. However we exercise our decision-making capacity, its exercise leaves us disposed to perform each of the actions making up a plan for the same reasons as motivated us to decide to perform the whole plan they compose. Free agents must be able to apply practical reason. But practical reason is plan-centred: it governs and recommends series of actions extending through time. Hence, as continuants, free agents must not only be able to exercise plan control. They equally need an ability to apply practical reason consistently in their agency through time. Our capacity to take decisions - our will - is, I conjecture, what, through our control of it, gives us both abilities. Through our control of our present decisions, and the influence of these decisions on our future actions, we ensure that we consistently apply practical reason not just now, in our deciding, but hereafter in our future action too. For our decisions cause the future actions decided upon by causing us to be moved in the future to act by the same considerations as now move us - through our decision-making - to cause those future actions. Decision-making allows us to control our future actions by perpetuating our present motivation and rationality into the future. The use of first-order actions to control our own future actions is quite different. It is something we resort to only when we lack confidence in our ability to perpetuate our present motivation through decision-making - when we doubt that just taking a decision would leave our future self as responsive as it should be to the considerations that motivate us now. Given such doubt, we use actions to manipulate our future self into doing what we want it to. Action-based methods are manipulative precisely because our future self is not left motivated to act as we want it to by reasons which it shares with us. Either actions simply bypass our future rationality altogether - as when we introduce Ulysses-style physical restraints or simply drug ourselves into action. Or the method used may constitute a form of moral arm-twisting. By acting now — by, say, telling our friends of our project to eat less, or promising our family that we will eat less - we are introducing quite new motivations, such as the fear of public humiliation, or the fear of promise-breaking, to goad ourselves into performing future actions

The Psychologising conception of freedom which at present, prior to the arm twisting, we favour on quite different grounds. We are getting our future self to act as we now want it to - but other than for any reasons that motivate us now. A FACULTY FOR APPLYING REASON

A free agent, I am arguing, must possess an action-causing faculty or power: a faculty or power by which she may apply a plan-centred practical reason by exercising plan control - and so control over her future actions. But practical reason, being plan-centred, governs and must be applied in her future action, as well as in the prior use of the faculty to control those future actions. So her exercise of the faculty by which she applies a plan-centred practical reason to control her future actions, must ensure that reason's consistent application in the very performance of those future actions also. In other words, a free agent must possess an action-causing faculty which meets the following conditions: (a) (i) The faculty must be exercisable simply on the basis of prior deliberation, and the agent must have control over how she exercises it. and (a) (ii) The faculty must be exercisable in a variety of ways; and its exercise in each distinct way must reliably cause a distinct future action. Why condition (a)? I have already argued that free agents must possess a plan control - a control that is available to them simply as deliberators; and that such control must come from the causal influence of some presently controlled agency on that future action. Condition (a) just spells out that demand. (b)

The causal influence of the faculty's exercise on future motivation and action would be counteracted by the agent's coming to hold beliefs which, if held initially, would have moved her not to exercise the faculty as she did. The faculty's causal influence on future action is belief dependent.

The rationale for condition (b) is clear enough. The job of this action-causing faculty is to apply a plan-centred practical reason, a reason which governs the agent's activity through time. So the

115

116

The psychology of freedom

faculty's job is to ensure the application of practical reason not in the present alone, but into the future as well. Now practical reason bids us be guided, where possible, by beliefs which are comprehensive and correct. So should it turn out, or at least appear, that the exercise of the action-causing faculty was moved by beliefs which were or could well have been, crucially false or incomplete, then that should counteract the effects of the faculty's exercise. Suppose that, through exercising the faculty, an agent ensures that thereafter he is motivated to do A. That motivation should not be one which the agent would retain irrespective of subsequent belief change. Beliefs which would have moved him not to exercise the faculty as he initially did, should counteract the effect of that exercise. (c)

The exercise of the faculty must be rationality-preserving in its effects on the agent's action explanatory attitudes and action. In particular, given that the faculty is exercised rationally, then its exercise should leave the agent disposed to act rationally too.

Why condition (c)? The job of the faculty is to ensure that, in the future, we consistently apply the recommendations of a plancentred practical reason - a reason which governs and justifies our action into the future. The faculty's function is to produce rational actions. Its rational use must, therefore, fulfil that function: it must leave us disposed to perform actions which, if performed, would be performed rationally too. (d)

In the absence of subsequent change in belief, the agent should be left with a persisting disposition to act for the same ends as motivated the faculty's exercise. The exercise of the faculty must be motivation-perpetuating in its effects on action.

The rationale for condition (d) is again clear. Free agents need a capacity for plan control. But they must also be able to apply practical reason consistently through time, in the actions which execute their plans. To allow free agents both capacities, the faculty which gives them plan control, and which is used to apply plancentred practical requirements by causing future actions, must also perpetuate its own plan-centred motivation in the actions caused.

The Psychologising conception of freedom

117

THE WILL AS A FACULTY FOR REASON-APPLYING AGENCY

Do we possess such a future action-causing faculty? Surely the will - our capacity to influence our actions through our prior decisionmaking - is just such a faculty. It is easy to see that, as the will is conceived in common-sense psychology, condition (a) is met. My capacity to take decisions is a capacity which I can exercise pretty much whenever I deliberate. I do not need any special resources for its exercise. I do not need ropes and gags, willing accomplices, drugs, cold showers, locks, or other special means for exercising self control - the means which would have to be available for me to exercise self control through present action. I can also exercise the capacity in a variety of distinct ways — in the taking of a variety of distinct decisions - and I have control over how it is exercised. Moreover the various ways I may exercise my decision-making faculty are distinguished, in part at least, by the distinct actions which they reliably produce. A decision to do A is distinguished from a decision to do B, so that the two merit their distinct contents, at least in part by virtue of the fact that the former reliably disposes and causes me to do A, the latter to do B. Condition (b) is met too. Suppose I decide upon an action only because I think it a means to certain desirable ends, and then cease to hold that view. Either I lose my belief that the action is a means to those ends, or I cease to judge those ends desirable. Then that change in my belief is going to counteract the causal influence of my original decision. I shall not go on to perform the action regardless.2 2

Notice though that this counteraction can itself be counteracted - by yet further, decisioninduced belief change. And that means that a decision can still lead me to act as decided even after I have come to hold beliefs which would certainly have moved me not to take the decision in the first place. Suppose I decide to drive to London by a certain road only because I believe that the road will be uncongested. And suppose that as I approach the road I see that, contrary to my initial belief, it is congested after all. It is certainly true that had I known at the outset that the road would be so congested, I would not have decided to take it. Nevertheless, I might still end up taking the road as originally decided and because of that original decision. Perhaps my decision to take that road led me to arrange to meet others along the road, and it is now too late to warn them of a change of route. Perhaps, having come so far, it is now too late to use another route. In this case the decision's initial effect on my motivation is still counteracted by my coming to believe that the road is congested. Were this my only new belief, my initial decision certainly would not ensure that I still took that road. But in this case the decision,

118

The psychology of freedom

It is intuitive that conditions (c) and (d) are met too. As we already observed in the last chapter, decisions are rationalitypreserving in their effects. If I take a decision to do A, and take this decision rationally, then I am left disposed to do A rationally too: given belief that the time decided on for action had arrived, I would do A - and my doing of A would be rational also. Not only is our exercise of the will rationality-preserving, but it is motivationperpetuating also. The same ends that originally motivated my deciding to do A are also going to motivate me to act as decided. In everyday life, we constantly assume that decisions are rationality-preserving and motivation-perpetuating. Once we have agreed that an agent's original decision to act was rational then, unless we suppose that she should subsequently have revised the assumptions on which that decision was based, we will not go on to quarrel with the rationality of her performance of the action decided upon. And once we know the reasons why an agent has taken a decision to act then, unless we expect her beliefs may change meantime, we are not ever at a loss to predict for what reasons she will be acting as decided. If the agent's decision to eat less has been motivated by a concern for her health, then that decision will also lead her to eat less out of the very same concern for her health. The same considerations that motivated her to decide to act will also motivate her to act as decided. Decision-making, as we conceive it, is not just free agency which gives us future action control. In virtue of being all of beliefdependent, motivation-perpetuating and rationality-preserving, decision-making has a plan-executory or reason-applying causal influence on future action. And that makes the will, as we ordinarily conceive it, a capacity or faculty for reason-applying agency.

through its effects on my action, has had further effects on my beliefs. And through these new beliefs I have acquired compensating reasons for continuing to act as originally decided. The decision has led me to believe that certain arrangements have been made with others which cannot be changed, or that it is impossible for me to follow a different route and still arrive in London by a certain time. And it is only because I now hold these new beliefs that I continue to act as decided. My thanks to Tim Williamson for leading me to make this point.

The Psychologising conception of freedom

119

EXPLAINING DEPENDENCE AND SURVIVAL

My claim is that Dependence can be explained in terms of rational requirements on free agents - requirements that derive from their own nature as continuants, and from the structure which their nature imposes on practical reason. Practical reason is plan-centred: it recommends actions as parts of sequences of action extending into the future. So to apply practical reason in the exercise of their action control, free agents need a faculty for plan control - a faculty which is available to them simply as deliberators, and by which they can exercise control over which sequences of future actions they perform. But free agents must also be able consistently to apply that very same plan-centred practical reason through time, in the very performance of those future actions. So their faculty for exercising plan control must be plan-executory in its effects on future action. By exercising this faculty agents must be able to ensure not only that they perform a given plan, but also that they consistently go on applying a plan-centred practical reason into the future, in the very performance of the actions making up that plan. The will is just such a reason-applying faculty. There is no other. For any capacity for agency which we can exercise just as deliberators, and which has a plan-executory effect on future action -which determines future action in a belief-dependent, motivationperpetuating and rationality-preserving way - is a faculty of will. In explaining what kind of future action-controlling capacity a free agent needs, we have also explained exactly what it is to possess a decision-making capacity at all. So now we see why action control depends on decision control why freedom of action depends on a second-order freedom of will. When we imagine away our control of our decision-making, we imagine away our freedom. And we imagine away our freedom because we imagine away our faculty for exercising future action control in a reason-applying way: for without decision control, the will is no a longer a future action controlling faculty at all. In chapter 3 we saw Sidgwick condescendingly describe as something 'vaguely thought' that a belief in Free Will requires us to maintain that at any moment we can alter our [future actions] by a sufficiently strong exertion [of will]

120

The psychology of freedom

But now we can see that there is nothing vague or confused about this thought. Remember our paradigm of freedom. A free agent is an agent who determines how he will act by his own free decisionmaking. Now we can see why. To determine our future actions through our own free decisions is to make effective exercise of our capacity for rational self-determination. It is to exercise effective control over how we act in a way which (a) applies a plan- and future action-centred reason but (b) does so consistent with our rationality through time. Our conception of the will is a conception of a faculty the influence of which on future action makes us free. What Sidgwick terms something Vaguely thought' can be given a clear explanation - the explanation provided by a unified theory of freedom and the will. And I claim that what explains Dependence is what explains Survival too. We can now complete the explanation, which we began at the end of the last chapter, of why decisions, even actiondetermining decisions, never take away our freedom to act other than as we have decided. There I claimed that what makes Survival so plausible, is the fact that decision-making is a faculty for agency which is rationality-preserving in its effects on future action. By exercising the will rationally, we thereby impose, through our own agency, a law of reason on ourselves. Through our own doing we ensure that in the future we shall be performing actions which reason recommends. We can now expand on this account of Survival. By exercising the will rationally, we not only ensure that we perform actions for which there are sufficient justifications. We also ensure that we shall be performing those actions for the right reasons — that we shall continue in the future to be motivated by the same justifications as rationally motivate us now. And because of that, the actions which our rational decisions cause us to perform will be fully rational too: they will be the actions which practical reason recommends, motivated by precisely the rational considerations that recommend them. Survival is true because in decision-making we, through our own doing, bring about the continuity of our rationality through time. When I take a decision I ensure, by what I myself do, that my self now and my self in the future share a common reason - that we are both motivated by the same considerations. Through my own agency, I thus ensure that I shall persist into the future as someone

The Psychologising conception of freedom

121

with whose rationality I can now immediately identify. How could the operation of such an agency end my freedom? HOW DECISIONS CAUSE THEIR EXECUTION

We have our reasons for performing the actions which execute our decisions, just as we will have had our reasons for the original decisions to perform them. A decision taken for reasons disposes and causes its taker to act, not blindly, but for reasons too. I have also claimed that decisions are rationality-preserving in their effects. Decisions to do A which are rationally taken dispose and cause their takers to do A rationally too. And that means that when a decision causes its maker to act for reasons, those reasons for so acting must be as good as the reasons which motivated the decision in the first place. So, decisions taken for good reasons cause the actions decided upon to be performed for equally good reasons too. But there is more than one way decisions might be thought to do that. I have suggested one way. A decision to do A ensures that the agent is motivated to do A by perpetuating its own motivation. The decision causes the agent to act as decided for the same reasons that motivated it. And that, in my view, is why decisions are rationality-preserving too - why decisions taken for good reasons cause actions which are performed for good reasons likewise. The reasons which motivate the decision and the action are the same. But there is another way - or so it might be thought. A decision to do A could cause the action decided upon by providing an additional motive or reason for performing it - a motive or reason that is good for any case where the decision to act was rationally taken. On the alternative theory, then, it is only by providing an additional motive for acting as decided, that decisions cause their execution. It is important to see why of these two theories of decision execution, my own Motivation-Perpetuation theory is to be preferred. The Motivation-Perpetuation Theory

If your decision to do A does get you subsequently to do A, you must end up acting as decided for some reason or other. One possibility

122

The psychology of freedom

is that when you do A, you will be doing A for the same reason you initially decided to do A. My claim is that this is exactly what your decision ensures. Desires for various ends E will have motivated you to decide to do A. The decision ensures that in the absence of intervening change in belief, you will remain motivated to do A in order that E. The decision perpetuates its original motivation. But what does that involve? In chapter 2 we saw that one's reasons for taking a decision to do A are characteristically nothing other than reasons for doing A. If I take a decision to go to the dentist, I need be motivated by nothing more than a desire to cure my tooth-ache, and a belief that going to the dentist would provide the cure. Now I am eventually going to argue that decisions may on occasion be motivated in a more complex way. But these other, more complex cases are far from typical.3 To keep matters simple, we shall for the moment develop the Motivation-Perpetuation theory by considering the characteristic case alone. In the characteristic case, then, I shall be motivated to decide to do A by nothing other than reasons for doing A: by desires for ends E combined with beliefs that doing A would further E - attitudes which already leave me with a desire to do A. The function of the decision, then, is to ensure that these attitudes which have motivated me to take it, and which already rationalise my doing A, continue to motivate me to perform the action which they rationalise. The decision ensures, therefore, that these attitudes continue to motivate me to deliberate about how to do A, and to decide on and to perform whatever other actions the doing of A requires or involves. All this further action will be motivated, just as was the initial decision, by a desire to attain There is an obvious objection to this conception of decisionmaking. If decisions do no more than perpetuate their own motivation, are they not redundant? Why cannot the beliefs and desires which, according to my theory, motivate both the decision and the action decided upon, motivate that action by themselves 3

See chapter 8 where I argue that there are atypical cases where one's reasons for taking a particular decision to act are not, prior to that decision, reasons for acting as decided. In chapter 9 I shall be extending my account of motivation-perpetuation to explain precisely how it works in these atypical cases too.

The Psychologising conception of freedom

123

without there having to be any prior decision? What value does decision-making add? The answer is that there are two important roles which decisions play. Decisions act as tie-breakers, and as desirestabilisers. (a) Decisions as tie-breakers

First, an agent's beliefs and desires may not, of themselves, determine how that agent will act. They may not on their own determine which future action the agent is motivated to perform. Action motivation I take to depend a priori on desire. An agent can be motivated to perform an action only if it is one of those he wants most strongly to perform. But there might be a number of such alternatives by way of action no one of which is preferred by the agent to any other. The actions might be ones between which the agent is indifferent - he really does not mind which he performs. The agent might, for example, be presented with a number of equally good means to a single end. In these cases the agent's beliefs and desires will not in themselves determine which action the agent will be performing. So what does settle the matter? At a time of action, the matter could be settled by the agent's simply performing one of the actions. But people also need to know in advance which actions they will be performing. They need to know this to plan the rest of what they do. In which case, if beliefs and desires alone do not already settle the matter - if there are a whole lot of options between which the agent is indifferent - it must be possible to do something now which breaks the tie. Actually performing one of the actions is not yet a possibility. The time for that has not arrived. So an agent must instead take a decision to perform one of the actions — a decision which is motivated by the agent's reasons for performing that particular action, reasons which that decision ensures will thereafter motivate and explain the action itself. Decisions, then, can function as tie-breakers. Where an agent's reasons for action do not on their own suffice to determine in advance how that agent will act, decision-making settles the matter. Notice that in so doing, the decision can leave the agent's feelings about the actions between which he has decided quite unchanged. Having decided on one out of a range of equally good alternative means, the agent's decision can still leave him as indifferent between the means as he was before. Deciding on one

124

The psychology of freedom

means in particular settles which means the agent is motivated to adopt. But to do that the decision does not have to leave the agent with a novel and suddenly acquired preference for employing that means over any other. The decision does not have to leave the agent wanting more strongly to adopt that means than any other. Why should it? That a particular means has been decided on will not in itself make any difference to how good it is as a means. This is a point to which we shall return. (b) Decisions as desire stabilisers

But decisions cannot be mere tie-breakers. For we bother to take decisions to act even in cases where there is no need for tiebreaking - in cases where we already wanted to perform the action decided upon more strongly than we want to perform any other. Decision-making occurs even in cases where our beliefs and desires already settle which action we are motivated to perform. When offered a choice between holidaying in Germany and France, I may already want to holiday in France more strongly than I want to holiday in Germany. But that will not stop me taking a decision to holiday in France, and that decision causing me to act as decided. So what role can decision-making play in such a case? There is only one role left for the decision to play. Even if, prior to any decision, we are already motivated to perform a particular future action A, there is still the question of whether we shall remain so motivated into the future. If how we shall act is to be settled in advance - as the co-ordination of our action over time demands - there must not be a possibility that our motivation will change. The point of taking a decision in these cases, then, is to determine action in advance by ensuring that we do remain so motivated. How does the decision do that? Since the agent's beliefs and desires already suffice to motivate a particular action, the contribution of the decision must be to ensure the retention of some or all of these attitudes. But which? It cannot be the function of a decision to ensure the retention of the beliefs that motivated it. These beliefs constitute the assumptions on which the decision is based. In particular, these beliefs may include practical judgments about which ends and actions are desirable. These beliefs may include the very view of practical reason's requirements which it is the function of decision-making to apply. Rather than ensuring the retention of the beliefs which

The Psychologising conception of freedom

125

motivated it, a decision's causal influence should be conditional on those beliefs being retained. The decision's function, therefore, must be to ensure the retention of the desires that motivated it. Decision-making functions as a desire-stabilising agency. For as long as the beliefs which have motivated a decision remain unchanged, the decision should ensure that the desires that motivated it remain unchanged also. And that way it can be true, as Bratman has rightly claimed, that intentions possess inertia. Intentions tend to persist until the time of the actions intended. That means that deciding to perform an action - the formation of an intention to act - ensures a persisting motivation to act as decided. Since motivation depends a priori on desire - since an agent can only be motivated to perform one of those actions he desires most strongly to perform - stabilising desire helps stabilise motivation. Stabilising desire ensures that from the time of decision onwards, the agent retains desires that allow him to be motivated to act as decided. The Additional Motive Theory

There is a quite different theory of how decisions cause their execution. On this Additional Motive theory, one's motives for executing a decision to do A are never simply the same motives for doing A which originally led one to take the decision. For, on this theory, that one has decided to do A always provides an additional motive for doing A - a motive which, prior to the decision, one did not have. And it is only by providing this additional motive that a decision ever causes its execution. This Additional Motive theory assumes that every decision maker holds a preference for decision execution - or at least for executing those decisions which they take rationally. The theory further assumes that taking a decision to do A will always lead one to believe that one has so decided. This belief then combines with one's preference for decision execution, and so one comes to hold an additional motive for doing A. One comes to want more strongly than not to do A because to do A would be to do what one has decided. It is only by thus adding to action motivation, and not by perpetuating their own motivation, that decisions cause us to act as decided. This theory assumes that the execution of decisions - or, at any

126

The psychology of freedom

rate, the execution of rational decisions — is desirable in itself. Why else should all decision makers, without exception, hold a preference for executing whatever decisions they might take? But we should reject the Additional Motive theory. For I reckon that the execution of decisions — even decisions that have been taken rationally - is not desirable in itself. Nor can we assume that all decision makers care about decision execution for its own sake. It is of course true that when we fail to do what we have decided, this typically leads us to feel regret. But that is because of the frustration of the ends that motivated us to take the decision. The failure to execute the decision need not bother us in itself. To see this, we need a case where despite one's failure to execute a decision, all the ends which motivated one to take the decision in the first place have been attained. In this sort of case, the decision's non-execution provides the only reason for regretting one's failure to do what one has decided. Nothing else has gone wrong. We shall find that in this sort of case, one has nothing to regret. It is not hard to produce a suitable case. Consider the following example. I am faced by an array of two kinds of biscuit, brands A and B. I want to take one, but only one packet of biscuits, and it really does not matter to me which packet or brand. So motivated by this desire and belief, I decide to pick out one particular packet - a packet of brand B stuck saliently amongst the brand A packets. Notice it is that particular packet which I have decided to pick out and so have reached for. I have not just snatched randomly at the array. And that is because - but only because - random snatching is a hopeless way of picking biscuit packets up. My sole concern is to pick up some biscuit packet or other — I really do not mind which. But doing that at all efficiently does mean deciding on and reaching out for one biscuit packet in particular. Now when I have taken my decision to pick out that particular B-brand packet, should it matter to me whether I in fact obtain it, as opposed to one of the other packets on display? It is not obvious that it should. Even after my decision, I can perfectly well still remain indifferent about which packet I pick out. Suppose when I get to the checkout counter I find that, contrary to what I intended, I have picked out a packet of brand A. As I was reaching, just for an instant another customer obstructed my view of the array. Then, quite by chance, although thanks to the momentary obstruction I didn't see it, a brand A packet - the one I

The Psychologising conception of freedom

127

have ended up with - fell from a higher shelf and dislodged the intended brand B packet. This discovery might leave me surprised. But it is not obvious that I have anything to regret. It is true that my intention to pick out that brand B packet was formed and held rationally. It is true, therefore, that I have failed to execute a rational decision. But why should that matter? Though I have failed to execute my decision, the end which motivated it - the end of getting hold of one biscuit packet or another - has still been perfectly attained. The taking of any one of these packets is as good a means to this end after the decision as before. Why should I regret obtaining a brand A packet as opposed to a brand B packet? If decision execution were desirable in itself, then rational agents would regret ending up with a brand A packet in this case. If they could pay some small enough sum in money or goods in advance to prevent such failures to execute their decisions - failures where, however, nothing else goes wrong - they should be willing to do so. After all, being rational they are supposed definitely to prefer such cases not to occur. But why waste any sum at all, no matter how small, on such expenditure? Decision execution, I suggest, is not intrinsically desirable. It is not a condition of my rationality that I should hold a preference for decision execution for its own sake. And even if, contrary to what I claim, it is irrational not to possess such a preference, the irrationality is not obvious. On the contrary, it is highly disputable. And this leads to a further, quite conclusive objection to the Additional Motive theory of decision execution. The theory purports to give the basic mechanics of decision execution. It purports to explain how, quite generally, the decisions of decision makers cause those decisions' execution. The preference to which it appeals - a preference for decision execution for its own sake - must therefore be possessed by decision makers generally. But the rationality of this preference is controversial. And so the preference for decision execution is one which decision makers can perfectly well lack. Decision makers who may be rational enough otherwise, can perfectly well fail to care at all about decision execution. All they need care about is attaining the ends - such as getting hold of biscuit packets -which motivate them to take their decisions in the first place. The causal powers of our decisions cannot, therefore, plausibly depend on our having any concern for decision execution.

128

The psychology of freedom

Some theories of the will liken taking a decision to a kind of promise which one makes to oneself. A decision is likened to a vow.4 Now it might well be true that the execution or fulfilment of a promise or vow is intrinsically desirable. We do regard the failure to execute a promise made to another, or even a vow made to oneself, as something regrettable in itself. And why this should be is a fascinating philosophical question which deserves discussion. But to initiate such a discussion here would only be a distraction. For can anyone, with a straight face, maintain that the discovery, at the checkout counter, that one has ended up with a biscuit packet other than the one intended is the discovery that, unwittingly, one has failed to be true to a vow? The Motivation-Perpetuation theory is much the most plausible theory of how decisions explain their execution. The only concerns which it attributes to decision makers generally are concerns which we anyway know them to possess - the desires for what can be perfectly ordinary ends such as food acquisition which motivate them to take their decisions in the first place. The MotivationPerpetuation theory just supposes that, through tie-breaking and desire-stabilisation, a decision perpetuates the motivating force of these desires, so that they thereafter motivate performance of the action decided upon as well. That way the theory avoids having to assume that decision makers have an additional concern for decision execution. And, since decision makers can perfectly well lack such a concern, that is the Motivation-Perpetuation theory's strength. Imagine yourself in the supermarket, tired out and at the end of a long day. You decide to pick out a particular biscuit packet and then pick that packet out. As you're doing all this, need you care about anything beyond getting hold of a packet of biscuits?5 4 5

My thanks to John Campbell for expressing this idea to me so vividly by using precisely this term. I think we may reject Velleman's ingenious but contrived assumption, in his Practical Re/lection, that we do all care about decision execution - that, as decision makers, we are all motivated to execute our decisions by a desire to prove right our decision-based expectations about how we shall act. Suppose you're in the supermarket, picking out that packet of biscuits you decided on. Why should you care at all whether you were right about which biscuit packet you would pick up? It plausibly does not matter one iota whether you were right. And since that it does matter is anyway not obvious, you cannot be assumed to care in any case. Happily, my Motivation-Perpetuation theory makes Velleman's debatable motivational assumption quite unnecessary. We can account for how decisions explain their execution without it.

The Psychologising conception of freedom

129

Two reasons why decision execution might look intrinsically desirable

My Motivation-Perpetuation theory does not actually require the assumption that decision execution is not desirable in itself. All it strictly requires is the assumption that, because the intrinsic desirability of decision execution is not obvious, decision makers may perfectly well fail to desire decision execution for its own sake. But, nevertheless, I do also maintain that decision execution is not desirable in itself. Why should anyone suppose otherwise? There are in fact two important considerations, in particular, which might make decision execution look intrinsically desirable. The first consideration is that decision execution is indeed generally desirable. The second is a central principle of practical reason which the supposed intrinsic desirability of decision execution might promise to explain. This is the principle that it is irrational to intend an end without also intending the means. I shall consider these in turn. (a) The general desirability of decision execution

It is an important fact that decision makers, or at least those of them who are sufficiently rational and reflective, and who are aware that they have taken a decision, can perfectly well develop, and precisely for that reason, an increased desire to act as decided. And that fact might well suggest that decision execution is desirable in itself. But the supposed intrinsic desirability of decision execution does not follow. The facts bear another explanation. We do not need to suppose that decision execution is something desirable in itself that rational decision makers desire decision execution for its own sake. Suppose that, emerging from a hangover, I remember that yesterday I did decide to go to a given coffee shop this morning at 10.30, but cannot at all remember why. The thought that I have taken that decision may well increase my desire to act as decided to go to the coffee shop at 10.30. But, of course, this is not because I regard executing decisions as intrinsically desirable. The desirability of acting as decided is entirely derived. All that concerns me is performing actions for which there are justifications other than that those actions have been decided upon by me. That I did decide on going to the coffee shop is, in this case, evidence, and nothing

130

The psychology of freedom

more than that, for there being just such other justifications for going there - justifications which, unfortunately, I have quite forgotten. I want to execute my remembered decision — but only as a means to other ends. This is, happily, an unusual sort of case. But we can generalise. We do not have to stick with odd cases where we have forgotten the original justifications for a decision. Even in normal cases where one knows exactly what justified one's decision, the belief that one has taken it can easily increase one's desire to act as decided. And again, this is not because one wants decision execution for its own sake. We saw the real reason in the last chapter. There I argued that one way of looking at decision-making in advance of action is in terms of betting. By deciding on my actions in advance, I effectively increase the likely stake on a bet that I shall perform the act decided on. My position, as decision-maker, is like that of a bettor who, as he influences the outcome of a bet, also increases his stake. Hence, I argued against philosophers such as Sidgwick, decision-makers' decisions should determine that they do act as decided. A decision which determines action in advance is a particularly rewarding source of action control. Decisions to do A later - decisions taken in advance of doing A - exercise a co-ordinatory influence. They give rise to persisting intentions. And so they make it more likely that before doing A, I shall have performed other actions on the assumption that I may or shall do A. Such actions tend to increase the benefit I derive from doing A. I decide to go on holiday well in advance of going. My holiday is thereby better prepared. So the holiday provides me with more enjoyment. On the other hand, taking a decision to do A later, because of this co-ordinatory influence, also increases the likelihood of my performing actions I shall regret performing unless, in the end, I do A. Hence forming an intention to do A in advance of doing A may increase the loss if A were, after all, not done. If I fail to go on holiday despite having formed an intention to go, I not only miss out on my holiday — I may also lose money put down as deposits, or face a disappointed family whose expectations of a holiday I have earlier aroused, and so forth. It is better for me not to go on holiday having never decided to, than not to go despite having decided to. Hence deciding to do A in advance of doing A often increases the likely stake on A's being done. It often increases the likely benefit

The Psychologising conception of freedom from doing A, and it increases the likely loss from not doing A. This increase in the likely stake on A's being done is the greater if: (a) the decision to do A leads the agent to be very confident that he will do A - for then forming the intention will make him rely the more completely on his eventually doing A, thereby increasing the stake; and (b) the decision is taken well in advance of rather than nearer the time of A's being done - for then there is more time for the agent to perform other actions on the assumption that he will do A, thereby increasing the stake. So deciding now, rather than later or not at all, to do A typically increases what an agent is likely to gain were he to do A - and correspondingly increases what the agent is likely to lose were he not to do A. So (i) is often true: (i) There is more justification for an agent's doing A later given that he now intends to do A than there is given that he has not yet formed such an intention. But if (i) is true, then (ii) is true too: (ii) Deciding to do A later increases an agent's justification for doing A later. So rational agents who come to believe that they have decided to do A can indeed often be left with an increased desire to do A. But we can now see that the real explanation for this increased desire has nothing to do with the supposed intrinsic desirability of decision execution. The explanation lies instead in the fact that actions have value only as parts of well co-ordinated plans — only when accompanied by the performance of other actions at other times which match. The failure to execute a decision looks bad to rational agents only because it is likely to prevent the consistent performance of such well co-ordinated plans. The explanation lies, then, in a value which actions have apart from the occurrence of any decision, and in the plan-centred structure of that value — and not in the intrinsic desirability of decision execution.

131

132

The psychology of freedom

The Motivation-Perpetuation theory can therefore be supplemented. Besides tie-breaking and stabilising the desires to act which motivated them, decisions can also cause their execution by adding to those desires - at least in a sufficiently rational and reflective agent who is aware of all his decisions as he takes them. For given that one believes that one has decided to do A, one's general concern that one's actions be co-ordinated through time that one perform whole plans - combined with one's knowledge of the co-ordinatory effects of one's decisions, can leave one with an increased desire to act as decided. But let me emphasise that this is a supplement to the Motivation-Perpetuation theory as thus far developed, and not a replacement of it. Why? To begin with, not all decision makers need possess a sophisticated grasp of the function and effect of their own decision-making. Although decision-making functions to co-ordinate action through time, we should not assume that all decision makers are going have a developed understanding of this fact. Not all decision makers are capable of being that reflective about decision-making and its consequences. And those decision makers that are capable, need not always take their decisions that reflectively. In which case deciding to do A will not necessarily lead them to want more strongly to do A. The only way these decision makers' decisions cause their execution is in the way already explained - byway of tiebreaking and desire-stabilisation. But, most importantly, one's belief in the likely co-ordinatory effect of one's decisions — a belief that is essential to the increase in one's desire to act as one has decided -presupposes the general truth of the Motivation-Perpetuation theory as explained so far. For if decisions did not anyway ensure a persisting motivation to act as decided, they could not leave one with a persisting intention to act - and then decisions could not have any co-ordinatory effect at all. Decisions can only ensure a persisting intention to act as decided if they perpetuate their own motivation — if they ensure that the ends which motivate the decision thereafter continue to motivate the agent to act as decided. It is a constant human tendency to mistake what is very commonly desirable, but only as a means to other things - money is a good example of this — for something desirable in itself. That the execution of rational decisions is so commonly desirable as a means to the performance of whole plans - as we have seen that it is

The Psychologising conception of freedom

133

- could then, quite naturally, lead many people to begin, with dubious rationality, to find decision execution something desirable in itself. Let me emphasise, by the way, that it is quite consistent with my position to allow that someone should come to hold a preference for decision execution for its own sake. And if someone did develop such a preference, that preference could indeed provide them with an additional motive for decision execution. My position is simply, first, that rationality does not (obviously) require us to share this preference; and secondly, that this preference is irrelevant to the basic mechanics of decision execution. Those of us who find decision execution, considered in itself, a thing indifferent can still take decisions to act; and our decisions to act can just as well cause their execution thereafter. (b) Intending an end and intending the means

There is a further consideration which might lead one, again quite mistakenly, to believe in the intrinsic desirability of decision execution. Does not deciding on an end make it irrational not also to intend whatever means are believed necessary? And does not this suggest that executing decisions is intrinsically desirable?6 To quote Bratman, suppose I have to decide between two routes to get to San Francisco: My belief-desire reasons in favour of taking route 101 to San Francisco may seem on reflection equal in weight to those in favour of 280. Still, I must decide. As it happens, I decide - albeit arbitrarily - in favour of 101. Now I must figure out how to get there: I reason from my intention to take 101 to an intention to turn right at Page Mill Road. In this means-end reasoning. I treat my prior intention to take route 101 as directly relevant to the rationality of my derivative intention to turn right at Page Mill. That is why I see my intention to turn right as rational and an intention to turn left (toward route 280) as irrational, whereas before I formed the intention to take route 101 this was not so. {Intention, Plans and Practical Reason, p. 23)

Once I have decided on an end, it also becomes true — when it was not true before - that performing the means would execute a prior decision. Now if that fact could add to the desirability of performing the means, it would also justify intending the means. And then we would have an explanation of why deciding on and 6

My thanks to Joe MintofTfor the suggestion.

134

The psychology of freedom

intending the end makes it irrational not also to intend the means. But there is another explanation for why it is irrational to intend the end without intending the means. An agent cannot apply reason as it concerns his action - cannot be practically rational unless his decisions to do A leave him motivated to intend and to perform whatever actions he believes necessary to doing A. Without inducing such motivation, the agent's decisions simply will not reliably cause their execution. And that means that the agent will be unable to use decision-making to determine in advance which actions he will perform. Consequently he will be unable to co-ordinate his actions over time. And so we have our explanation for why failing to intend the means to an intended end constitutes irrationality. To fail to intend the means to an intended end constitutes irrationality because it is to fail properly to exercise one's capacity for applying reason. Notice - to obtain this explanation we do not need to assume the intrinsic desirability of executing decisions. We need appeal only to what it is to apply reason. We should distinguish those rational requirements that are end-derived, from rational requirements that are rationality-constitutive.

End-derived requirements are the justifications which motivate rational agents to decide on and perform actions. These requirements presuppose the desirability of particular ends. They justify intending or performing action on the grounds that there is some desirable end which that action is likely to further. Rationality-constitutive requirements, by contrast, do not presuppose the desirability of any particular end. They are not what motivate us to form particular intentions or to perform particular actions. Rationality-constitutive requirements just specify the proper functioning of our capacity for applying end-derived requirements. Rationality-constitutive requirements just specify what it is to apply reason. One example of such is the requirement that if we intend the end, we should also intend the means. There are other rationality-constitutive requirements besides the requirement that if we intend the end, we should also intend the means. For example, there is the requirement that we act continently or non-acratically. That is, out of any range of alternative actions which we believe available to us, we should not perform an action which we judge less justified. This is plainly a rationality-

The Psychologising conception of freedom constitutive requirement. The capacity to act as one's practical judgments dictate is part of one's capacity to apply reason. To act acratically is to fail properly to exercise this capacity. To explain the requirement that we should act continently, then, we do not have to assume that it is intrinsically desirable to perform the actions we judge most desirable - that the performance of an action can somehow be made more desirable simply by our judging that it is more desirable. Indeed, this assumption seems to be false. Judging that it is more desirable to holiday in Germany than in Francb will not of itself add to the desirability of holidaying in France, or diminish the desirability of holidaying in Germany. We do not need then to assume that a special desirability attaches to executing decisions once one has taken them. There is no such desirability. What motivates rational agents to stick with and execute their decisions is nothing more than what motivated those decisions in the first place. So when Bratman's rational decision maker decides to take route 101, and then, as a result, decides to turn right at Page Mill, all that happens is this. Motivated by her desire to get to San Francisco by some route or other, the agent decides to take route 101. Then, thanks to her decision, that same desire motivates her to do what she believes necessary to taking route 101. It motivates her to turn right at Page Mill. And so that is what she decides to do. Why need anything more be going on in a rational decision maker's mind? CONCLUSION

This chapter has justified and explained the Psychologising conception of freedom which we ordinarily possess. It has justified our belief in Dependence. It has explained why our freedom of action should depend on our having the capacity to control which actions we perform through the exercise of a free will — through taking free decisions. And the chapter has justified our belief in Survival. It has explained why the exercise of this capacity which makes us free, cannot take away our freedom. To explain all this, I have had to develop a new theory of the will. The will, I have claimed, is a faculty for second-order agency - a faculty which is distinguished from any capacity for firstorder agency which we might possess, by the fact that it has a reason-applying influence on future action. In particular, the will

135

136

The psychology of freedom

generates future action by perpetuating its own motivation. In exercising the will, we ensure that the reasons which motivated our decision to execute a plan, motivate us thereafter to perform the plan-executive actions decided upon. The theory of the will which I have developed is independently plausible. And that is because it fits with a general and quite fundamental intuition about decision-making. This is the intuition that decision-making has an executive function in relation to our action. Decision-making is all about applying reason as it concerns our subsequent action. The function of taking decisions about which actions we shall perform, is to ensure that we act rationally thereafter. And that is just what motivation-perpetuating and rationality-preserving decisions do ensure. By perpetuating their own motivation, decisions which are rationally taken cause actions which are rational too. My theory of the will, then, is true to decision-making as we ordinarily conceive it. But that ordinary conception of the will is far from a straightforward one. Indeed, it threatens to be incoherent. Our Psychologising conception of our freedom presupposes, I have argued, that we possess a capacity for decision-making which combines two features: (a) decision-making fulfils an executive function in relation to our first-order agency. Its function is to apply reason as it concerns our subsequent action — thanks to the motivation-perpetuating influence of the will, rational decisions to act leave us disposed to act rationally thereafter. (b) decision-making is itself a form of agency. Decisions to act are actions themselves. The problem which is about to face us, is that these two features seem to be inconsistent. It seems impossible for decision-making to combine them. This tension within the common-sense psychology of the will is the subject of the next chapter.

CHAPTER 5

Decision rationality and action rationality

DECISION-MAKING AND RATIONALITY

I have claimed that our Psychologising conception of our freedom is linked to a particular theory of the will — to a conception of decision-making as a special kind of agency which is reason-applying in its effects on action: - the will gives us control over our future action - so to take a particular decision to act is itself deliberately to perform an action. - but the will is also a reason-applying or executive faculty: its function is to apply reason as it concerns our subsequent action - thanks to the motivation-perpetuating influence of the will, rational decisions to act leave us disposed to act rationally thereafter. But is this psychology of the will a coherent one? There is a case for supposing that it is not. The claim that decision-making has an executive function - that a decision's function is to apply reason as it concerns subsequent action - may not after all be consistent with the claim that a decision to act is an action itself. For each of these two claims seems to imply a quite different theory of decision rationality. In chapter 2 I suggested that we explain agency in terms of the hypothesis that it is governed by reason in a distinctively practical way. There is a distinctively practical reason. I also suggested that reason in its distinctively practical form is means-end justifiability. Agency is justified in terms of the likelihood of its furthering desirable ends - the Means-End theory of agency and agency rationality is true. So if decisions to act are actions themselves, this general theory of agency implies that those decisions are themselves means-end justifiable occurrences. Deciding to do A is justified in terms of the likelihood that so deciding would further

138

The psychology of freedom

desirable ends. So decisions to act are governed by reason in the same general way as the first-order agency of our actions and attempts at action. I call this model of decision rationality the Action model.

But the idea that decisions to act have an executive function suggests a quite different model of decision rationality. If decisions to act have the function of applying reason as it concerns subsequent action, then surely what makes deciding to perform a particular action rational is, purely and simply, the rationality of performing that subsequent action. We explain the rationality of deciding to do A in terms of the rationality of doing A. But that means claiming that decisions to do A are justified by what makes doing A rational - by the likelihood that doing A would further desirable ends. So intentions and the decisions which form them are governed by reason in the same general way as pro attitudes such as desires and practical judgments. I call this model of decision rationality the Pro Attitude model. This then is our problem. What makes it rational to decide to do A - the fact that that decision itself would further desirable ends? or the fact that doing A — the action decided upon — would further desirable ends? Each one of these two competing theories of decision rationality, it seems, is supported by one element of the common-sense psychology of the will. But the theories are competing. Only one of them can be true. So which theory of decision rationality is true? And can the true theory, whichever it is, be made consistent with the common-sense psychology of the will as a whole? Before considering the two competing models of decision rationality in more detail, let me first say something about action rationality. RATIONALITY, JUSTIFICATION AND MOTIVATION

My claim about action rationality is simple. Two things, at least, are necessary for an action to be performed rationally. First, a rational action must be an action for which there is sufficient justification: the action must be the, or a, right action. Secondly, if an action is rational, the agent must have been moved to perform it by this justification, and not by some other non-justifying consideration: the action must have been performed for the right reasons. Let us first take justification: what is it for there to be some

Decision rationality and action rationality

139

justification for performing an action A? The following principle JUSTIFY provides a rough account: JUSTIFY

Any justification for doing A consists in the likelihood that doing A would further a desirable end E.

According to JUSTIFY, action is means-end justifiable -justifiable in terms of the ends which performing it is likely to further. Some comments, first about desirability and then about end furtherance. At this point I assume no particular doctrine about desirability. It might be that desirability is explained as Teleological or Consequentialist theories explain it — in terms of a quantity of good which rational agents seek to maximise or increase. On this view the desirability of an end consists in its attainment increasing the amount of good. But it may be that the desirability of an end, at least in some cases, has to be explained in non-Teleological terms. Perhaps, in these cases, a desirable end's attainment would meet a set of constraints on rational agency that have no direct connexion with an increase in good.1 By furthering an end I mean something like raising the objective probability of the end's attainment. This might be through causing the end's attainment - but end furtherance need not involve causation. Doing A might further end E by helping constitute the attainment of E: attaining E might, for example, consist simply in A's being done. Notice that what justifies an action is not end furtherance, but likely end furtherance. By 'likely' I mean what the evidence suggests, or what is epistemically as opposed to objectively probable. Consider the following example which I take from Michael Bratman. An agent is canoeing down a river, and arrives at a fork along one branch of which there lies a waterfall deadly to canoeists. If the waterfall is along branch A, then taking branch B will further survival. Let us take it that surviving is a desirable end. Does it follow that the desirability of surviving justifies taking branch B? It is not clear that this does follow. For example, all the evidence may suggest that the waterfall lies along branch B: perhaps ordinarily reliable authorities have so informed the agent, 1

In chapter 6,1 discuss Teleological theories of decision rationality: theories which explain justifications for taking decisions in terms of the likelihood that the decisions would increase good. I take the term 'Teleological' from Broome's Weighing Goods.

140

The psychology of freedom

or perhaps an illusion makes it sound exactly as if the waterfall is along branch B. In which case, the agent is surely unlucky enough to have every justification for taking branch A. Unqualified end furtherance - what ends an action actually would further — may be relevant to action value — to which action increases good. In our example, the best action is most certainly to take branch B if the agent will not otherwise survive. But the justified action may still be to take branch A if the evidence suggests that so acting would further survival: likely or evidentially supported end furtherance is what matters to action justification. As rational agents we may want, at least sometimes, to increase good. But even then we must, as rational agents, rely on the evidence for what ends our actions would further. And we may be unlucky: the most justified action need not be the best. But, having made the distinction between end furtherance and likely end furtherance, I shall for brevity ignore it in what immediately follows. I shall write loosely of end furtherance when, in discussing action justifications, I should strictly be writing of likely end furtherance. Justification alone is not enough to make an action rational. Even if there is sufficient justification for my doing A, I may still end up doing A irrationally. The ends which justify my doing A may have nothing to do with the reasons why I actually do A - with what moves me to do A. To determine the rationality of an action we need to consider, not only what justification there might be for performing the action, but also the reasons which actually move the agent to perform it. How do reasons for action work? The following principle, REASON, gives a rough account — the same account as was given at the beginning of chapter 1: REASON

Any reason for doing A consists in a desire for an end E combined with belief that doing A would further E.

The attitudes which, according to REASON, constitute reasons for doing A, are purposes or means-end related reasons for doing A. They consist in a desire for an end and a belief that doing A would be a means to that end. So a rational action is an action motivated by desires for ends which, through their desirability, also sufficiently justify the action. I shall in what follows write, for brevity, of ends motivating and justifying actions. Notice that when I do so, I mean that ends justify

Decision rationality and action rationality

141

through their desirability - a factor which may or may not have anything to do with the agent's own psychological attitudes. 2 And I mean that the ends motivate through being desired by the agent — a factor which certainly does depend on the agent's own psychological attitudes. In rational action, then, there is an identity between action-justifying and action-motivating ends. In the light of this outline of action rationality, let us now turn to decision rationality. I take it that decisions, like actions, are justified by desirable ends and that, as when we act, we are moved to take decisions by our desires for ends. To explain the status of decision-making as a reason-applying agency, we need therefore to examine the connexion between the ends which provide our justifications and reasons for action, and the ends which provide our justifications and reasons for deciding to act. And so we arrive at our two rival models of decision rationality. TWO MODELS OF HOW REASON GOVERNS THE WILL

(a) The Pro Attitude model

One model of how reason governs the will patterns decision rationality on the rationality of pro attitudes such as desires and practical judgments. Hence I term the model the Pro Attitude model. This model claims that a decision to perform a particular action is justified by exactly what, according to JUSTIFY, justifies performing the action decided upon - by the desirability of ends which that action would further.3 The Pro Attitude model depends on a particular account of practical deliberation. As Bratman has put it: in deliberation about the future we deliberate about what to do then, not what to intend now, though of course a decision about what to do later leads to an intention now so to act later . . . This means that in such deliberation about the future the desire-belief reasons we are to consider are reasons for various ways we might act later. (Intention, Plans and Practical Reason, p. 103) 2 3

It is a controversial question whether the desirability of ends has to be explained in terms of what ends agents actually desire to attain - a controversy I address elsewhere. Much prominent work on decision and intention assumes the Pro Attitude model. See, for example Anscombe, Intention; Davidson, 'Intending'; and Bratman, Intention, Plans and Practical Reason.

142

The psychology of freedom

On this account, the practical deliberation which moves us to take a decision to perform some action A is, simply, deliberation about whether to do A. The practical deliberation which motivates our second-order agency, then, is just about whether to perform the first-order agency which that second-order agency would elicit. In other words, the subject matter of the deliberation which guides our decision-making is nothing other than what justifications there might be for the actions between which we must decide. But if this is so, then what justifies our deciding to perform one action rather than another - the justifications which we consider in the deliberation leading up to decision - is going to be what justifies our performing that action rather than another. Our justifications for deciding to do A, rather than any other action, are going simply to be our justifications for doing A, rather than any other action. Hence we derive the Pro Attitude model's Identity thesis about decision and action justifications: J-IDENTITY

our justifications at t for then deciding or intending to do A are identical with our justifications at t for later doing A.

Justifications for doing A, as explained by JUSTIFY, are desirable ends which doing A would further. It follows, on the Pro Attitude model, that any justification for deciding to do A is going to be a desirable end which doing A would further. Hence, though doing A is means-end justifiable, deciding to do A is not. Justification for deciding to do A depends on what ends are furthered by doing something else - by doing A, rather than by deciding to do A. The Pro Attitude model says that the deliberation which explains and motivates our decisions to do A is, simply, deliberation about whether to do A. So it claims also that the reasons which move us to take a decision to act are our reasons for performing the action upon which we decide. We are moved to decide to act by the very same attitudes which, at the time for action, would move us to act as decided. Hence the Pro Attitude model proposes another Identity thesis, this time about decision and action reasons: R-IDENTITY

our reasons at t for then deciding or intending to do A are identical with our reasons at t for later doing A.

Reasons for doing A, as explained by REASON, are desires for ends combined with beliefs that doing A would further those ends. It

Decision rationality and action rationality

143

follows, on the Pro Attitude model, that any reasons for deciding to do A are going to be desires for ends combined with beliefs that doing A would further those ends. The claim that a given activity is means-end justifiable is closely related to - though importantly not the same as - the claim that it is purposive. To say that an activity is purposive is, as we have seen, to say that it is motivated by attitudes which constitute means-end related reasons for performing it. An agent who does A purposively is motivated by a desire for an end E and a belief that doing A might further the end E: he is doing A in order that E. Now it is clear that if a given activity is means-end justifiable then that activity may also take purposive form. If the justification for an activity depends on what ends it furthers - if it comes from the fact that the activity is likely to further some desirable ends — then it may surely be done as a means to those ends. Means—end justifiability implies the possibility of purposiveness. But, as we saw at the end of chapter 2, agency can be means-end justifiable without its actually being purposive - without the agent's actually performing it in order to further any end. Our example was the agency of conation - the agency of trying to perform bodily actions. Trying is a form of agency. From the fact that we try to move our hand it follows, whether or not we are successful in doing what we are trying to do, that we have at least done something. Trying is also means-end justifiable. Whether or not we are justified in trying to do A depends on what ends trying to do A would further - and not simply on what ends doing A would further. If trying to do A might well lead to disastrous failure - if we might well fail to do A and incur great loss thereby — then no matter what desirable ends doing A might further, we might still lack sufficient justification for trying to do A. And sometimes trying is purposive as well. The attitudes which motivate me to try to do A may, at least on occasion, be desires for ends and beliefs about how trying to do A would further those ends. I may try to do A because I have deliberated about whether to try and concluded that trying would be worth it. But our trying need not take purposive form - and often it does not. Often the attitudes which motivate us to try to do A may simply be desires for ends and beliefs about how doing A would further those ends. Agents very often try to act other than on the basis of beliefs about the likely

144

The psychology of freedom

effects of their trying - perhaps even without having formed any such beliefs at all. It may well be that the possibility of their trying to do A and failing has not occurred to them - that they have been considering only the possible effects of doing A. It is also intuitive that if an activity may be purposive - if an agent could ever perform the activity as a means to some end - then it is means-end justifiable. After all, if an agent could do A in order that E, then surely the desirability of E and the fact that doing A might further E will in turn provide the agent with some justification for doing A. If no justification at all is provided, then it must be true that, despite the desirability of E, and the fact that doing A would further E, A simply cannot be done in order that E. There is a restriction on the purposes for which A may be done as a means. So the possibility of an activity taking purposive form implies its means—end justifiability; correspondingly, that an activity is not means-end justifiable implies that it cannot take purposive form. On the Pro Attitude model, particular decisions to act are in no case justified by ends which they further, no matter how desirable those ends might be. So, according to the same model, it is correspondingly true that particular decisions to act simply cannot be taken in order to further ends. The Pro Attitude Model of decision rationality, in denying that the taking of particular decisions to act is means-end justifiable, also denies that such decision-making may ever be purposive. Agents may perform actions as means to ends. But, according to the Pro Attitude model, they cannot take decisions to act as means to ends. (b) The Action model

The Pro Attitude model explains our justifications for deciding to do A in terms of our justifications for doing A, as explained by JUSTIFY. But there is an alternative model of decision rationality. This Action model includes decision-making in the range of agency governed by JUSTIFY.4 JUSTIFY says that our justifications for doing 4

Recent work which assumes the Action model includes that of Gauthier, Kavka, and Lewis. But these writers, whom I discuss in chapter 6, propose Teleological theories of decision rationality. And that, as we shall see, makes their theories inconsistent with the claim that decision-making serves to apply reason as it governs subsequent action.

Decision rationality and action rationality

145

A are desirable ends which doing A would further. So the Action model proposes ACTION: ACTION

Any justification at t for then deciding or intending to do A consists in the likelihood at t that so deciding or intending would further a desirable end E.

Just as an action is justified by the desirability of ends which are furthered by performing that action, so too a decision to act is justified by the desirability of ends which are furthered by taking that decision. It follows that, on the Action model, decision-making is as much a means-end justifiable agency as is the action to which it gives rise. Hence, like the action to which it gives rise, decision-making may be purposive. An agent may take a decision as a means to attaining some end. And the practical deliberation which leads to a decision to do A need not be exhausted by deliberation about whether to do A. Once an agent has deliberated about what ends would be furthered by doing A, there may be further deliberation yet to do: she may have to deliberate about what ends would be furthered by now deciding to do A. The deliberation which gives rise to our second-order agency need not be simply about whether to perform the first-order agency which that second-order agency would elicit; it can also be about whether to perform that second-order agency. Practical deliberation need not be exhausted by deliberation about what action to perform. There may be further deliberation about what decision to take or what intention to form. Does the Action model claim that decision-making is always purposive? And need the deliberation which precedes our decisions always include deliberation about what decision to take? Remember what held true of trying. Trying is means-end justifiable: the Action model provides the true model for the rationality of trying. Hence trying may be purposive; and the deliberation which precedes our attempts at action may be deliberation about what attempt to make. But quite compatibly with this, many cases of trying are not actually purposive; many attempts at action are preceded only by deliberation about what action to perform, and not by any deliberation about what attempt to make. So too, surely, for decision-making. Even if decision rationality is as the Action model describes, not all our decisions to act need be taken purposively - as means to ends. The deliberation which

146

The psychology of freedom

precedes our decisions to do A may only be deliberation about whether to do A, and not about whether to decide to do A as well. Indeed, I noted contra Frankfurt, that decision-making characteristically is non-purposive. Much of the practical deliberation which precedes our taking a particular decision to act surely is just about how to act. That is one reason why, whether or not it actually turns out to be true, the Pro Attitude model of decision rationality has such a strong initial plausibility. DECISIONS TO ACT AS REASON-APPLYING ACTIONS

Now we have outlined the Pro Attitude and Action models of decision rationality, which best explains how decisions can be what we ordinarily suppose them to be - reason-applying actions? The common-sense theory of the will contains, remember, the following two claims : - taking a particular decision to act counts as an action itself- as something which one deliberately does, and in doing which one may exercise future action control; and - decisions to act have an executive function: taking a decision about how to act serves to apply reason as it concerns subsequent action. A decision to do A, if rationally taken, leaves one disposed to do A rationally thereafter. And this a decision does by perpetuating its own motivation - by ensuring that the ends which motivated one to decide to do A, and which justified taking that decision, also motivate one to do A thereafter. Decisions serve to apply reason as it concerns subsequent action by having a reason-applying, and so rationality-preserving and motivationperpetuating effect on future action.

And the difficulty, as we shall now see, is how to reconcile the idea that decisions to act are actions themselves, with the idea that decisions have this reason-applying influence on future action. THE ACTION MODEL

Suppose the Action model of decision rationality is true. Any justification for deciding to do A is, therefore, a desirable end E which deciding to do A would further: deciding is means-end justifiable. The first point to be made, then, is that the Action model seems adequately to explain how decision-making might be a genuine case of agency. And so the Action model raises no special

Decision rationality and action rationality

147

problem about how decision-making can be free - a source of future action control. What counts as our agency? An hypothesis which I adopted in chapter 2 is that there is a distinctively practical reason - a distinctive way in which reason governs our practice or agency. If this is right, we can explain agency in terms of practical reason. Any case of motivated X-ing counts as our agency - as something we do - in virtue of the fact that X-ing is governed by reason as a practice. The distinction between the active and the passive, then, is to be explained in terms of the different ways in which the active and the passive are governed by reason — in terms of the different ways in which the active and the passive are justified. Action is a paradigm agency. The actions to which our decisions lead are amongst what counts as our agency, if anything does. How does reason govern action? As we have already seen, action is means-end justifiable. The justification for the actions which we perform depends on what ends those actions would further. And that, we have seen, is true of our first-order agency in general. Our conations or tryings are also means-end justifiable. That is plausibly why we consider tryings as cases of agency, even where they are not actually purposive. So if decision-making too is means-end justifiable, then it very plausibly counts as agency likewise. So if we adopt the Action model of decision rationality, the case for decisions forming part of our agency is going to be powerful — perhaps as powerful as the case for the agency of the actions and the attempts at action which those decisions cause. Difficulties arise, however, when we come to explain how decisions have a reason-applying and so rationality-preserving and motivation-perpetuating effect on subsequent action. The toxin puzzle

As a version of Kavka's toxin puzzle5 will now show, it is not immediately clear how the Action model is consistent with the idea that decision-making is a reason-applying activity. An agent is offered £1 m for deciding now that tomorrow he will drink a very mild toxin. I shall call drinking the mild toxin tomorrow doing A. The prize is won irrespective of whether, once 5

See Kavka's 'The toxin puzzle', Analysis 1983.

148

The psychology of freedom

the decision to do A has been taken, A is actually done. The effects of the mild toxin are disagreeable, but negligibly so in comparison to the desirability of £1 m. Winning the prize is clearly a desirable end which would be furthered by deciding to do A. Equally clearly, it is a desirable end which would not be furthered by actually doing A. The question posed by the toxin puzzle is: does the prize justify deciding to do A? Would a rational agent be moved to decide to do A in order to win the prize? It looks as though the Action model does allow the prize offer to justify taking the decision. But, if so, that creates problems for the claim that decisions are rationalitypreserving and motivation-perpetuating in their effects. It is clear enough why the Action model leaves the prize offer looking a plausible decision justifier. The Action model, after all, says that the taking of a particular decision to act is justified in terms of desirable ends which that decision would further. So will not the fact that winning the prize is a desirable end which deciding to do A would further, constitute sufficient justification for taking that decision? Since the bad effects of the toxin are so negligible then, even if the decision did lead to A's being done, that would hardly detract from the gain of the prize. Why should not a rational agent decide to do A in order to win the prize? Consider a form of first-order agency which is apparently very like that of taking a decision: the action of taking a decision drug. As introduced in chapter 3, a decision drug is a drug which mimics a decision. Taking a decision to do A is an activity which causes us to do A by ensuring that we are left motivated so to act. So too is taking a decision drug - taking the decision drug for a given action A also causes us to do A by ensuring that we are left motivated to do A. And, as with a decision, the causal influence of a decision drug is belief-dependent. The antidote to the drug for doing A is coming to hold beliefs which imply the falsehood or incompleteness of assumptions on which the drug's taking was based - beliefs which, had they been held at the outset, would have moved the agent not to take the decision drug. So, it seems, taking a decision and taking a decision drug are similar forms of action-influencing agency. Each is a doing over which we may exercise control, and which in turn gives us control over future action. Each kind of agency influences action in broadly the same way; in particular, the causal influence of each is subject to the same condition. We may say then that decision-taking and

Decision rationality and action rationality

149

decision drug taking are activities which - at least in the context of the toxin puzzle - further materially the same ends. If we espouse an Action model of decision rationality - if we suppose that, like performing an action, taking a decision is justified by desirable ends which it furthers - we might want to model our view of what would justify taking a decision on what would justify taking a decision drug. After all, the following thought about how practical reason governs means-end justifiable agency is highly intuitive. Take any two means-end justifiable activities - and suppose that the two activities differ in what would justify their performance. Then the difference in justifications must be explicable in terms of a difference in the ends furthered by the two activities. If a form of agency is means-end justifiable then, trivially, its justifications come from ends furthered. So must not differences in justification arise out of differences in ends furthered? Consider a £1 m prize offer just for taking a decision drug for doing A, where A is again the action of drinking a very mild toxin tomorrow. The prize offer is again conditional only on the drug's being taken. It is won irrespective of whether A is ever done. It is highly plausible that such an offer is sufficient justification for taking the drug. In which case, why is it not sufficient justification for taking the decision? For surely practical reason governs all our means-end justifiable agency alike. If two such activities further materially the same ends under the same conditions, why will not a rational agent be led by reason to perform one if, and only if, reason would lead him to perform the other? Suppose we do not discriminate. The prize offer, we maintain, is sufficient justification for taking the decision to do A just as it is sufficient justification for taking the decision drug for doing A. A rational agent will therefore be moved by the prize offer to decide to do A. Can decisions have a reason-applying causal influence?

The decision to do A, if taken, will then, like any such decision, leave the agent motivated to do A. But can the decision have influenced the agent's desires to act and action as the common-sense model of the will demands - in a way which is both motivation-perpetuating and rationality-preserving?

150

The psychology of freedom

Ex hypothesis and unlike deciding to do A, actually doing A does not help the agent win the prize; nor has the agent been given any justification for supposing otherwise. However, were the decision to be motivation-perpetuating, the agent would have to be left desiring to do A because of a belief that doing A, like taking the decision, would help him win the prize. For, according to REASON, only when the action A is motivated by such belief can that action be motivated, as was the prior decision, by a desire to win the prize. Suppose, then, that the decision did leave the agent with the belief that doing A would help him win the prize. Then the influence of the decision could hardly be rationality-preserving: for the decision itself could have given the agent no warrant for such a belief. The decision, if motivation-perpetuating, could not be rationalitypreserving as well. Suppose the prize-motivated decision to do A was not motivationperpetuating - could it then be rationality-preserving alone? We have supposed, for the sake of argument, that the prize-motivated decision was rational. But would the execution of it be rational too? To answer this question, we must consider what would justify doing A as decided. If JUSTIFY is true, doing A could not be justified by the end which justified the decision - by the end of winning the prize. For doing A would not further that end - it would not help win the prize. But perhaps there is some other justification for doing A — some distinct and equally desirable end which doing A might still further. However, we must now ask what end this could possibly be. One end furthered by doing A would be that of executing a prior, rational decision to do A. Perhaps rational decisions to do A make doing A rational, not because doing A would further desirable ends already furthered by the prior decision, but simply because it is desirable to execute rational decisions. In other words, perhaps rational decisions to do A make doing A rational because the execution of rational decisions is desirable in itself. A rational agent's desire in acting as decided could, therefore, be simply to execute a prior and rational decision. But, when in the last chapter I rejected the Additional Motive theory of decision execution, I there argued that decision execution is not intrinsically desirable. If it were, then we should rationally regret our failure to execute a decision, even in cases where no other bad consequences arise. But the biscuit-packet example

Decision rationality and action rationality

151

suggested that when no other bad consequences arise, there is nothing to regret. Notice that we would not accept that the execution of rationally taken decision drugs is intrinsically desirable. As I noted at the end of chapter 3, taking a decision drug, and taking it rationally, does not make it any more desirable to perform the action which the decision drug is designed to produce. Consider again the version of the toxin puzzle in which it is taking a decision drug, rather than taking a decision, which is rewarded. Suppose to win the £1 m prize offered just for taking that decision drug, I quite rationally take the drug for doing A - the drug which will cause me subsequently to do A, to take a toxin, as would a decision. That I take the drug rationally certainly will not do anything to justify doing A thereafter. Why should it become desirable to perform a given action just because performing it is a standard consequence of - the 'execution' of - a rationally taken drug? We assumed that because they seemed to be means-end justifiable activities which furthered relevantly the same ends, we should not discriminate between taking a decision to do A and taking a decision drug for doing A. If practical reason recommended that we take the decision drug so as to win the prize, then it would also recommend that we take the decision so as to win the prize. But having refused to discriminate at the outset, we cannot start discriminating now. The causal influence on action of a prizemotivated decision drug would not be rationality-preserving. So why should the causal influence on action of a prize-motivated decision be rationality-preserving?6 In so far, then, as the Action model leaves it plausible that the 6

Joe Mintoff has suggested to me that the execution of a prize-driven decision would be desirable, as the execution of a prize-driven decision drug would not be, because deciding to act has a distinctive connexion with intention - a connexion which explains why decision execution, and decision execution alone, is intrinsically desirable. To decide to act is to form an intention to act. And it is intention execution which is intrinsically desirable. But there is a problem with this suggestion. As I observed in chapter i, an intention is nothing more than the state of being decided - the state of being motivated to act through some past second-order action. Both decision taking and decision drug taking ensure a persisting motivation to act. It is just that when this motivation results from a decision from a past action which is second rather than first-order-we say that the agent concerned holds an intention. Appealing to intention, then, does not do anything to support the assertion that decision execution is intrinsically desirable as decision drug execution would not be. It is just a round-about way of repeating that assertion - an assertion which I have anyway already suggested is false.

152

The psychology of freedom

toxin puzzle's decision prize should be a decision justifier, so it also threatens our natural belief that decision-making is a rationalitypreserving activity. Suppose a prize offered simply for deciding to do A could justify taking that decision, and so could move a rational person to take it. Then, it seems, the action A which that rational decision left one motivated to perform could perfectly well be irrational. Rational decisions to act could leave one disposed and motivated to perform irrational actions. So if decisions to act are to be rationality-preserving - if, when taken rationally, decisions to act are to leave one disposed to act equally rationally thereafter - there had better be a restriction on what can make taking a decision rational. Decision prizes that do nothing to justify acting as decided had better not justify the decisions which they reward. REASON-APPLY

The rationality of deciding to perform an action A, we naturally assume, guarantees the rationality of doing A thereafter. We never admit the sense of a decision to act, without allowing that performing the action decided upon is sensible too. Rationally taken decisions to act must leave agents disposed to act rationally. Now for that to be true, the rationality of deciding to do A must guarantee the existence of ends sufficient to justify doing A thereafter. And it had better be these ends which, thanks to the decision, also motivate the agent to do A. For, as we have already noted, rational action is action which is motivated by ends which sufficiently justify its performance. So, in any case where a decision to do A has been rationally taken, what end or ends can be relied upon to provide sufficient justification for doing A? Clearly, it cannot be the end of executing a rationally taken decision. For we have found no reason to suppose that the execution of decisions - even the execution of rational decisions - is desirable in itself. Neither the fact of the decision nor its rationality can of themselves make doing A desirable. Justification for doing A, then, can only come from what made the decision rational in the first place - from whatever ends provided sufficient justification for deciding to do A. Once the decision has been taken, these same decision-justifying ends must also provide at least as much justification for doing A.

Decision rationality and action rationality

153

In other words, if the rationality of deciding to perform an action A is to guarantee the rationality of doing A thereafter, there had better not be ends which justify deciding to do A without, ipso facto, providing at least as much justification for doing A thereafter. With respect to any action A, R E A S O N - A P P L Y must be true: REASON-APPLY

any end E that justifies deciding to do A must, supposing that decision is taken, also provide at least as much justification for doing A.

places an important restriction on the ends which can justify taking a decision to act. Ends can justify taking decisions only if they would thereafter justify the actions decided upon. And that means, by J U S T I F Y , that ends can justify taking decisions only if they would thereafter be furthered by the actions decided upon. It follows that a huge cash prize offered just for taking a given decision to act cannot justify taking that decision. And that is because winning the prize could not justify acting as decided thereafter. The prize would be won in any case, whether or not the agent executed his decision. Were REASON-APPLY false, then I see no reason why the rationality of taking a decision should guarantee the rationality of acting as decided. There could perfectly well be ends, such as the winning of huge prizes thereby, sufficient to justify deciding to do A - ends which, therefore, would move a rational agent to take that decision — without there being sufficient justification for doing A thereafter. Compare decision drugs. There is no analogue of R E A S O N - A P P L Y which restricts justifications for decision drug taking. Ends, such as the prospect of winning huge cash prizes thereby, can easily justify taking a decision drug despite the fact that they would not justify thereafter performing the action which the drug motivates. Which is precisely why the rationality of taking a decision drug does not guarantee the rationality of thereafter acting as one's drugged to. Decisions whose rationality was unrestricted by R E A S O N - A P P L Y would be no different — unless the execution of rational decisions were desirable in itself, as the execution of rationally taken decision drugs is not. But that is a possibility which we have already rejected. So if rationally taken decisions to do A are to dispose us to do A rationally thereafter, REASON-APPLY had better be true. And not only that, but decisions to act had better be motivation-perpetuating as well. As we have already agreed, the rational action which a REASON-APPLY

154

The psychology of freedom

rational decision moves us to perform must be motivated by the ends which justify its performance. Now if REASON-APPLY is true, these action-justifying ends are going to be the very same ends which justified the initial decision to act. Rationality-preserving decisions had better ensure, then, that the ends which motivated taking them, are also the ends which motivate performance of the actions decided upon. I have argued that REASON-APPLY must be true from the claim that decisions to act are rationality-preserving in their effects - that their rationality guarantees the rationality of acting as decided. If REASON-APPLY were false, then rational decisions to act could leave us disposed to act irrationally. But suppose I were wrong on that point - perhaps because, contrary to what I have argued, the execution of rational decisions really was intrinsically desirable. Then, I claim, that still would not diminish the plausibility of REASON-APPLY. For REASON-APPLY

is, surely, an immediate consequence of the idea that the will serves an executive function in relation to our action - that decision-making is about applying reason as it governs our subsequent action. For if that is what decision-making is about, how could there be counterexamples to REASON-APPLY? — how could anything make it rational to decide to do A which wasn't, by that very fact, a justification for doing A thereafter? Decisions and decision commands

That REASON-APPLY is true would explain something very important about decision-making that O'Shaughnessy has noticed. In the course of arguing that taking a decision to act is not an action itself, O'Shaughnessy has observed, surely rightly, that it does not make sense to order someone to take a particular decision: In short: there is no activity of'deciding to do $', where O is an act; even though there is an event of deciding-to-^. This receives confirmation in the fact that there is no order: 'Decide to raise your arm.' (The Will, vol. 2, p. 300)

Now whether or not O'Shaughnessy is right in his inference to the non-agency of decision-making - an issue to which we shall shortly return - he must surely be right that 'Decide to raise your arm' is not a sensible command. And that distinguishes decision-making

Decision rationality and action rationality

155

from our first-order agency. There is nothing weird about the commands 'Raise your arm 5 or 'Try to raise your arm'. But there is something thoroughly weird about 'Decide to raise your arm' or 'Intend to raise your arm'. And that is easily explained by the fact that decision-making has the function of applying reason as it governs subsequent action - so that decision-making is itself governed by REASON-APPLY.

Commanding someone to X makes sense only where such a command could, in principle, justify that person's X-ing. That X-ing would comply with the command must be the sort of consideration which - in the absence of some overriding justification for not X-ing, and if there were enough authority in the commander would justify X-ing. Now when I command you simply to take a decision, you comply with my command simply by taking that decision, irrespective of how thereafter you act. But then, even in the absence of any overriding justification for not taking the decision commanded, and no matter how weighty my authority, the desirability of complying with my command could not possibly justify taking the decision. For my command, considered as a decision justifier, violates REASONAPPLY. It violates REASON-APPLY because once the decision is taken, the desirability of complying with my command could not do anything to justify thereafter acting as decided. The reason, then, that there is no sensible command 'Decide to raise your arm', is simply that, as a decision justifier, such a command would be inconsistent with the function, which the will has, of applying reason as it governs our subsequent action. The command parades as a decision justifier; but no matter what authority might be behind the command, it can do nothing to justify performing the action decided upon. And that is exactly why, as O'Shaughnessy puts it, there is no such command. The Action model and REASON-APPLY

If the Action model is to be consistent with the idea that decisionmaking is a reason-applying activity - that decision-making serves to apply reason as it governs subsequent action - then the model must be supplemented by REASON-APPLY. The Action model claims that any justification for deciding to do A consists in the likelihood that so deciding would further some desirable end. But if REASON-

156

The psychology of freedom

is true too, then not all desirable ends which deciding to do A would further can provide any justification for so deciding. Justification for deciding to do A can only come from those ends which, once the decision was taken, would provide at least as much justification for doing A thereafter. In other words, if decisions to act are means-end justifiable, as the Action model claims them to be, there must be a special restriction on the ends which might justify taking them. Decision-making, if means-end justifiable at all, must be especial-purpose action. That is, not every desirable end which a given decision to act would further can provide any justification for taking that decision. Justification for the decision can only come from those ends which satisfy REASON-APPLY — which would justify acting as decided thereafter. And the ends which can motivate taking a decision must correspondingly be limited. Any possible motive for taking a decision must ipsofacto be a motive for acting as decided thereafter. So prizes offered simply for deciding to do A, no matter how large they might be, cannot provide any justification whatsoever for taking that decision, or motivate us to take it. And that is because those prizes cannot do anything to justify or motivate doing A thereafter. If the Action model is true, decision-taking and decision drug taking are both going to be means-end justifiable actions which further similar ends. But decision drug taking is certainly not a special-purpose action as decision-making is supposed to be. A £1 m prize offered just for taking a decision drug will certainly provide some justification for taking the drug. As we have seen, there is no analogue of REASON-APPLY restricting the ends which can justify decision drug taking. So, if we are to supplement the Action model with REASON-APPLY, we had better provide some explanation why, despite the fact that both are very similar means-end justifiable actions, there should be a constraint on what ends can justify decision-making that does not similarly apply to decision drug taking. As yet we lack such an explanation. And that leaves a certain tension between the Action model and REASON-APPLY. If both decision drug taking and decision-taking are actions which are justified in terms of the ends which they further, and if both actions further similar ends, prima facie there should be no difference between the rationality of the two actions. So if there is a difference because REASON-APPLY is true, might not APPLY

Decision rationality and action rationality

157

that be because, unlike the taking of a decision drug, the taking of a decision simply is not justified by reference to the ends which it furthers? Might not the Action model of decision rationality be false? And that thought takes us back to the Pro Attitude model. THE PRO ATTITUDE MODEL

The toxin puzzle pits our intuition that deciding to act is itself an action, just as taking a decision drug is an action, against our intuition that decision-making has an executive function - that deciding how to act is about applying reason as it concerns our subsequent action. As I suggested in chapter 2, we conceive agency to be that which is governed by reason in a distinctively practical form. And the most obvious candidate for reason in its distinctively practical form is means-end justifiability - which is, after all, the way in which reason governs our first-order action generally. The thought that deciding to act is an action itself leads us, therefore, to see deciding to act as means-end justifiable too. And that then strongly inclines us towards seeing the prize offer as a potential decision justifier. For an analogous prize offer could always provide some justification for performing a first-order action. There may be some story about why, though it too is means-end justifiable, the second-order action of our decision-making should be different in this respect. But it has yet to be told. On the other hand, admitting that the prize offer is a possible decision justifier is also counter-intuitive. And that is because, besides conceiving of decisions as actions, we have an equally strong conviction that decision-making serves an executive function - that the point of taking decisions about which actions to perform just is to apply reason as it concerns our subsequent action. The prize offer, if a decision justifier at all, would be a decision justifier that didn't do anything to justify acting as decided. But if decisionmaking is all about applying reason as it concerns our subsequent action, there simply cannot be decision justifiers like that. So it is not essential to the toxin puzzle, that the action to be decided upon - taking the toxin - is deeply irrational. The puzzle arises from the mere fact that the prize offer - the alleged decision justifier - does nothing to justify acting as decided. Given that the will's function is executive - that the point of decision-making is to

158

The psychology of freedom

apply reason as it governs our subsequent action - there is an immediate problem about admitting any such decision prize as a decision justifier or motivator. The problem arises whether or not the action to be decided upon is irrational as well. This point is reinforced if we turn to O'Shaughnessy's decisioncommands, which give rise to the same conflict of intuitions as do decision prizes. It makes perfect sense to command people to perform first-order actions. I could quite sensibly order you to take a decision drug, were there one available to be taken. So - it is very natural to think - if decisions to act are means-end justifiable actions too, it should make equal sense to order people just to take decisions as well. But commands just to take decisions to act do not make sense — and they fail to make sense whether the action decided upon is rational or irrational. 'Decide to raise your hand' or 'Intend to raise your hand' are weird commands, just as is 'Decide to drink a toxin tomorrow'. Commanding someone simply to take a decision does not make sense - and it does not make sense whatever the action to be decided upon. And that is because such commands to decide can do nothing to justify acting as decided. As pretended decision justifiers, O'Shaughnessy's decision-commands, like Kavka's decision prizes, violate REASON-APPLY. AS pretended decision justifiers, they are inconsistent with the reason-applying function of the will. Now the Pro Attitude model does appear to do justice to our conviction that decision-making is a reason-applying activity - that the point of decision-making is to ensure the rationality of our subsequent action. And it is in precisely this fact that the great plausibility of the Pro Attitude model lies. There is no question but that the Pro Attitude model rules decision prizes and decision commands out as decision justifiers. And that is because the Pro Attitude model explains justifications for deciding to do A directly in terms of justifications for doing A - in terms of desirable ends which doing A would further. Winning decision prizes or obeying decision commands are not desirable ends which doing A would further. Hence, the desirability of winning decision prizes or obeying decision commands thereby cannot be any justification for deciding to do A. The Pro Attitude model says that our justifications for deciding to do A are, simply, our justifications for doing A. In saying this, the model admirably fits our intuition that decision-making is indeed

Decision rationality and action rationality

159

about applying reason as it concerns subsequent action. For if that is the function of decision-making, what else should determine the rationality of deciding to perform an action than the rationality of the action decided upon? REASON-APPLY says that ends which justify decisions to do A must justify doing A thereafter. The Pro Attitude model says that the ends which justify our taking a decision to do A just are those which justify later doing A. So does not the Pro Attitude model restrict possible decision justifications in line with REASON-APPLY? So the need to explain how decision-making can be reasonapplying in relation to subsequent action may well incline us, at least initially, towards the Pro Attitude model. But, of course, we then come up against our intuition that decisions to act are actions themselves. And on this score the Pro Attitude model does much less well. A problem for the agency of decision

If the Pro Attitude model is true, decisions are not means-end justifiable. That means that decisions cannot be purposive. The attitudes which may motivate and explain our decisions to do A must be purposes for doing A; they cannot be purposes for deciding to do A. But can there be any agency which is not means-end justifiable, and which we could not ever perform as a means to some end? The Means-End model of agency rationality, which implies that all agency is means-end justifiable, is highly plausible. And that is because means-end justification is so plausibly reason in its distinctively practical, agency-governing form. As we saw in chapter 2, it is essential to, and distinctive of, agency that it should always provide the possibility of exercising control. So the form in which reason governs agency must be appropriate to that fact. Reason in its practical form must be such as would govern any exercise of control. And means—end justifiability does seem to be rationality in the appropriate form. For what does the rational exercise of control involve, but producing outcomes which are desirable, and preventing outcomes that are undesirable? And means-end justifiable doings, appropriately enough, are made rational precisely by the fact that they would produce desirable outcomes and prevent undesirable ones.

160

The psychology of freedom

Many philosophers have made inferences from the active nature of an occurrence which, whether or not they are valid, certainly presuppose the means-end justifiability of agency. Consider belief. Philosophers often dispute about whether belief is ever a form of agency within our control - whether we can ever control what beliefs we form as we control what actions we perform. And they typically put the issue in terms which presuppose the means-end justifiability of agency. Sometimes philosophers ask, as does Bernard Williams (see his 'Deciding to believe'), whether we can ever decide to believe that p, and form the belief just on the basis of this decision. This way of putting the issue suggests that anything which counts as agency must be subject to the will: it must be something which, at least in principle, we could do or not simply on the basis of deciding whether to do it. But if something is, or may be subject to, the will - if an agent can do it simply on the basis of deciding to do it - then surely it can be done as a means to an end, and is justified by ends which doing it would further. Indeed, I shall later be arguing that subjection to the will has precisely this implication. Jonathan Bennett (see his 'Why is belief involuntary?') puts the issue differently. When asking whether we ever control our beliefs, he makes the issue rest on whether belief can be prize-driven. He asks whether we can ever come to believe that p on the basis of a desire to win a prize, and a belief that believing that p would win us that prize. This way of putting the issue suggests that anything which counts as an agency must be a general-purpose agency: if there is any desirable end, such as winning a prize, which it would further, then it must be performable - and so rationally performable - as a means to that end. And this again implies that its performance is means-end justifiable. Finally, O'Shaughnessy, we saw, assumes that if decisions to act are actions themselves, then they must be commandable. But commands, as we have seen, are appropriate only if it is possible, at least in principle, for the command to justify what is commanded. But then what is commanded must be justifiable by the fact that it would comply with a command, i.e. by the fact that its occurrence would further that particular end. Whatever is commandable must be means-end justifiable also. So O'Shaughnessy's view again implies that agency must be means-end justifiable. So Williams, Bennett, and O'Shaughnessy are all committed to

Decision rationality and action rationality

161

an inference from something's counting as an agency to its being means-end justifiable. The intuitiveness of this inference raises a question about the Pro Attitude model of decision rationality: is it really consistent with our common-sense conception of decisionmaking as agency - as something we do, over which we may exercise control as we exercise control over action? Even if the inference from agency to means-end justifiability is correct, it is of course not obvious that Williams', Bennett's, and O'Shaughnessy's other assumptions are correct. It is not at all clear that, even in principle, all agency need be subject to the will. For we suppose that taking a particular decision to act is itself a form of agency - it is something that we deliberately do. But I have suggested that taking a particular decision to act is not something which is subject to the will - it is not something which we can do on the basis of a prior decision to do it. And as regards Bennett's and O'Shaughnessy's understanding of agency, we have already seen that decision-making, if means-end justifiable, must be a specialpurpose agency. There must be a special restriction - that placed by REASON-APPLY - on what ends might be relevant to a decision's justification or motivate its being taken. If there is any hope at all of making sense of the common-sense psychology of decisionmaking, it had better not be a condition on something's counting as agency, that it could be motivated by prizes for doing it or by commands that it be done. The popularity of the Means-End model of agency rationality is easy to understand. So what can a defender of the Pro Attitude model say to defend the agency of decision-making when she denies its means-end justifiability? She must obviously find some alternative to the Means-End model of agency. Now our general hypothesis is that reason governs our agency in a distinctive way: there is a distinctively practical reason. So a supporter of the Pro Attitude model must find some alternative account of how, in general, reason governs our agency - of how reason takes a distinctively practical form. There must be some way in which reason governs agency which is plausibly distinctive of agency, but which does not necessarily involve justification of the agency by ends which the agency furthers. Practical reason must in some way unite the means-end justifiable agency of our first-order action with what the Pro Attitude theorist holds to be the nonmeans-end justifiable agency of our decision-making.

162

The psychology of freedom

The Means-End model claims that any form of agency is justified by reference to the ends which it furthers. Since any means-end justifiable agency can be performed purposively, and any purposive agency is means-end justifiable, this conception of agency rationality is linked to a conception of the agent as a purposive pursuer of ends. But other models of practical reason suggest themselves, also. These alternative models generalise other aspects of reason as it governs first-order action. For example, we may on occasion have equal justification for performing each one of a range of alternative actions, so that it is rationally up to us which we do. In the supermarket, I can have equal justification for picking out any one of a range of biscuit packets; it does not matter which I pick out. Action seems, potentially at least, to be underdetermined by reason. It is always possible that reason may leave it up to the rational agent what she shall do. We have then the Autonomy-within-Reason model of practical reason. Reason governs an agent's X-ing practically where there is the general possibility that reason should leave up to the rational agent whether or not she Xs. This conception of agency rationality is linked to a conception of an agent as, when active, potentially autonomous within the demands made by reason. If practical judgments and desires are passive on this model of agency - perhaps they are, though I do not wish to settle the question here - that is going to be because what practical judgments and desires it is rational for us to hold is entirely determined by the desirability of the objects of those practical judgments and desires. If two options are equally desirable, for example, it is not rationally open to us to prefer one or judge it more desirable than the other. The rational agent must be indifferent between them, or judge them equally desirable. Whereas, on this model of agency, intention formation counts as active, precisely because in this sort of case it is rationally left open to us which option we intend. One can think of yet further models of how reason is distinctively practical. What of the Plural-Justification model of agency rationality? Our action seems to be governed by a range of irreducibly distinct kinds of justification. There is not one single end in terms of which a given action A must be justified if it ever is so. The performance of A may be justified by a plurality of agency-justifying desirable ends. To do A might be just, or it might be pleasurable or elegant.

Decision rationality and action rationality

163

We have no conception of a single desirable end - the good or the desirable in general - that is prior to our conception of these distinct ends, and to which all these distinct ends count as means. The Plural-Justification model claims that it is this which is distinctive of agency rationality. This conception of agency rationality is linked to a conception of an agent as a deliberator between ends, and not merely a mechanical means-end deliberator. On this view, intention counts as active, just as action does. For the same variety of ends that justify actions can justify the intentions to perform them. What of desires and practical judgments? Here we might distinguish one's overall desire for an object, or one's judgment about how desirable an object is overall, from judgments and desires for the object that are responses to one particular value - such as a judgment that an object is desirable in so far as it is pleasurable, or a desire felt for an object in so far as it is believed pleasurable. On this model, it is the overall practical judgments and desires that count as active, for these are justified by reference to the desirability of a variety of distinct ends. Whereas the remaining desires and practical judgments count as passive. Each model of agency rationality looks as though it will draw the boundary between the active and the passive in a slightly different way. Each model of agency rationality of course includes first-order action in the active, as any such model should. But each leaves the occurrence of second-order actions a possibility too - a possibility which is consistent with the Pro Attitude model of decision rationality. CONCLUSION

Our common-sense theory of the will has it that taking a particular decision to act is an action itself. And that encourages us to place decisions to act in a class to which the actions which they explain clearly belong - the class of means-end justifiable occurrences. And so our intuitions lead us to the Action model of decision rationality. On the other hand our common-sense theory of the will also has it that decision-making has an executive, reason-applying function the function of applying reason as it concerns subsequent action. And so we incline instead to the Pro Attitude model of decision rationality.

164

The psychology of freedom

So can decisions really be what we normally suppose them to be - a special kind of reason-applying action? Is there such a thing as a reason-applying agency of the will? We have seen what we would have to do to defend the commonsense psychology of the will within the framework of the Action model. We would have to explain why decision-making should be what I have called a special-purpose action. We would have to explain why, though means-end justifiable like first-order actions, some desirable ends which decisions further should be wholly irrelevant to their justification. A proponent of the Action model had better be able to tell us where this restriction on decision justifications comes from, given that there appears to be no analogous restriction on justifications for our first-order action. We have also seen what we would have to do to defend the common-sense psychology of the will within the framework of the Pro Attitude model. The Pro Attitude model would have to be accompanied by a revision of our theory of agency. Agency would still be understood in terms of its distinctive rationality - as that which is governed by reason in a distinctively practical way. But we would have to develop a quite different theory of what it is for reason to be practical. Reason in its practical form would no longer consist in means-end justifiability, but in some other form of justification. At least on the basis of the argument so far, a concern to explain our freedom need not, of itself, push us in the direction of either the Pro Attitude or the Action models in particular. For, as I have claimed, to explain the connexions between decision-making and freedom, we need to defend both the agency of decision-making, and its reason-applying role in relation to the action which it explains. Otherwise we lose our explanation of why Survival and Dependence are true. Dependence, we have seen, is explained in terms of the requirement that free agents exercise future action control in a reason-applying way. Decision control is supposed to be what gives us that reason-applying future action control: hence action control depends on our having decision control. Survival is explained by the fact that the exercise of a faculty for reasonapplying future action control can never be freedom depriving. An adequate theory of freedom requires, therefore, that we do equal justice both to the status of decision-making as a form of

Decision rationality and action rationality

165

action, and to the reason-applying and plan-executory nature of its effects. I shall eventually be defending our common sense theory of the will within the framework of the Action model of decision rationality. The Action model of decision rationality, I shall be arguing, is the true model of decision rationality. And that is because the Action model actually provides a better account than its rival of the executive, reason-applying function of the will. It is the Action model, not the Pro Attitude model, which does justice to the way decisions apply reason as it concerns our subsequent action. My argument for the Action model will, in particular, appeal to a general theory of practical reason - a theory which provides a principled explanation of why the Action model should be supplemented by REASON-APPLY. My argument for the Action model, therefore, will be entirely consistent with our natural belief that decision-making is reason-applying in function. But first we should examine some other versions of the Action model in the existing philosophical literature. These versions of the Action model, devised by Lewis, Gauthier, and Kavka, are very different from my own. These theories, in my view, have effectively abandoned half the common-sense psychology of the will. They have given up the idea that taking decisions about how to act has a special, reason-applying function - the function of ensuring the rationality of the first-order action which it explains.

CHAPTER 6

Decision-making and Teleology

AGENCY AND TELEOLOGY

Until recently most accounts of decision rationality in Anglo-Saxon philosophy assumed the Pro Attitude model. Justifications for decisions were explained in terms of justifications for acting as decided. It is easy to understand why. We conceive of decisionmaking as having a reason-applying, executive function. And, as we have seen, it seems to be the Pro Attitude model which is consistent with this conception of the will. It is true that decision-making, as we conceive it, also counts as a form of agency - as a second-order action - and that this active status is not obviously consistent with a Pro Attitude model of decision rationality. But, at least prior to the work of Harry Frankfurt, comparatively few Anglo-Saxon accounts of the will were much concerned with the active nature of decisions - with our freedom of will. Many theories of the will assumed what in chapter i I termed an Enlightenment psychology - a psychology in which the occurrence of second-order agency is ignored or denied. To the extent that such theories took any view of freedom at all, the view taken was broadly Hobbesian: we had freedom of action, but no freedom of will. There is a further explanation for neglect of the Action model of decision rationality. This is the popularity of Teleological theories of agency rationality. A Teleological theory explains the rationality of agency in terms of its increasing the amount of good. According to a Teleological theory, an action is rational simply in virtue of the amount of good it is likely to produce - in virtue of its maximising, or at least sufficiently increasing, expected good. But the argument so far suggests that a Teleological model of agency rationality cannot apply to the will. For the agency of the will has got to be reason-applying in relation to subsequent action. 166

Decision-making and Teleology

167

That means that decision justifications must satisfy REASONAPPLY. Not all desirable ends which decisions might further are eligible to be decision justifiers, but only those which would justify acting as decided thereafter. Decision-making, if it is means-end justifiable at all, must be a special-purpose agency. So a given decision to act may be likely to increase good - it may further desirable ends such as bringing its taker great rewards - without that fact providing any justification whatsoever for taking the decision. Yet, on a Teleological conception, any such likely increase in good must count towards the justification we have for performing the agency. Consider for example any standard form of Probabilistic Decision Theory -PD T -understood as a theory of rationality. PDT in its standard form is a straightforwardly Teleological theory of agency rationality. It supposes that any rational agency to which it applies maximises expected good. In PDT, the ends which an action A might possibly further are reported by propositions describing 'outcomes' - outcomes being the various alternative ways the world might be if A is done. The desirability of ends consists in their goodness - a goodness which is measured by numerical utilities attaching to the outcomes in which those ends are attained. The expected goodness of doing A is then measured by an expected utility - a utility defined as a function of the utilities attaching to the possible outcomes of doing A, taken together with the probabilities of those outcomes arising if A is done. More precisely, the expected utility of doing A is a weighted average of the utilities attaching to the possible outcomes of doing A, the weights being the varying probabilities of each outcome arising if A is done. The higher the probability of a given outcome O arising if A is done, the more closely A's expected utility will approximate to the utility of O. Finally, action rationality is defined in terms of expected utility. It is rational to do A if and only if A's expected utility is at least as high as that attaching to any alternative action.1 So where an action A furthers a desirable end E, then, in decisiontheoretic terms, that fact is bound to provide some justification for doing A. For E's attainment will be reported by outcome I ignore many disputes between rival versions of PDT over, for example, the precise nature of the function from outcome utilities to act utilities.

168

The psychology of freedom

propositions on which doing A confers higher probability than would performing other activities; and because these outcome propositions report the attainment of a desirable end, their utility will be the higher also. The net effect will be to add to A's expected utility relative to the expected utility attaching to other activities, and so contribute towards justifying doing A rather than any other agency. Hence PDT does not allow for the possibility that some ends furthered by an action will not help determine its rationality at all. Outcomes and their utilities always contribute to determining the rationality of any action that makes those outcomes more probable. The notion of special-purpose action is alien to PDT. So suppose Teleology provides a quite general model of agency rationality - a model which applies to all our agency, of whatever order. Then it is impossible to conceive of decision-making as both reason-applying in relation to subsequent action, and as a form of agency itself. If decision-making does serve an executive function in relation to the action which it explains - which is indeed what we naturally suppose - and if Teleology is true of agency in general, then decisions just cannot be actions too. Historically, therefore, decision theorists found it convenient to ignore or deny the second-order agency of our decision-making altogether, and concentrate instead on the more Teleologically tractable first-order agency of our action. Decision theory has, until recently, better deserved the title of Action theory. First-order action has been the only agency which it countenanced. Decisions and intentions were swallowed up into the generality of an agent's utility-measured desires. For example, the founding text of the subject, Ramsey's 'Truth and probability' discusses only how two kinds of numerically measured attitude — beliefs and desires — combine to guide action. Decisions to act are not distinguished as a yet further kind of action-explanatory occurrence, let alone included as actions themselves. This failure to make explicit mention of decisions and the intentions they form was based partly on theoretical economy. But it was surely encouraged also by PDT's Teleology. Recently, however, the second-order agency of the will has been rediscovered by decision theorists. Predictably, the rediscovery has taken Teleological form. Rational decision-making is now presented as a goodness- or utility-maximising action itself. PDT has accommodated second-order action by going second-order

Decision-making and Teleology

169

itself. Gauthier, Kavka, and Lewis have all proposed versions of the Action model of decision rationality which take Teleological form. Second-order PDT is supposed to have much to teach us. A rational foundation is provided for altruism, according to Gauthier; a new paradox about rationality - the 'paradox of deterrence' - has been revealed, according to Kavka. But, pace Gauthier, there is no new foundation for altruism; Gauthier's arguments rest only on an ad hoc distortion of our common-sense conception of the will. And, pace Kavka, there is no paradox of deterrence. There is only the absurdity which follows from denying that decision-making is a reason-applying activity - an activity which serves to apply reason as it governs our subsequent action. In what follows, we shall first consider Gauthier's original Teleological theory of decision rationality - an account of rational decision-making as a utility-maximising agency. On the basis of this theory, Gauthier argued that decisions may rationally be used as means to deterring attacks: decisions may be used as means in deterrent strategies. Gauthier's theory of decision rationality combines a Teleological account of decision rationality with our ordinary assumption that decisions are rationality-preserving in their effects. If sound, this theory would provide a defence to the self-interested of the rationality of at least some altruistic actions. Gauthier's theory would explain, in terms that appeal to selfinterest, the rationality of at least some actions by which agents sacrifice their self-interest for the sake of others. Secondly, we shall consider the views of Kavka and Lewis, views which share Gauthier's Teleological conception of decision rationality but which - I shall suggest - are more consistent in their revision of our ordinary conception of the will. Kavka and Lewis have abandoned altogether the doctrine that decision-making is reason-applying in its effects on subsequent action. They deny that decision-making is even a rationality-preserving activity. Finally we shall consider Gauthier's most recent Teleological theory of decision rationality. This theory is importantly different from his original theory. If defensible, it might allow us to combine a Teleological theory of decision rationality with the doctrine that decisions are reason-applying in their effects. But it is not defensible. If there is any truth to the common-sense psychology of the will, Teleological theories of decision rationality are false. Second-order PDT has nothing to tell us about decision rationality.

170

The psychology of freedom MAXIMISING DECISIONS

In Gauthier's 'Deterrence, maximisation and rationality', we find a Teleological version of the Action model of decision rationality. Decision-making and intention formation is means-end justifiable; and it is justified by its maximising expected good. Gauthier used this theory to explain how conditional decisions might rationally be used to deter. We often take conditional decisions. Instead of deciding nonconditionally to do A, we decide to do A if C. We do this when we do not yet feel sure or believe outright that C, but wish to settle in advance how we shall act if C proves to be true. Now, on my theory of the will, a conditional decision must be as much a reasonapplying activity as any non-conditional decision. A conditional decision must, like any non-conditional decision, be plan-executory in its effects on motivation and action. It must ensure that the agent is motivated to do A on coming to believe C - and cause such disposition and action in a reason-applying or plan-executory way. So the effects of the decision on motivation and action must be belief dependent. The decision must ensure that the agent would do A if he came to believe C - but only provided that the agent didn't abandon the beliefs on which the decision to do A if C was originally based. The decision must also be rationality-preserving in its effects on motivation and action. If it is rational for the agent to decide to do A if C, then taking that decision must leave the agent disposed to act rationally thereafter. It must thereafter be rational to do A if C. Finally, the decision must be motivation-perpetuating. The agent's motives for deciding to do A if C must also be what would motivate him to do A were he to come to believe C. It is clear enough how the Pro Attitude model applies to conditional decisions. According to the Pro Attitude model, any justification for a decision to do A if C is a desirable end which doing A would further if C were the case. It is equally clear how the Action model applies. According to the Action model, any justification for a decision to do A if C is a desirable end which deciding to do A if C would further. Now, amongst the ends which deciding to do A if C might further and which might be thought, by supporters of the Action model, to justify that decision, is a deterrent end - the end of preventing C.

Decision-making and Teleology

171

Suppose that your doing A if C is your retaliating if I attack. Then one of the highly desirable effects of your deciding to do A if C may be preventing C - preventing my attack. For your deciding to do A if C will leave you intending - and so disposed - to do A if C. And this disposition may well have a deterrent effect on me. Suppose that A involves some vile and deadly retaliation. Then it may well be that I shall avoid attacking you if I come to believe that you have that disposition. Indeed, it might be that nothing short of your really possessing such a disposition would prevent me attacking you. Threats to do A if C unaccompanied by the disposition would be identified as such by me and ignored. Preventing C's truth is obviously not a justification for your deciding to do A if C which the Pro Attitude model admits. Preventing C's truth is certainly not an end which would be furthered by your doing A when C. Indeed, a supporter of the Pro Attitude model may well deny that there is any justification for the retaliatory decision. Suppose the consequences of your doing A — of retaliating — would destroy both attacker and defender alike. It is plausible that there would then be no justification for doing A when C. A supporter of the Pro Attitude model would therefore deny that your deciding to do A if C was justified at all. But suppose that it is certain, or near certain, that your taking that decision, and only your taking that decision, would further the desirable end of preventing C's truth - of deterring my attack. Gauthier would argue that, in such a case, the desirability of preventing C's truth could justify your deciding to do A if C. The decision would almost certainly not produce the disastrous retaliation decided upon; and it would be your only means of saving yourself from enslavement or worse. Relative to the alternatives deciding not to do A if C or remaining undecided - your deciding to do A if C could maximise expected utility. In which case, according to Gauthier, you, as a rational agent, will decide to do A if C: It may be utility maximising to form . . . the retaliatory intention; therefore it may be rational to form the intention; ('Deterrence, maximisation and rationality', p. 304) This claim about why the decision is justified very obviously assumes the Action model of decision rationality. But, of course, more than the Action model is assumed by Gauthier's argument. As

172

The psychology of freedom

I warned, we have a Second-Order PDT: a Teleological conception of agency rationality has been applied to the agency of the will. Let us say that a deterrent decision is a conditional decision to do A if C which is motivated by a desire to deter - to prevent C. So the term 'deterrent decision' refers not to the decision's deterrent effects, but instead to the agent's motivation for the decision. Gauthier's doctrine is that there can be deterrent decisions, and that, when they occur, they may be rational. Gauthier also argues that given the rationality of the retaliatory decision, executing the decision were an attack to occur would be rational too: It may be utility maximising to form . . . the retaliatory intention; therefore it may be rational to form the intention; if it is rational to form the intention, it is rational to act on the intention. ('Deterrence, maximisation and rationality', p. 304) So deterrent decisions, when rational, induce dispositions to execute them which are correspondingly rational; and should the agent thereafter act as he is rationally disposed to act - should C prove to be true, and the agent do retaliatory A - then that retaliatory action will be rational too. Are deterrent decisions possible at all? If, like decision-making generally, conditional decision-making is a reason-applying agency, deterrent decisions must be impossible. First, a deterrent decision cannot be motivation-perpetuating in its effects on action. The motive for the decision to do retaliatory A - the prevention of C's truth - cannot be the motive for doing A once C is believed true. Preventing C's truth is not an end furthered by doing A when C is true - and no agent would suppose otherwise. Not only can a deterrent decision's effect on action not be motivation-perpetuating. Its effect on action cannot be beliefdependent either. The belief which motivates a deterrent decision to do A if C is belief in the decision's deterrent efficacy: it is belief that if the decision were taken, then this would prevent C's truth. Now the deterrent decision must leave the agent disposed to do A on coming to believe that C is true. But on coming to believe C - on coming to believe that an attack had occurred - any agent would at once come to disbelieve in his decision's deterrent efficacy. Now this disbelief, if held initially, would certainly have moved the agent not to take the decision. So a deterrent decision must motivate its taker

Decision-making and Teleology

173

to act as decided even after coming to hold a belief that - according to our belief-dependence condition - should counteract its influence. It is quite certain that no other disposition would ever deter an attacker. Therefore the influence of a deterrent decision on motivation and action cannot be belief-dependent. In this respect a deterrent decision is even further from being reasonapplying in its effects than is a decision drug. For decision drugs, as I have introduced them, at least influence subsequent action in a belief-dependent way. A decision drug's influence is counteracted by the drug taker's coming to hold beliefs which, if held initially, would have moved him not to take the drug. On Gauthier's Teleological conception of decision rationality, deterrent decisions are supposed to be justified by their likely deterrent effects. Deterrent decisions may be rational because of those likely effects. But deterrent decisions must dispose the agent to perform an action even under circumstances in which the rationale for the original decision to act - evidence for and belief in the decision's deterrent effects - has been entirely removed. It is mysterious why the rationality of the decision should guarantee the rationality of such a disposition, or of the retaliatory action to which that disposition would lead. Gauthier has taken the common-sense conception of decisionmaking as a plan-executory agency, and gutted that conception, preserving only its shell. He has preserved our ordinary belief that decision-making counts as agency - but within a Teleological theory of its rationality. In order to accommodate this Teleological theory, Gauthier has abandoned our ordinary beliefs in beliefdependence and motivation-perpetuation. But he has arbitrarily retained and traded on our ordinary belief in rationality-preservation. We ordinarily suppose that rational decisions induce rational dispositions to act as decided. Gauthier wants us to continue supposing that. We ordinarily suppose that provided certain conditions are met - provided the beliefs and justifications for belief which base our decisions are not lost - rational decisions lead to rational decision executions. Gauthier wants us to continue supposing this too. Not only that, but he even wants us to infer from the rationality of a decision to the rationality of its execution in cases where these conditions are not met - in cases where the decision execution can only occur given loss of justification for the beliefs which based the decision.

174

The psychology of freedom

Why is Gauthier so anxious to insist that utility-maximising decisions guarantee the rationality of the action decided upon? The answer lies in moral theory. Gauthier needs this guarantee to defend to the self-interested the rationality of altruism.2 Let doing A be an act which sacrifices the agent's self-interest and let C be any circumstance in which doing A would benefit others. Let me emphasise that doing A does sacrifice the agent's interest; A does leave the agent a net loser. So A certainly is not an action from which the agent is a loser in the short term only - from which in the long term he gains. Doing A might therefore include respecting others' property when the agent could take and keep it quite undetected. The agent is bound to be a net loser from showing the property of others such respect. How can the rationality of performing such acts A when C be defended to the self-interested? It may be utility-maximising, in self-interested terms, for an agent to decide to do altruistic A whenever C. This is because of the altruistic disposition to do A if C which such a decision induces. Others will detect that the agent has this altruistic disposition, and reward the agent for it with their help and co-operation. The agent may rationally expect to gain more from the co-operation of others than he will lose when, from time to time, C proves to be true and his disposition leads him to do A. Let us suppose that taking the altruistic decision does maximise expected utility. Gauthier's Teleological theory of decision rationality will again allow him to infer to the rationality of the altruistic decision - and then from the rationality of the altruistic decision to the rationality of the resultant altruistic disposition, and of the altruistic actions to which that disposition leads. Let us say that a decision to act altruistically which is taken in order to further self-interest is a. selfishly altruistic decision. It is clear that, if decisions are plan-serving, then selfishly altruistic decisions are impossible. For they will not be motivation-perpetuating in their effects on action. The selfishly altruistic decisions are taken in order to further self-interest. But the altruistic acts to which they lead do not further self-interest, and agents are unlikely to suppose otherwise. It is less obvious that selfishly altruistic decisions cannot have a 2

See Gauthier's Morals by Agreement, pp. 184-9 o n this point.

Decision-making and Teleology

175

belief-dependent influence on action. Deterrent decisions to do A if C, we have seen, are taken on the basis of belief that they will prevent C's truth. Hence they cannot be executed without the agent's losing the belief on which he based his decision. But this is not obviously true of selfishly altruistic decisions to do A whenever C. Agents do not necessarily take such decisions on the basis of belief that taking them will prevent any instance of C - that taking them will prevent any occasions arising in which the decisionmaker's sacrifices would benefit others. What does motivate the decisions is the expectation that the agent will be a net beneficiary of taking them. So the decisions are motivated by belief that instances where C is true will not arise too often, and at too great a cost to the agent. Within limits, therefore, the execution of these decisions is entirely consistent with the retention of the beliefs on which the decisions were based. Suppose a selfishly altruistic decision is strictly belief-dependent in its effects on action. Then the agent will not be disposed to altruism under conditions which are too demanding on his resources -where there is too much call on him for self-sacrifice, or where the compensatory benefits from the altruistic disposition are too few. For belief that such conditions would arise, if held at the time of the altruistic decision, would have prevented the selfinterested agent from taking it. But that means that the disposition induced by such a decision is rather qualifiedly altruistic. The agent is indeed disposed to perform some genuinely altruistic actions - to perform some actions which sacrifice his interests for others when he has nothing to gain thereby. But this disposition to altruism is no unconditional disposition: it is not the disposition to altruism possessed by those whose altruism flows from the heart. And it is the rationality of such an unconditional disposition towards altruism, and of the self-sacrifice to which such a disposition may lead, which many philosophers have rightly wanted to defend. Selfishly altruistic decisions which induced such a disposition could not have an influence on action which was belief-dependent. Gauthier's Teleological theory of decision rationality commits him to denying REASON-APPLY, REASON-APPLY restricts the ends which can justify deciding to do A to those ends which provide at least as much justification for doing A thereafter. But Gauthier is committed to denying that there is any such restriction on the ends which can justify our taking a decision. Deterrent ends supposedly

176

The psychology of freedom

justify deterrent decisions; but they certainly do not justify executing deterrent decisions which have failed to deter. Selfinterest supposedly justifies selfishly altruistic decisions; but it does not justify executing those decisions. Gauthier must therefore explain the rationality of decision execution other than in terms of the justifications there were for the original decision. Yet in his account of deterrent decision-making, Gauthier failed to provide any such explanation. He simply assumed what REASON-APPLY would explain - the rationality of executing rational decisions. We might naturally wonder whether, having gutted our ordinary conception of a decision of most of its substance, this is an assumption Gauthier can sensibly continue to make. KAVKA AND LEWIS

Kavka and Lewis are more consistently revisionary in their theory of decision rationality. Like Gauthier, they defend a Teleological conception of decision rationality. But they deny that decisions which count as rational on a Teleological basis are necessarily rationality-preserving in their effects. Suppose a decision to do A is taken, and rationally taken because it maximises expected utility. Is the action decided upon thereafter going to be rational too? Not necessarily. As Lewis puts it: decisions are actions - and so too are the actions executing decisions. Each action is governed by a rationality which is Teleological: each action is, if rational, utility-maximising. Now the decision or intention formation may be utility-maximising when the action decided upon certainly would not be. Hence, in such a case: the action of forming the intention . . . should be performed, and . . . the action [intended] should not be. ('Devil's bargains and the real world', P- 143) If both of deciding to act and acting as decided must, when done rationally, maximise expected utility, how can decisions be rationality-preserving? A decision to do A would only be rationalitypreserving if, given that such a decision maximised expected utility and was taken, then - in the absence of change, or justification for change in the beliefs basing the decision - doing A would thereafter maximise expected utility also. But we can easily find cases where deciding to do A could be utility-maximising without doing A

Decision-making and Teleology

177

thereafter being utility-maximising also. The toxin puzzle provides an example. Assume that the good, as measured by utility, is exhausted by what furthers the agent's self-interest. Then deciding to consume the toxin maximises expected utility: the agent is likely to win £1 m thereby, for an insignificant cost. But consuming the toxin thereafter certainly would not maximise utility: by consuming the toxin the agent incurs only a cost. Kavka and Lewis both agree that deterrent decisions are possible and may be rational: deterrent decisions may be justified by their likely deterrent effects. They also agree that retaliating as decided may not be rational even if the original decision to retaliate was taken rationally. The deterrent decision might be utilitymaximising without its being true that retaliating as decided would thereafter be utility-maximising as well. And so we arrive at the paradox of deterrence, as Lewis and Kavka term it. Paradoxically, they claim, there may be sufficient justification for forming an intention to act which, thereafter, there is no justification for executing. But this is not a paradox that can trouble us if we abandon Teleology and suppose - as we should - that decision-making is a reasonapplying agency. If decision-making is a reason-applying agency then R E A S O N - A P P L Y must be true. But if R E A S O N - A P P L Y is true, there cannot be justifications for taking a decision which do not, once the decision is taken, provide corresponding justification for acting as decided. 3 So Lewis and Kavka refuse to infer from the rationality of a decision to the rationality of the action decided upon. But what about the rationality of the intervening state of intention? H e r e is what Kavka claims about the intention formed: an intention that is deliberately formed, resides at the intersection of two distinguishable actions. It is the beginning of the act that is its object and it is the end of the act that is its formation. As such, it may be assessed as rational (or moral) or not, according to whether either of two different acts promotes the agent's (or morality's) ends . . . the assessment of the rationality (or morality) of the agent's intentions will depend upon whether these intentions are treated as components of their object-acts or 3

Kavka in 'Some paradoxes of deterrence', p. 19 lists the following philosophers as denying that the paradox of deterrence could ever arise - as denying that there could ever be sufficient justification for forming an intention to act which, thereafter, there is no justification for executing: Abelard, Aquinas, Butler, Kant, Bentham, and Sidgwick! This pretty catholic consensus against him and Lewis appears not to disturb.

178

The psychology of freedom

their formation-acts. If treated as both, conflicts can occur. It is usual and proper to assess the practical rationality of an agent, at a given time, according to the degree of correspondence between his intentions and the reasons he has for performing the acts that are the objects of those intentions. As a result, puzzles . . . emerge when, for purposes of moral analysis, an agent's intentions are viewed partly as components of their formation-acts. ('Some paradoxes of deterrence', p. 23) In the case of a rationally taken deterrent decision, then, we could appraise the rationality of the intention formed in either of two ways. We could appeal to the fact that the utility-maximising decision which formed it was taken rationally - in which case the deterrent intention comes out as correspondingly rational. Or we can appeal to the fact that the utility-minimising deterrent action intended would be very irrational, in which case the intention comes out as correspondingly irrational too. It would be 'usual and proper' to determine intention rationality by reference to the rationality or otherwise of the action intended. But perhaps, in Kavka's view, it is not improper or absurd to refer to the rationality or otherwise of the initial decision instead? If so, there is no single fact of the matter about whether the deterrent intention is rational. But within the framework of Lewis' and Kavka's theory, there is obviously a rather more straightforward way of determining intention rationality than that. One just claims that it is rational at any time to intend to do A if and only if then holding the intention is utility-maximising. For, as I have already argued, intention is an inherently active state. It consists in being and remaining decided, and so in refraining from the activity of changing one's mind and deciding otherwise. If Teleology is the right model of the rationality of deciding in the first place, why is not it the right model of the rationality of remaining decided — of refraining from deciding otherwise? In which case, holding deterrent intentions is going to remain rational for as long as they are likely to work - and very irrational once it becomes clear that the enemy is not deterred. And, in the toxin puzzle, that would make it rational initially to hold an intention to take the toxin. But holding the intention after the prize was won would be irrational, because the intention would no longer be utility-maximising. For example, if it were a condition of winning the prize that an intention to drink the toxin be held until, say, shortly before the intended time of toxin-taking, then the

Decision-making and Teleology

179

intention to take the toxin would be held rationally up to that time, but irrationally thereafter. Why then not follow Lewis and Kavka? If decisions to act are actions themselves, why not assume that justifications for taking decisions work exactly like justifications for performing the actions which our decisions explain? If actions are made rational by their maximising good, so too are decisions to act. But the price of making that assumption is clear. We would be forced to give up entirely the idea, which motivates so much support for the Pro Attitude model, that decision-making is a reasonapplying activity - that decisions serve the rationality of the actions which they explain. But why should we so readily give up this idea? For it is an idea that seems to be true. The whole point of taking decisions about which actions we shall perform, after all, is to ensure that we end up performing the right actions. Our ordinary conception of the will ties justifications for our decisions to act to providing justification for the subsequent actions decided upon. As Teleologists about decision rationality, Lewis and Kavka quite consistently abandon this tie. But, in so doing, they are abandoning a quite central feature of our ordinary conception of the will - a feature no less central to it than the idea that decisions are actions themselves. It is a feature of common sense psychology which we should accommodate if we can. We may not succeed. But Lewis and Kavka have not even begun to try. INDIRECT TELEOLOGY AND GAUTHIER-PLANS

It is worth considering a more subtle Teleological theory of decision rationality suggested by Gauthier's more recent work.4 This theory of decision rationality is not directly Teleological. It does not suppose that rational decision-making is a utility-maximising agency. Instead it tries to explain decision and action rationality on the same indirectly Teleological basis - a basis which appeals to the I refer to Gauthier's 1989 paper 'In the neighbourhood of the Newcomb predictor'. I have changed the example used from the Newcomb problem to the toxin puzzle. It is also fair to emphasise that the theory is very much one that is suggested by that article. Gauthier has not yet to my knowledge published a theory in the exact form presented here. And, in any case, as he has revealed to me in correspondence, his views on various relevant issues are far from fixed. The theory which I propose under his name is of interest in any case. But it is of course possible that Gauthier would now express the theory differently.

180

The psychology of freedom

utility-maximising properties of plans. Hence, as we shall see, this new theory of decision rationality is not exactly a Teleological version of the Action model. It is instead a quite new Teleological theory of decision rationality - a theory which is importantly different from either of the Action or the Pro Attitude models. This latest indirectly Teleological theory of decision rationality is subtle. It promises, at least at first, to do the impossible. It promises to make a broadly Teleological theory of decision rationality consistent with my own theory of decision-making as a reason-applying agency. But this indirect Teleology comes at too great a price. It forces us to give up a bedrock theory of action rationality. Consider the toxin puzzle again. And let us continue to assume that the good, as measured by utility, is exhausted by the agent's self-interest. Deciding to consume the toxin tomorrow - to do A maximises the agent's expected utility. By deciding to do A the agent wins £1 m at little cost. But doing A thereafter is not utilitymaximising: to doing A there attaches only a cost. Nevertheless, the agent would still be vastly better off now taking the decision to do A and thereafter doing A than he would be now deciding not to do A and thereafter not doing A. If he now decided to do A and did A, the agent would be left with £1 m at the cost of mild discomfort. If he now decided not to do A and refrained from doing A, he would be spared some discomfort - but at the cost of £1 m. Call a Gauthierplan5 the combination of an action and a prior decision to perform it. So in the toxin puzzle one Gauthier-plan — GI — is the decision taken now to do A later, combined with later doing A. An alternative Gauthier-plan - G2 - is the decision taken now not to do A later, combined with the later refraining from doing A. GI clearly maximises expected utility compared to G2. The claim suggested by Gauthier's recent work is, very roughly, that a self-interested agent who is rational will pursue self-interest — maximise expected utility — not in respect of decisions, nor in respect of actions, but instead in respect of Gauthier-plans. When deliberating prior to a decision about whether to do A, the agent will consider, not simply which decision or which action would further his self-interest, but which Gauthier-plan would further his self-interest. He will take the decision contained by whichever is the self-interest furthering Gauthier-plan. And thereafter he will 5

My terminology.

Decision-making and Teleology

181

perform the action which executes that decision - again because it is part of the same self-interest furthering Gauthier-plan. In the toxin puzzle, that means deciding to do A and doing A: the agent will decide to consume the toxin, and then consume it as decided. So what moves a rational agent to take a decision is desires for the ends furthered by the Gauthier-plan containing that decision. And the action which results from the decision will again be moved by exactly the same desires - desires for the same ends furthered by that very same Gauthier-plan: Theory U [Gauthier's theory] gives you, as at least some of your reasons for [acting], your preferences for the outcomes of plans. ('In the neighbourhood of the Newcomb predictor', p. 192) Up till now we have assumed REASON. We have assumed that what moves an agent to perform an action A is always a desire for some end combined with belief that doing A would further that end. So, in the toxin puzzle, a desire for self-interest could not possibly move the agent to do A - to consume the toxin. For the agent believes that doing A would not further his self-interest. Any action A which resulted from a prize-motivated decision to do A would therefore have to be motivated by a desire for some other end - such as by a desire to execute the prior decision. Self-interest might provide the agent's motive for his prize-winning decision; but it could not, thereafter, be his motive for acting as decided. But Gauthier, it seems, is denying that REASON is true. When the agent in the toxin puzzle does A having decided to, his action, according to Gauthier, is supposed to be motivated by the same preference as motivated the prior decision. And this preference is a preference for the utility-maximising outcome of the Gauthier-plan containing that action and the prior decision. This is a preference for an outcome in which the agent wins £1 m at the cost of mild discomfort over an outcome in which the agent avoids discomfort but misses out on £1 m. This is a preference for what is in the agent's self-interest. On Gauthier's theory, therefore, a desire for self-interest is supposed to motivate decision and action alike. And this even though self-interest would not actually be furthered by doing A. A rational agent's motive for doing A, according to Gauthier, need not consist in a desire for an end E and a belief that doing A would further E. It may consist instead in a desire for E,

182

The psychology of freedom

and a belief that E would be furthered by a Gauthier-plan - a plan containing the action A and a prior decision to do A. It follows that JUSTIFY is also false. An agent's desire for an end E can motivate him rationally to do A although doing A would not further that end. So the desirability of E can justify doing A even though A would not further E. Desirable ends can justify performing actions, not because the actions would further those ends, but because Gauthier-plans containing those actions would further those ends. Thus, in the toxin puzzle, a desire for selfinterest is what moves a rational agent to do A. So self-interest is what justifies doing A - though only because self-interest is furthered by the Gauthier-plan containing that action and the prior decision to perform it. So, according to Gauthier, a rational agent's decision is justified and motivated by the ends furthered by a Gauthier-plan — a plan which contains both it and the action which executes it. And, once the decision is taken, the action which then executes that decision is justified and motivated by the very same ends - the ends furthered by the very same Gauthier-plan containing the action and that prior decision. Denying JUSTIFY and REASON allows for the combination of doctrines which, I have so far argued, are inconsistent. Gauthier can claim that prize-motivated decision-making is possible and may be rational. In the toxin puzzle, it is rational for an agent who desires the prize to take the prize-winning decision. But, despite allowing for rationally prize-motivated decision-making, Gauthier's new theory still has REASON-APPLY come out true. Once a decision is taken, any ends which justified the decision thereafter provide corresponding justification for the action decided upon. So prizemotivated decisions maybe motivation-perpetuating and rationalitypreserving in their effects on action. Notice that this new indirectly Teleological model of decision rationality will not allow for rational deterrent decisions. It will not allow for conditional decisions which are rationally motivated by belief in their deterrent effect. It is easy to see why. The justifications for a conditional decision, on this new theory, like those for any decision, must be determined by reference to the expected utility of a Gauthier-plan. This Gauthier-plan will, as usual, include both the decision and action which executes it. Now a decision to do A if C can only be executed if C. So, to determine justification for

Decision-making and Teleology

183

the decision, we must consider the expected utility of the following combination of events: the decision is taken, C is true, and A is done. But that means that the likely deterrent effect of a decision to do A if C - the decision's propensity to prevent C's truth - is ignored in determining the agent's justification for taking the decision. Decision justifications are determined by reference to a situation where C is true - where the decision has no deterrent effect. And of course, while, on this new indirectly Teleological model, the likely deterrent effects of a conditional decision are not taken into account in determining its justification, the unpleasant consequences of actually executing it - of doing A when C certainly are. Why should we believe this novel theory of decision and action rationality? Gauthier argues that pursuing ends at the level of Gauthier-plans is more efficient: it produces more good or utility. In the toxin puzzle the agent who pursues self-interest at the level of Gauthier-plans, does better than an agent who pursues it only at the level of actions. But this consideration is not decisive. This is just to assume that reason governs Gauthier-plans Teleologically in the way that, in standard PDT, it is supposed to govern action. This is just to assume that Gauthier-plans are justified, along with the decisions and actions which constitute them, to the extent that those Gauthier-plans increase good. And Gauthier needs to argue for this crucial assumption. But there is every reason for supposing that Gauthier's assumption is false, for were Gauthier's assumption true, then REASON and JUSTIFY would be false. Actions would be justified by desirable ends furthered by Gauthier-plans, and not only by desirable ends furthered by performing those actions. Rational agents would be motivated to act by their desires for the outcomes of Gauthierplans, not only by their desires for the outcomes of actions. Yet REASON and JUSTIFY are very plausibly true - they are as evident, as any propositions can be in the disputed study of practical reason. JUSTIFY, in particular, is a luminous claim about action rationality which we assume every day. Whenever we persuade people into performing an action, we always appeal to what ends the action itself would further. We always point out that the recommended action would cause or constitute the attainment of some desirable end. No theory of action rationality can credibly deny JUSTIFY.

184

The psychology of freedom Plans and the practical primacy of action

Gauthier condemns the pursuit of self-interest - the maximising of expected utility - at the level of actions as destructive of what he calls 'autonomy'. The supposedly rational agent who performs actions only if those actions maximise expected utility: becomes the victim rather than the master of her circumstances, unable to introduce into her life any coherence beyond that realisable in a sequence of maximising [actions]. As we noted, her plans can be nothing more than projections of such sequences. On such a view, [that rational agents perform actions only if they maximise expected utility] the traditional link between rationality and autonomy is lost. ('In the neighbourhood of the Newcomb predictor' p. 194) What is Gauthier's understanding of what he calls 'autonomy' and its connexion with rationality? It seems somewhat like my own conception of freedom of action and its connexion with rationality. Both Gauthier and I suppose that practical reason is plan-centred. Practical reason characteristically recommends agency to us in the form of plans - sequences of agency extending through time. We also agree that freedom of action requires that an agent have a capacity to apply a plan-centred practical reason. And we both suppose that the function of decision-making is to help us apply practical reason as it recommends plans. But our conceptions of a plan are importantly different. And since, for both of us, decision-making's function is to help us apply reason's recommendations of plans, this difference in our conception of plans leads to an important disagreement about the function and nature of the will. My conception of a plan has all along been of a sequence of actions - first-order agency - extending through time. So plans on my understanding include actions only, and not our prior decisions and intentions to perform those actions as well. That means that, on my conception of a plan, the function of decision-making is to help apply practical reason as it concerns the first-order agency of our action. Hence, on my theory, action is the agency with which practical reason is primarily concerned. Decision making simply facilitates the application of practical reason as it concerns action. All along I have been teaching what I shall term the doctrine of the practical primacy of action.

Decision-making and Teleology

185

Whereas on Gauthier's conception of a plan, a decision or intention is as much a component of a plan as the action decided upon. Practical reason, in so far as it recommends plans, recommends equally both decisions to act and the actions decided upon. Action is accorded no practical primacy, but is on the same level as decision. Decision-making is not for Gauthier what it is for me - a special executive agency for applying reason as it concerns our subsequent action. In the next chapter, I shall be arguing further for the practical primacy of action. Unless we teach action's practical primacy, and see practical reason's concern with decision-making as secondary, we shall find it impossible to give a consistent defence of our Psychologising conception of our own freedom. In particular, as I shall argue, we shall not be able to refute Hobbes' Regress argument against the freedom and agency of the will. Unless we teach action's practical primacy, we shall not be able to defend the very belief in an agency of the will which Gauthier's theory of decision rationality assumes. CONCLUSION

I shall be arguing that decision-making is a means-end justifiable agency - that the Action model and not the Pro Attitude model of decision rationality is true. But that cannot mean that rational decision-making is a goodness- or utility-maximising activity. The true theory of decision rationality cannot be Teleological in any form, direct or indirect - not at any rate if both REASON-APPLY and JUSTIFY are true. REASON-APPLY restricts the ends which can justify deciding to do A to those ends which would thereafter justify doing A. And then JUSTIFY restricts the ends which can justify doing A to those ends which would be furthered by doing A. Now decisions to do A, and so too the Gauthier-plans of which they form a part, can easily further goodness-maximising ends - such as winning huge prizes - which would not be furthered by doing A thereafter. Teleology implies that those goodness-maximising ends would justify deciding to do A. REASON-APPLY and JUSTIFY imply that they cannot. So, if REASON-APPLY and JUSTIFY are true, Teleological theories of decision rationality must be false. Remember why REASON-APPLY looks plausible. We naturally

186

The psychology of freedom

conceive of the will as having an executive function. The function of taking decisions to act is to apply reason as it concerns subsequent action. That means that justifications for deciding to perform an action must provide corresponding justification for performing the action decided upon thereafter. A convincing theory of decision rationality has to reconcile our belief in the agency of the will with our belief in its executive function. If we end up defending a version of the Action model - as I shall be doing - it must be the Action model in a form consistent with REASON-APPLY. Over the last three chapters of the book I shall be gradually developing a new theory of decision rationality. This theory will do precisely what extant versions of the Action model do not. It will explain the agency of the will in a way that is entirely consistent with the will's executive function. The credibility of this new theory will lie both in its internal coherence, and in the fact that it does full justice to the intuitions that have led many philosophers wrongly to reject the Action model, and to espouse the Pro Attitude model of decision rationality instead — in many cases denying the agency of the will altogether in so doing.

CHAPTER 7

The Regress argument

INTRODUCTION

In this chapter and the next, we shall be addressing the sceptical arguments against a freedom and second-order agency of the will. This chapter will discuss and refute Hobbes' Regress argument. The next chapter will refute the argument from Non-Purposiveness and the Reduction argument. The purpose of these two chapters, however, goes beyond refuting scepticism about second-order agency. The chapters will also serve to develop the case for the theory of decision rationality which I favour - a theory which by combining the Action model with REASON-APPLY, does justice not only to the agency of the will, but also to its executive, reason-applying function. This chapter is going to add to the case for REASON-APPLY by showing that the principle does work in explaining important and very intuitive limitations to the scope of the will. The next chapter is then going to be devoted to arguing for the Action model itself. Then, once scepticism about second-order agency has been disposed of by these two chapters, the book's final chapter can be devoted to establishing that, far from being in tension with one another, REASON-APPLY and the Action model of decision rationality are entirely consistent. THE REGRESS ARGUMENT

Chapter 2 introduced Hobbes' Regress argument against freedom of will. Hobbes began by assuming that, in order to have freedom of action, one must have a capacity to act as one wills. He inferred that, therefore, in order to have freedom of will, one must have a capacity to will as one wills. But, Hobbes claimed, there is no such capacity: 187

188

The psychology of freedom

I acknowledge this liberty, that I can do if I will, but to say, I can will if I will, I take to be an absurd speech. (£Of liberty and necessity', p. 240) In Hobbes' view, our freedom of will - of decision - depended on the possibility of decisions to decide determining our decisions in the same way that our decisions to act determine our actions: The question is whether the will to write, or the will to forbear, come upon a man according to his will. ('Of liberty and necessity', p. 240) Hobbes denied that our decisions to decide could so determine our decisions. We cannot, for example, take a decision to write something just on the basis of having beforehand decided so to decide. So, Hobbes concluded, there is no freedom of will. The Regress argument does not threaten freedom of will alone. The very possibility of a second-order agency of the will is at stake too. For, as I suggested in chapter 2, it is hard to see how secondorder agency can be possible if second-order freedom is ruled out — and ruled out by the very nature of decision-making itself. The nature of anything which counts as genuine agency — as a genuine doing - must at least be consistent with our having control over whether we do it. If decision-making, by its very nature, precludes our ever possessing and exercising control over which decisions we take, why suppose that taking decisions is something we do?

The Regress argument raises and connects two important problems. The first problem is about the nature of freedom. Some sort of capacity to act as we decide is plausibly necessary to our freedom of action - to the freedom of our first-order agency. The question which the Regress argument raises is whether this plausible condition on freedom of action generalises to our secondorder agency as well. Do free decision makers need a capacity to decide as they decide? The second problem is about the nature of the will itself. The problem, specifically, is about the scope of the will. What can we do on the basis of deciding to do it? Showing where the Regress argument goes wrong requires a theory which can address both problems - precisely the theory of freedom and the will which we have been developing so far. In this chapter we shall both apply this theory, and develop it further.

The Regress argu men t

189

THE STRUCTURE OF THE REGRESS ARGUMENT

The Regress argument makes an important assumption about the nature of freedom. It assumes that any condition on freedom of action is also a condition on freedom of will. This is an assumption which has widely been made by philosophers, whether or not they believe in freedom of will. It has been made by disbelievers in free will like Hobbes. But it has also been made by believers. Frankfurt is one such believer, and he has claimed: It seems to me both natural and useful to construe the question of whether a person's will is free in close analogy to the question of whether an agent enjoys freedom of action. ('Freedom of the will and the concept of a person', p. 20)

And it is clear enough what Frankfurt means by this talk of close analogy. Conditions on freedom of action are going to be conditions on freedom of will as well - as we saw in chapter 2 from Frankfurt's own account of freedom of will: Now freedom of action is (roughly at least) the freedom to do what one wants to do. Analogously, then, the statement that a person enjoys freedom of will means (also roughly) that he is free to want what he wants to want. ('Freedom of the will and the concept of a person', p. 20) No one, of course, would deny that there is going to be some analogy between freedom of will, if we have it, and freedom of action. The question is whether, as Frankfurt supposes, the analogy must be that close. Let us call the assumption that any condition on freedom of action is also a condition on the freedom of our agency generally, and so on freedom of will, the Close Analogy assumption. A second assumption behind the Regress argument, is that it is a condition on freedom of action that action be subject to the will. Put very roughly, the assumption is that we are free agents only because we can act as we decide — only because our actions are determined by, or at least determinable by, prior decisions to act. In this rough form, this assumption is highly intuitive. It is less clear what, put more precisely, the will subjection of action comes to — and why exactly the subjection of action to the will should be a condition on the freedom of our action. As we shall see, there is more than one version of the Regress argument; and versions differ, in particular, in their view of why the will subjection of action should be a condition on freedom of action.

190

The psychology of freedom

The Close Analogy assumption takes us from the doctrine that a free agent's action is subject to the will to a claim about a free agent's agency generally: to the claim that it is a condition on the freedom of any agency, including that of the will, that that agency be subject to the will. So our decisions to act - if taken freely should be determinable by decisions to decide, as our actions are determinable by decisions to act. I shall call this doctrine the Will Subjection theory of freedom,

Hobbes believed the Will Subjection theory of freedom. But the theory has had many defenders besides Hobbes. Frankfurt takes the will to consist in holding desires which are sufficient to motivate us to do the thing willed. And he clearly supposes that anything we do freely, including holding will-constituting desires, may, like action, be done because its doing is willed, as Frankfurt understands the will. Frankfurt is another believer in the Will Subjection theory of freedom. Using the phrase 'free action' or 'free agency' to mean any action or agency over which an agent has control, we can represent the Regress argument as an argument of the following form: (i)

Any free action must be subject to the will.

Then we have the Close Analogy assumption: (ii) Any condition on free action is a condition on free agency generally. (i) and (ii) entail the Will Subjection theory of freedom: (iii) Any free agency must be subject to the will. But (iv) No decision may be subject to the will. (iii) and (iv) then entail (v)

No decision can constitute free agency.

To make the Regress argument work, we first have to find some sense in which both premisses (i) and (iv) come out true — some way in which free action, at least, must be subject to the will, but in which decision-making may not be. Now I shall argue that there certainly is a sense in which premisses (i) and (iv) are both true. But that is not enough to make

The Regress argument

191

the Regress argument work. Where the argument fails is with premiss (ii) - the Close Analogy assumption. The Close Analogy assumption is a venerable one. Its popularity among theorists of freedom is easy to explain. For a very important and natural reason - a reason which I shall be discussing later on - we take action to be a paradigm of our agency generally. We tend to assume that any genuine agency must work like action unless we are shown definite grounds why it should not. Scepticism about second-order agency has always exploited precisely this paradigmatic status which action has for us. The sceptical arguments quite generally depend on pointing out that what is commonsensically supposed to be active decision-making is crucially unlike action in various respects - and that therefore it cannot be genuine agency. To meet such scepticism we need a general model of agency which can explain how there can be agency that is unlike action in the relevant respects. Without such a model, scepticism about second-order agency simply will not go away. As for scepticism about second-order agency generally, so for that expressed by the Regress argument in particular. If we possess an intuition that some condition, such as subjection to the will, is a condition on our freedom of action, then taking action as our paradigm for agency generally, we are going to generalise the condition — unless we have definite justification for not generalising it. And philosophers have hitherto failed to come up with the definite justification required. Hence the Regress argument's continuing force. It is easy to see where any such justification would have to come from. A theory of freedom would have to tell us exactly why the condition on our freedom of action applies. And then a theory of decision-making and its relation to action would have to tell us of relevant differences between the agency of our decision-making and that of our action — differences which preclude the condition applying to our freedom of decision-making also. In other words, justification for abandoning the Close Analogy assumption would have to come from a unified theory both of freedom and of our agency in its various forms, first and higher order. The Close Analogy assumption has been made by philosophers^wte de mieux — for lack of any such unified theory. But this lack we are now in the process of supplying.

192

The psychology of freedom HOW IS ACTION SUBJECT TO THE WILL?

The Regress argument depends on there being a way in which actions are subject to the will, but decisions are not. What way might that be? Hobbes denied that 'the will to write, or the will to forbear, come upon a man according to his will'. Now Hobbes was not claiming, I take it, that decisions to do A can never be explained by prior decisions to take those decisions. That claim would be false. There is nothing in principle to stop us taking decisions to take specific decisions — decisions to decide to do A; nor, in principle, is there anything to stop such decisions thereafter explaining the first-order decisions decided upon. Suppose - to consider a version of the toxin puzzle again - that you now offer me £1 m for deciding tomorrow to drink a toxin the day after tomorrow: I win the prize if and only if tomorrow I take the decision to drink the toxin. Now I might well decide today to take that prize-winning decision tomorrow, and do so quite rationally- after all, if I manage to take that prize-winning decision tomorrow, I will win £im. On the basis of taking that decision to decide today, I might then perform some action in order to get myself to take the prize-winning decision. For example, I might visit a hypnotist, and get him to hypnotise me so that I will then be motivated to decide to take the toxin. My decision to take the prizewinning decision could thus explain my visit to the hypnotist — and, thanks to that visit, my taking the prize-winning decision. In this example, my decision to decide to do A is also a decision to act. It is a decision to perform some action which will then cause me to decide to do A. And that is precisely how the decision to decide explains the decision decided upon - by way of an intervening action, which is what actually causes me to be motivated to take the decision decided upon. The decision to decide explains a motivation to take the decision decided upon — but only in an indirect, action-dependent way.

Decisions to act, by contrast, explain the motivation to perform the action decided upon in an action-independent way — by directly explaining the required motivation to act. When we decide to perform an action A, that decision alone is enough to ensure that we are left motivated so to act. The decision does not first have to cause us to perform some intervening action by which, in turn, we

The Regress argument

193

cause ourselves to be motivated to do A. To cause us to perform actions, decisions do not have to cause us to perform other, motivation-inducing actions first. That must be true, otherwise our decisions could not get us to perform actions at all. There is a way, then, in which our decisions can determine our actions - a way in which it is much less obvious that our decisions can determine our decisions. Our decisions can determine our actions in an action-independent way - we can perform actions just on the basis of deciding to perform them. And that is because deciding to perform an action A is of itself sufficient to leave us motivated to do A. Once one has decided to act, one is left motivated to act as decided, and the action decided upon will thereafter follow - provided, of course, one does not subsequently change one's mind, and one has the required capacities and know-how. Decisions do not plausibly determine our decisions in the same way. The difficulty does not arise only when the decision decided upon is odd or irrational - as when the decision decided upon is a decision to perform a silly action such as toxin drinking. The difficulty is quite general. I cannot take any specific decisions, no matter how innocuous and ordinary, just on the basis of some prior decision to take them. Return to the example of chapter 2.1 cannot take a decision that tomorrow at 10 o'clock I shall decide to visit the dentist - and then expect to take that decision tomorrow just on the basis of today's decision to take it. Now that is not because I lack any capacity or know-how which might be required to take this perfectly ordinary decision decided upon. Nor is it because I cannot take decisions to take specific decisions at all. We have just looked at an example of how I might, and quite rationally, take such a decision to decide. The explanation must be that though I can take decisions to take specific decisions, taking them will not of itself cause me to be motivated to take the decision decided upon. Deciding that tomorrow at 10 o'clock I shall decide to visit the dentist will not of itself cause me to be motivated to take that decision tomorrow. That must be true: otherwise I could take the decision to decide, thereby induce the required motivation - and then so motivated, at 10 o'clock tomorrow, decide as decided. Since that is not possible, there is an evident limitation on the effects which decisions can have on our motivation. And this is a limitation which must somehow be explained.

194

The psychology of freedom

Deciding to act directly causes us to be motivated to act as decided. Deciding to take a particular decision to act does not similarly cause us to be motivated to decide as decided. But decision-making, whenever we do it, is motivation-affecting; it always leaves us with a persisting intention - and so with some kind of persisting motivation to execute our decision. Now, as we have seen, the motivation which a decision to take a specific decision leaves is not a motivation to do what would constitute its execution it is not a motivation to take that specific decision. The motivation left must, instead, be a motivation to do what would cause its execution. A decision to take a specific decision must leave us motivated to act in a way that would cause ourselves to take that decision decided upon. Decisions to take specific decisions are decisions to act in a decision-causing way. Return to the example we considered above of the decision to take a prize-winning decision - an example of how one might actually come to take a decision to take a specific decision. The effect of that decision to decide is to leave one motivated to perform some action which will cause the decision decided upon. One is left motivated to get oneself hypnotised into taking the decision. And so quite generally - if I do ever decide to take a decision to drink a toxin, or if I do ever decide to take a decision to visit the dentist, that decision will leave me motivated, not to take the decision decided upon, but to perform an action which will cause me to take that decision. The decision to decide, in other words, will also be a decision to act in a decision-causing way. And so when such a decision to decide does explain the taking of that decision decided upon, that will be through having motivated an intervening, decision-causing action. Let us say that any activity A is subject to the will if and only if a decision to do A would directly cause the decision maker to be motivated to do A. Notice that will subjection, as I have just defined it, does not guarantee that a decision to do A would cause its maker actually to do A. Having decided to do A, I might change my mind and abandon my decision. Or I might lack capacities or know-how needed for doing A. But if doing A is subject to the will, then, for as long as I remain decided on doing A, I shall be motivated to do A: if I fail to perform the agency decided upon, it will not be for lack of the motivation required. Where any agency A is subject to the will, and I do possess such capacities and know-how as are required for

The Regress argument

195

doing A, then I shall have a capacity to do A at will. I only have to decide to A and - unless my will changes - 1 shall do A. As we have just seen, deciding to perform an action would directly cause one to be motivated to perform that action. Deciding to take a particular decision to act, by contrast, would not directly cause one to be motivated to take that decision. We have found a way in which our actions clearly are subject to the will, whereas our decisions are not. We have now to explain exactly why decisions are not subject to the will as actions are. We need an account of what limits the scope of the will. WHY DECISIONS, AND PRO ATTITUDES TOO, ARE NOT SUBJECT TO THE WILL

Not only are not decisions subject to the will. As we have already seen, desires and practical judgments do not seem to be subject to the will either. Suppose I decide that tomorrow at 10 o'clock I shall form a desire to visit the dentist - or suppose I decide that I shall then make a judgment that visiting the dentist would be desirable. I cannot expect that just on the basis of so deciding, I shall tomorrow at 10 o'clock form the desire or make the practical judgment decided upon. Taking a decision to desire, or taking a decision to judge something desirable, will not directly motivate the desire or judgment decided upon. It might be argued that there is a simple reason why practical judgments and desires are not subject to the will. The reason is that for something to be subject to the will, it has to be a form of agency - and making practical judgments and forming desires are not agency. But this argument is far too superficial. For one thing, although I agree that making practical judgments and forming desires are not agency, the claim that they are not is not uncontroversial. And, in any case, the non-agency of practical judgments and desires cannot be the whole story. For decisions - I certainly want to argue - do constitute genuine agency, but are not subject to the will either. Now it would be very strange indeed if what prevented decisions being subject to the will had nothing to do with what prevented desires and practical judgments from being subject to the will. It is a sensible starting assumption - even if we might have to abandon it eventually - that there is some general explanation for why none of decisions, practical judgments, or

196

The psychology of freedom

desires are subject to the will. But if there is a general explanation, and yet decision-making is a form of agency, that explanation cannot lie simply in the non-agency of the events and states concerned. It is certainly true that only agency can be subject to the will. But a theory of the will's scope - of what we can do just on the basis of deciding to do it - should explain that fact, and not simply assume it. The account of the will's scope which I am about to give provides the needed explanation. Will subjection presupposes that what is subject to the will has certain characteristics - characteristics which only agency can possess, but which decisions, albeit agency themselves, lack. By digging deeper, we can therefore explain what the more superficial argument which I am rejecting only presupposes. We can explain why whatever it is we do just on the basis of deciding to do it, must be - as ordinary English makes it impossible not to acknowledge — a genuine doing. We need a criterion for determining what can be subject to the will in the sense I have defined. And we need a criterion that does not beg the issue of whether only agency can be subject to the will. Notice, though, that, while not wishing to beg this question, I want also to continue discussion in English that, while abstract, remains reasonably natural. So, in what follows, I shall continue to talk of what the agent has decided to do, of what the will motivates the agent to do, and of the agent deciding to do A. But nothing in the argument will rest on the use of these locutions. The argument does not at any stage assume that only agency is subject to the will. That is something which the argument establishes. What is to be our criterion? We find it in our doctrine that decision-making must be reason-applying in its effects. In particular, decisions must be rationality-preserving causes of what the agent decides to do. Rationally deciding to do A must cause the agent to be motivated to do A rationally too. At least this is so in any case where the motivation for what we have decided to do is a direct result of the decision to do it - where there is no intervening action by which the agent causes that motivation. Sometimes, of course, I can rationally decide to do something irrational. I may have been promised a reward for doing what is irrational - as in the toxin puzzle, where I am offered a prize for deciding to take the toxin. If the reward for the irrationality is big enough, I certainly can rationally decide to do what is irrational, and end up performing the irrational doing decided upon as a

The Regress argument

197

result. But when I take such a decision, I am deciding to manipulate myself into doing something irrational. The rational decision will lead to the irrational doing decided upon by way of some intervening and rational action - an action by which I cause my performance of the irrational doing by causing myself to be motivated to perform it. The rational decision will lead to the irrational doing decided upon in an action-dependent way. It is where decision-making can give rise to the motivation for what is decided upon directly - causing the motivation for it other than by way of causing some intervening and motivation-affecting action - that rationally deciding to do something must leave it equally rational for the agent to do what he has decided. But that means we have the criterion of will subjection which we need. We can determine whether a given doing really can be subject to the will. Doing A may be subject to the will - deciding to do A would directly cause us to be motivated to do A — only if the rationality of deciding to do A guarantees the rationality of doing A thereafter. We have already seen in chapter 5 what that involves, REASONAPPLY must be true: suppose that a given end E provides justification for the decision to do A; then, once the decision has been taken, that end E must provide at least as much justification for doing A. When might an end justify deciding to A? Consider our toxin puzzle. The agent is offered a prize for deciding to drink the toxin. Now we have supposed that this prize provides the agent with some justification for deciding to take that prize-winning decision - for deciding to decide to drink the toxin. Indeed if the prize is big enough, the agent will surely have not some but sufficient justification for deciding to take the decision. His decision to take the prize-winning decision can then be perfectly rational. Why then does the prize justify deciding to decide to drink the toxin? The answer is clear. The prize means that there is a desirable end which would be furthered by the decision decided upon. We can generalise: the fact that doing A would further some desirable end E, so that it is accordingly desirable to do A, provides some justification at least for deciding to do A.1 1

For the record, this claim is in fact a simplification - though not in ways material to the present argument. Strictly, the claim is true only on condition that deciding to do A would or at least might cause the agent to do A; and that doing A would still further E were a prior

198

The psychology of freedom

So it is easy to see what must be true if doing A is to meet our criterion for will subjection. If doing A can be subject to the will, then any end E which would justify a decision to do A - any desirable end E which doing A would further — must, once the decision is taken, provide at least as much justification for doing A. Hence, doing A must be something which would be justified by any desirable end E which it furthered. For any such end will already have justified the decision to do A. So, if doing A is subject to the will, doing A must be justifiable by desirable ends which doing A would further: doing A must be means-end justifiable. Furthermore, there must also be no special restrictions on what ends could justify doing A. Any desirable end E which doing A would further must be eligible as a justification for doing A. It is clear that decisions will not meet our criterion for being subject to the will. And this is because decisions, if means-end justifiable at all, are what I call special-purpose activities. Not any desirable end E which deciding to perform an action A would further can justify deciding to do A. As we saw in chapter 5, unless there is this special restriction on the ends which might justify taking a decision, decision-making will not be reason-applying in its effects on action. Consider again the £1 m prize for my deciding tomorrow to drink the toxin the day after tomorrow. That prize surely provides some justification for my deciding now to take that prize-winning decision tomorrow. Winning the prize is certainly a desirable end which deciding to drink the toxin would further. So - as we have seen - I might rationally decide to take the prize-winning decision and then set about finding ways to manipulate myself into taking it. But, as has already been argued in chapter 5, if decision-making is to be a reason-applying activity, the prize cannot justify the prize-winning decision itself. Which is why I have to find ways to manipulate myself into taking that decision. If I could take the decision to do A taken. These further conditions must be met if the desirability of E is to constitute a justification for deciding to do A which the Action model recognises - if E is to be a desirable end which deciding to do A would further. The Action model and the consequent need for these qualifying assumptions are argued for in chapter 8. Notice that my claim in the text leaves it open that there are desirable ends which justify deciding to do A but which would not, at least prior to a decision to do A being taken, be furthered by doing A. Such decision-justifying ends are also discussed in chapter 8.

The Regress argument

199

decision simply on the basis of deciding to take it, decision-making would no longer be a reason-applying activity. Perfectly rational decisions could leave me motivated to do the irrational - in this case, to take prize-winning but irrational decisions. Not only have we explained why decisions and intentions cannot be subject to the will. We also have an explanation why desires and practical judgments cannot be subject to the will either. As I have argued, only means—end justifiable occurrences can be subject to the will. And desires and practical judgments are not means-end justifiable occurrences at all. They are pro attitudes - they are justified in terms of the desirable ends furthered by their objects, and not by the desirable ends which they themselves further. Finally, we have the explanation we want of why only agency can be subject to the will: why what deciding to do something would motivate us to do must be a genuine doing - a form of agency. For I take it that if what we are motivated to do is means—end justifiable, then it is ipso facto a form of agency. As I claimed in chapter 2, means-end justifiability seems to be reason in its distinctively practical form. Occurrences such as actions and tryings are evidently forms of agency because not only are they performed for reasons — motivated in some way, whether purposive or not — but they are means—end justifiable as well; and the same is true, I shall be arguing in the next chapter, of decisions. But then an occurrence which, though motivated, is not a case of agency, cannot be means-end justifiable either. In which case, the motivation for the occurrence cannot come directly from a decision that it occur. Nonagency cannot be subject to the will. To sum up so far. Where doing A is subject to the will, deciding to do A would of itself suffice to leave the agent motivated to do A. But decision-making is a reason-applying activity. It must be rationality-preserving in its effects on motivation, and on what that motivation explains. Where doing A is subject to the will, the rationality of deciding to do A must guarantee the rationality of doing A: any end which justifies deciding to do A must thereafter justify doing A also. And that means that doing A can only be subject to the will if it is a form of means-end justifiable agency which is not special-purpose. There must be no special restrictions on the ends which can justify the doing of A. Since decisions are special-purpose activities, the taking of specific decisions cannot itself be subject to the will. The motivation

200

The psychology of freedom

which deciding to take a specific decision would cause in us, cannot be a motivation to decide. It can only be a motivation to act in a decision-causing way. So any such decision to decide is going to be a decision to act in a decision-causing way, and gives rise to the decision decided upon by way of causing that intervening action. DECIDING TO MAKE UP ONES MIND

I may not be able to take specific decisions just on the basis of prior decisions to take them. But what about taking a decision one way or the other? What about making up one's mind? We do often take decisions to decide one way or the other. We do often take decisions on when and about what we shall make up our minds. And that is because it matters when we take decisions about things - and in what order things get decided. We need to ensure that we take decisions in time to take advantage of all the available options. But we also need to ensure that we do not take the decisions either before we have the information on which they need to be based, or at a time when we are least able to process that information efficiently. In particular, we need to make sure that decisions on which other decisions need to be based are taken first. We need, for example, to make sure that we have already decided where to go on holiday by the time we make our minds up about what route maps to buy. Not only do we take decisions when and about what to make up our minds. Such decisions seem reliably to lead to their execution. And this seems to be a contrast with decisions to take specific decisions. Just deciding to decide to drink a toxin, for example, will not get me to take the decision decided upon. To execute a decision to decide to drink a toxin, I should first have to perform some selfmanipulative, rationality-compromising action. I should have to get myself hypnotised, or take some decision-inducing drug. But I can execute a decision to make my mind up without having to go in for any such self-manipulation along the way. In between deciding to make my mind up and actually making my mind up, I do not have to appeal to hypnotists to hypnotise me into taking a definite decision. Nor do I have to drug myself into making my mind up. This contrast might suggest that, while the taking of specific decisions is not subject to the will, making one's mind up - taking a decision one way or the other - is. But the suggestion is false. One

The Regress argument

201

cannot make one's mind up at will. Deciding to make one's mind up about whether to do A - to take a decision one way or the other will not directly cause the motivation which taking such a decision one way or the other requires. It will not of itself leave one with either the motivation to decide to do A or the motivation to decide not to do A. The reason why is simple enough. It is exactly what explained why we cannot take specific decisions at will. Decisionmaking has to be rationality-preserving in its effects. Where doing A is subject to the will, the rationality of deciding to do A must guarantee the rationality of doing A. But the rationality of deciding to make one's mind up does not guarantee the rationality of actually making one's mind up. Consider the following example. Suppose that you offer me a prize - the usual £1 m - for taking a decision tomorrow, one way or the other, about whether to drink a substance T next week. The prize is not conditional on my taking any particular decision; nor is it conditional on my executing whatever decision I take. I simply have to make up my mind — and do it tomorrow. The problem is that I do not yet know, and will not know until the day after tomorrow, exactly what the effects of drinking T will be. I know that there are exactly two possibilities. Either I have condition X - and T is a rather unpleasant toxin to those with condition X. Or I have condition Y which may cause equally unpleasant symptoms - and T is a tonic cure for Y. Which of these two possibilities holds whether T is toxin or tonic - is something which I shall only discover after taking the prize-winning decision. Clearly, if I took a decision tomorrow about whether to drink T, and faithfully executed it, I might well end up having an unpleasant time - an unpleasantness I should have avoided by waiting to find out about my condition and the nature of T. On the other hand, let us suppose, the unpleasantness would not be so great as to offset the benefit of the £1 m I should gain by taking some decision or other tomorrow. So whether or not my decision tomorrow would be executed, I have sufficient justification for deciding today that I shall take a decision one way or the other tomorrow. 2 Suppose then that, quite rationally, I decide today to take a 2

Again - as supporters of the Action model will insist - this is assuming that the decision to take a decision tomorrow one way or the other might lead (whether through or independently of some intervening action) to such a decision being taken tomorrow.

202

The psychology of freedom

decision tomorrow about whether to drink T. And suppose that, as a result, I am led to make up my mind tomorrow about whether I shall drink T. Will my taking a decision one way or the other be rational too? Clearly not. For tomorrow I shall know perfectly well that before I act I shall be receiving relevant information about what the consequences of my drinking T would be - information which tomorrow I will not yet possess. And I shall know that it will be this information, and not any decision taken tomorrow, which, if I remain rational, will determine whether or not I drink T. How, knowing all this tomorrow as I shall, can it be rational for me to make up my mind then, in advance of the information, about whether I shall drink T? Decisions to do A cannot be taken by rational decision-makers who expect, before acting, to receive relevant information which they as yet lack; and who expect that it is this information which, irrespective of any earlier decision, will then determine whether, rationally, they should do A. So unless my rationality is compromised first, I will not tomorrow take a decision either way about whether to drink T - nor beforehand will I be motivated to take any such decision. But decisionmaking is a rationality-preserving agency: it cannot on its own motivate me to do what is irrational. Decision-making can only do that by way of causing some intervening action - an action by which I then make myself irrational. Hence I cannot take a decision one way or the other at will. If I could, perfectly rational decisions to make up my mind at a certain time - to take a decision one way or the other-could, of themselves, leave me motivated to make up my mind quite irrationally at that time. But then decision-making would not be a reason-applying agency: it would not be rationalitypreserving in its effects on the agent's subsequent motivation and agency. So we cannot make our minds up at will. A decision to make up one's mind will not directly motivate one to make one's mind up. It will not induce in one the motivation required for taking a decision. So, in one respect at least, decisions to make one's mind up must be exactly like decisions to take specific decisions. They must be decisions to act in a decision-causing way - and the motivation which they cause must be a motivation to act in a decision-causing way. If there is anything special about deciding to make up one's mind, then, it must have to do with the distinctive nature of the action decided upon.

The Regress argument

203

What action is that? The answer is not very difficult. Clearly, in making one's mind up one does two things. One reaches a decision; but prior to that one deliberates - one considers and weighs the various decision-relevant pros and cons. That deliberation then is the decision-causing action on which we decide when we decide to make up our minds, and which our decision to make our minds up leaves us motivated to perform. When one decides to make up one's mind about whether to do A, one surely decides to deliberate about whether to do A, and thereby reach a decision. It is then this action of deliberating, and not the initial decision to make up our minds, which, if anything ever does, eventually leaves us with the motivation which taking a definite decision about whether to do A requires - which then leaves us motivated actually to make our minds up. It is deliberation, and not the initial decision to make up our minds, which eventually leaves us, either with the motivation to decide to do A, or with the motivation to decide not to do A. Deliberation is a mental action which is importantly connected with decision-making. Our decisions are commonly caused by it. But notice that deliberation is nevertheless a first-order action and that makes it quite different from the second-order agency of decision-making which it explains. Why is deliberation not itself a case of second-order agency? Deliberation is first-order action rather than second, because deliberation does not itself constitute the formation of any particular action rationalising psychological attitude. To deliberate is not itself to form any particular intention, desire, or practical judgment. Rather deliberation is a typical cause of the formation of such psychological states — and can, of course, occur inconclusively, without any such states resulting. Unlike the actions - hypnosis, drug-taking and the like - by which I might cause myself to take a specific decision, deliberation is certainly not a form of self-manipulation. Deliberation is not an action by which one compromises one's continuing rationality. Quite the contrary, it is precisely an action by which one facilitates one's continuing rationality. Moreover, unlike manipulative strategies such as drug-taking and hypnosis, deliberation requires no special or recondite means. Deliberation is a mental action which is available to every decision maker - which every decision maker can perform at will. And that helps explain why decisions to make up one's mind, when taken, so reliably lead to their execution. The decision-

204

The psychology of freedom

causing action which we decide upon is precisely one which, in general, we can perform at will. But that is not the whole story of course. It is also important that the deliberation decided upon does frequently lead, in turn, to a definite decision being taken. Again, though, it is no accident that this is the case. For when one decides to make up one's mind at some given future time, one precisely picks a future time at which serious deliberation, motivated by a real desire to reach a definite decision then, would be likely to lead to a definite decision. One precisely picks a time at which a rational person - as one expects oneself to be - who deliberated seriously then, could and would reach a definite decision on the basis of that deliberation. One may of course pick wrong. Perhaps, at the time one deliberates, it becomes clear that one does not yet have all the information one expected to have. And so it may, after all, not yet be rational to make up one's mind. If that is the case, and one is rational, one will not actually make up one's mind. The final stage of the process - the stage of actually taking a decision - will not result, because one still lacks the motivation required. And that in turn will be precisely because one has deliberated. Far from motivating one to take a definite decision, in this case deliberating only ensures one's recognition of the irrationality of taking a decision so soon - and so helps prevent a decision yet being taken. Go back to the decision to make up one's mind about whether to drink T - the T that might be toxin or tonic, though we do not yet know which. Supposing one did take a decision to take the prizewinning decision about whether to drink T. Then to execute the decision, one would have to manipulate oneself into taking it. One would have to visit hypnotists and the like. For mere deliberation would not be enough to do the trick. And that is because to win the prize one has to make one's mind up at a time when, plainly, no definite decision about whether to drink T is yet rational. Just as deliberation can occur without resultant decision-making, so too decision-making can occur without prior deliberation. One can take a decision, one way or the other, without having to deliberate first. And when one does then, on my theory, one's decision cannot be the result of a prior decision to make one's mind up. For, I have claimed, such decisions are decisions to act - most commonly to deliberate; and when such they lead to their execution only by way of intervening deliberation.

The Regress argument

205

But what my theory predicts is independently plausible anyway. When do we take decisions without deliberating at all first? That happens when the circumstances leave it so obvious that we should take a decision and what decision we should take, that we simply do not need to deliberate first. In which case it is the obviousness of that decision which explains why we take it, and so why we come to a decision at all one way or the other - and not any prior decision to make up our minds then. If a robber convincingly threatens to kill me unless I give him my £10,1 do not need to deliberate before deciding to hand over the money. It is too obvious what I should do. And it is that obviousness of what I should do - that if I do not hand over the money I shall die for a petty £10 - and not any prior decision to make up my mind then, which explains why I come to an immediate decision about whether to comply. We began by thinking that, unlike taking specific decisions, taking a decision, one way or the other - making our minds up - is subject to the will. We see now that it is not. What initially looked like decision-making which is subject to the will has proved, on examination, to be decision-making which results from the will by way of action - by way of intervening deliberation. But if the will is to be reason-applying in its effects, that is just as well. A decision to make up my mind that led me to reach a definite decision, but which did so by by-passing any intervening consideration of the relevant pros and cons, would hardly be a rationality-preserving cause of that decision. The will, then, is a faculty for second-order, action-generating agency. And that is all that it is. It is not also faculty for third-order, will agency-generating agency. There is no such third-order agency of which we are capable. When we exercise the will, we always exercise it in the taking of decisions to act — in the forming of intentions to performjirst-order agency. And any further will agency generated by that intention formation, is generated in an actiondependent way, through that first-order agency. The will, I suspect, is never subject to the will. We have considered deciding to make up our minds - deciding to take a decision. What about another case where the will appears to be subject to the will - where decisions appear to influence decision-making in an action-independent way: what about deciding not to take a decision yet? That is easily dealt with: that is my reasons for deciding motivating me simply to refrain from

206

The psychology of freedom

taking any decision to act. In this case there is a perfectly economical explanation why I do not take any decision to act: my belief-desire constituted reasons for not yet taking any decision. We certainly do not have to postulate some special higher-order decision to which my decisions to act are subject. And so for any other case. Consider deciding to postpone making up my mind about whether to do A until next week. Such a decision to postpone making one's mind up is easily explained. I simply refrain from now taking any decision about whether to do A, deciding also to refrain from further deliberation about whether to do A until next week. Again, there is no need to postulate a higher-order decision with an action-independent effect on further decision-making. The truth here is simple. The will is not subject to the will. The agency of the will comes in, at most, its action-generating, second-order form. There is no directly will agency-motivating, third-order agency of the will. WHY MUST FREE ACTION BE SUBJECT TO THE WILL?

We have been considering how action is subject to the will, but decision-making is not. But, for the Regress argument to work, it must also be true that the will subjection of action is a condition of its freedom. Unless proponents of the Regress argument can establish this further claim, they cannot use the Close Analogy assumption to argue for the essential unfreedom of decisionmaking. So, must free action be subject to the will? In fact, the necessity of action's will subjection to its freedom has already been established. The point was proved by the argument in chapter 4 for the truth of Dependence - for why freedom of action depends on freedom of will. As I argued in that chapter, our having freedom of action at all depends on our having a future action control which is reason-applying. And it is decision-making which supplies us with the reason-applying future action control that a free agent needs. But decision-making provides us with this future action control only because action is subject to the will - only because by deciding to act we ensure that we are thereafter motivated to act as decided. If deciding to act did not leave us motivated to act as decided, decisions could not cause future actions at all; and then the will could not provide the reason-applying future action control which free agents need.

The Regress argument

207

My argument for why free action must be subject to the will is, then, my argument for Dependence - for why freedom of action depends on a prior freedom of will. But this raises an obvious question about Hobbes, the original proponent of the Regress argument whose premiss (i) we have just established. Hobbes also believed that freedom of action requires the will subjection of action. But Hobbes denied Dependence. So, of course, he would not have accepted my Dependence-based argument for premiss (i). Why, then did Hobbes believe that premiss? Premiss (i) makes the possession of control over how one acts depend on one's actions being explicable by prior intentions to perform them. Hobbes agreed - but only because action's actually being explained by the will was, in his view, essential to its character as action. For Hobbes, our actions just are those changes in us which are explained and motivated by prior intentions that they occur. And so if actions weren't subject to the will — if they were not thus explicable by it - they could not occur at all. Hobbes' belief in premiss (i) was based, then, on his own peculiar conception of the nature of action and of the will. Other philosophers hold that actual explanation by the will is essential, not necessarily to action's identity as action, but rather to its deliberateness or intentionality. In their view we cannot be acting deliberately or intentionally, unless our action is actually explained by some prior intention to act. On this view, action as such need not be a product of the will. But intentional action must be. Again, it is easy to see how this view of action intentionality would commit us to belief in premiss (i). It is a condition of having control over our actions that that control can be exercised. And one condition of our exercising control over our actions is surely that they be performed deliberately or intentionally. To be exercising control over how loudly I speak, for example, I must be speaking at a given volume intentionally or deliberately. But that means, on the view of action intentionality we are now considering, that for us to be exercising our action control, our action must actually be explained by a prior intention to act. Hence our possession of action control in the first place presupposes that our actions be so explicable - that they be subject to the will. So besides my own explanation of why premiss (i) is true - of why freedom of action presupposes action's subjection to the will - we

208

The psychology of freedom

have two others. Each defends premiss (i) by claiming that we cannot be exercising action control in the performance of action either we cannot be acting at all, or we cannot be acting intentionally - unless our action is actually explained by a prior intention to act. Now it is important that these two further accounts of why premiss (i) is true are both false. We can be exercising our freedom of action without our action actually being explained by any prior intention to act. Freedom of action presupposes, in its exercise as well as its possession, nothing more than the explicability of actions by prior intentions to perform them. Or so I shall eventually argue. My immediate task then is to show that my account of why free action must be subject to the will — the true account which appeals to Dependence - does not generalise to imply that free decisions, also, must be subject to the will. Then, before we can finish with the Regress argument, I need to return to these two alternative accounts of why free action must be subject to the will. I need to show that these two alternative accounts are anyway false - and for that reason cannot be used to establish the will subjection of free decision-making either. NEED FREE DECISIONS BE SUBJECT TO THE WILL?

If our freedom of action requires the will subjection of action, will not our freedom of decision require the will subjection of decision? The Close Analogy assumption predicts that since it is a condition on our freedom of action, will subjection must be a condition on the freedom of our agency generally. But we have seen that the will cannot be subject to the will. We cannot take decisions at will simply on the basis of deciding to take them. The Close Analogy assumption would lead us to conclude that, therefore, the will cannot be free. But the Close Analogy assumption is clearly dubious. In chapter 4 I argued that a free agent must have an executive control. She must have a capacity to apply practical reason in the exercise of her control over what she does. So practical reason is going to place important conditions on the freedom of any given agency. These will be conditions which derive from the need to apply practical reason as it concerns that agency. But that means that the Close Analogy assumption could easily be false. It could be

The Regress argument

209

false if the demands which practical reason places on the agency of our decision-making differ significantly from the demands it places on action. For then the capacity needed to apply reason as it concerns decision-making could be importantly different from the capacity needed to apply reason as it concerns action. Practical reason, as it concerns action, is plan-centred. It characteristically recommends a given action as a member of a sequence of actions - a sequence which extends through time into the future. So reason, when it recommends that we perform actions, often recommends that we perform them later rather than now. That is why, in applying those plan-centred recommendations of action, we have to be able to control action in advance. We need a reasonapplying future action control - a control which we can exercise just on the basis of our present deliberation. We have to be able to do something now which ensures that we perform a given action — but without yet performing that action itself. We have to be able to decide that we will perform the action in advance of performing it. As free agents, we need a deliberatively based future action control to apply practical reason as it concerns our actions. That is a control with which the will provides us - but only if action is subject to the will. So our freedom of action requires the will subjection of action. If we are to apply practical reason as it concerns our decisions, do we similarly need future decision control? Maybe we do - but, I shall now argue, only to a limited extent, and not in a way that would require the will subjection of our decisions. Decision-making, I have supposed, is an executive, reasonapplying agency. Its function is to help apply practical reason as it concerns action. So when practical reason recommends that we take decisions, this is characteristically because so doing will help us apply practical reason's recommendations of actions. Decisionmaking is an action co-ordinatory agency which is at the service of our action. Its function is to ensure that the actions which we perform at any given time are those which are justified given the actions which we perform at other times. As I claimed against Gauthier in chapter 6, action is the agency with which practical reason is primarily concerned. There is a practical primacy of action. It is this practical primacy of action which explains why we treat action as the paradigm of our agency - the paradigmatic status of action, I suggested earlier, being what lies behind the appeal of the

21 o

The psychology of freedom

Close Analogy assumption. We instinctively pattern our view of agency generally on our view of action in particular. And I doubt that we should so readily do this were action not the agency with which our practical reason is primarily concerned. However, it is precisely action's practical primacy which also raises questions about the Close Analogy assumption. For if reason's concern with the second-order agency of decision-making is secondary, then the capacities needed to apply reason as it concerns our decisions may not be exactly those needed to apply reason as it concerns actions. The practical reason I need to apply in exercising control over my decision-making is primarily concerned with action: it recommends decisions as action co-ordinators. So as a free decision maker I need only exercise control over my decisions in such a way as co-ordinates my actions over time. It follows that I shall only need to exercise future decision control if the exercise of such control is required effectively to co-ordinate action. To co-ordinate our actions we certainly need some future decision control. As we have already seen, we need to be able to control when and about what we make up our minds. We need to control the timing and order of our future decisions. That way we ensure that we only take decisions at times when we have enough of the information we need, and adequate opportunity for the deliberation required - but that we also make up our minds in time to execute whatever decision we end up taking. But that is a future decision control which, as we have also seen, we ordinarily have - and in a form which does not presuppose the will subjection of the will. What gives us this future decision control is our control over one future action in particular - our control of our future deliberation. By deciding now about what to deliberate and when, we exercise control over what in the future we take decisions about, and when. Granted, we cannot make our minds up at will. Decisions to make up our minds have to lead to their execution by way of intervening deliberation. So they lead to their execution only if deliberation at the relevant time could motivate a definite decision. And that deliberation will not do, in a rational agent at least, if reaching a decision at that future time would be irrational - if, for example, it would be rational to postpone any decision until more information had arrived. But if reaching a definite decision at a given future

The Regress argument

211

time would be irrational, why should we now make ourselves do it? The exercise of a deliberation-independent control over when we make up our minds would typically not be in the service of rational action. There is one form of future decision control which our future deliberation control will not ordinarily give us. And that is control over which specific future decisions we take. Controlling whether and when I deliberate about doing A may give me some control over whether and when I do reach some decision, one way or the other, about doing A. But it will not give me control over which specific decision I reach. The course and outcome of my future deliberation, and so which decision I reach on its basis, is going to be guided by the decision-relevant justifications available to me at the time. What those might be, and so which decision I shall reach as a result of considering them, is ordinarily both outside my control, and unknown to me in advance. We could regularly and reliably exercise control over which specific future decisions we took if, contrary to what I have so far been arguing, the will were subject to the will. Then, simply by deciding to take a specific future decision, I could leave myself motivated to decide exactly as I had decided - and so use decisionmaking to control which future decisions I take, just as I use it to control which future actions I perform. But does the co-ordination of our actions through time require the exercise of such future decision control? It is hard to see that it does. What action co-ordinatory work is done by taking a specific decision to act? Taking such a decision settles in advance how one will act by ensuring that thereafter one acts as decided: deciding to do A settles how one will act by ensuring that thereafter one does A. Now, assuming that the will were subject to the will - assuming that our decision-making of itself gave us control over our specific future decisions -what action coordinatory work would be done by taking a decision to decide to do A? Presumably, exactly the same action co-ordinatory work as a decision to do A — but done at one remove. One's decision to decide to do A would ensure that one thereafter decided to do A; and then that decision to do A would in turn ensure that one did A, thereby settling one's future action. But then why, at the outset, take a decision to decide to do A? Why not just decide to do A? Suppose action co-ordination requires that I settle now whether

212

The psychology of freedom

in the future I shall be doing A. If by decision-making I can settle the matter at all, I can perfectly well settle the matter by simply taking a decision to act. I need only decide, say, to do A. Decisions to decide to do A are quite superfluous to action co-ordinatory needs. It might, of course, be that no decision to do A taken now would affect how I act in the future. Perhaps the relevant action occurs too far into the future for my present decisions to act to affect whether I perform it. But if this is the case, it certainly will not help to postpone my decision to do A, deciding instead to take it at a later time. Why should present decisions to decide manage to determine future action at one remove, if present decisions to act could not determine it in any case? A ready and reliable control of which specific future decisions we took would indeed require that the will be subject to the will. But we certainly do not need that sort of future decision control to apply practical reason as it concerns action. We do not need it to co-ordinate our actions through time. So we do not need it to have freedom of decision either. It looks, then, as though we can have freedom of decision without the will subjection of the will. Pace Hobbes and Frankfurt, there is no reason to suppose that the conditions for freedom of will are closely analogous to those for freedom of action. Not every condition on freedom of action is also a condition on freedom of will. The Close Analogy assumption is false. HOBBES REDUCTIONIST REGRESS ARGUMENT

Hobbes supposed that free action must be subject to the will, and so explicable by an intention to perform it, because he thought that any action must actually be so explained. As Hobbes claimed: a Voluntary Act is that, which proceedeth from the Will, and no other. (Leviathan, p. 44)

Action, for Hobbes, is purposive. Any action is performed in order to further ends — ends which the agent desires to attain and which he believes his action will further. Now, in virtue of holding beliefs and desires which count as purposes for performing a particular action A, the agent counts as desiring to do A. And a willing of that action A is simply the possession of such a desire to do A which is strong enough to motivate and explain doing A. An action's being

The Regress argument

213

explained by the will to perform it comes, then, in Hobbes' view, to nothing more than its being purposive. Hobbes' belief in the will subjection of free action was based, then, on nothing more than his reductionist account of the will as an action-motivating form of desire. His belief that any free agency - including that of the will -would have to be subject to the will, was based on a combination of this reductionist account of the will with Hobbes' further belief that if action is purposive, then so too must be agency in general. Why did Hobbes suppose that decision-making - the will cannot itself be subject to the will? Hobbes failed to provide a very clear explanation. He certainly did not deploy my account of what limits the scope of the will. He did not appeal to the doctrine that the will must be rationality-preserving in its effects. Instead Hobbes simply dismissed the possibility that a particular decision to act be taken just on the basis of a prior decision to take it as 'absurd speech'. But there are two possible lines of thought which he might have been following. The first and most obvious line of thought is that were specific decisions directly explained by prior decisions to take them, this would give rise to a vicious causal regress. Suppose the explanation for a particular decision to act does ever lie in a prior decision to take it - then will not that prior decision to take it require exactly the same kind of explanation in turn? If we need to explain some decisions to act in terms of prior decisions to take them, will not this need arise at the level of the higher-order decisions also - and so on ad infinitum? In which case we arrive at the absurdity that, in order for a decision to occur, an infinitely long causal chain of prior and increasingly higher-order decisions must first have occurred. But maybe some decisions can arise out of prior and higher-order decision-making, without all decisions doing so. Why cannot just some decisions be explained by decisions to take them? Yet Hobbes' tone suggests that talk of any subjection of the will to the will is 'absurd speech'. There is a second line of thought - a line of thought which, I suspect, was really behind Hobbes' denial of the will subjection of decision-making. This line of thought would explain why there cannot be any - not even just some - decisions which are explained by prior decisions to take them. We have seen that the will — decision and intention — is constituted, in Hobbes' view, by desires

214

The psychology of freedom

to perform the actions willed: Hobbes was a Will-Desire reductionist. And, as we have also seen, Hobbes supposed that explanation by the will came to the same thing as purposiveness: action is explained by an intention to act in that it is explained by beliefdesire constituted purposes for acting. So, for the will ever to explain the will, desires to act - which on Hobbes' view constitute our will - would themselves have to be purposive. But, so Hobbes' line of thought might go, desiring is not something which may be purposive; desiring is not something which, like action, we may do in order to attain ends. The purposes which explain our desires to do A are purposes for doing A - not purposes for desiring to do A. (Hobbes was surely right on the last point at least, that desiring is not something we do purposively. Wanting things is not something we do in order to attain some end. Why cannot desire be purposive? In chapter 5 I argued that an occurrence may be purposive only if it is means-end justifiable. Now desire, we have agreed, is a pro attitude, and not a means-end justifiable occurrence. Desiring to do A is not justified by desirable ends which so desiring furthers. So if desiring is not means-end justifiable, then desires certainly cannot be purposive either.) If this line of thought was Hobbes', we see a second point at which his version of the Regress argument depended crucially on his WillDesire reduction. That reduction was the basis, not only for the claim that free action must be subject to the will, but also for Hobbes' denial that the will itself can be subject to the will. Hobbes simply combined his Will-Desire reduction with the plausible assumption that desiring is not itself a form of purposive agency. Desires cannot be purposive; so desires cannot arise just through being willed; so nor can the willings, the decisions and intentions, which those desires constitute. In chapter 2 I introduced a Reduction argument against freedom of will. This argument sought to establish that willing is not free precisely on the grounds that willing is a kind of desire - and that desiring is not a form of agency, and so not something which we can do freely. On our present interpretation, Hobbes' Regress argument against freedom of will is merely a variant form of this Reduction argument — a particularly debatable variant which assumes the Will Subjection theory of freedom. I suspect that Hobbes' disbelief in freedom of will was fundamentally driven by his

The Regress argument

215

reductive conception of the will - that the Regress argument, as Hobbes understood it, was just a version of the Reduction argument. We shall be returning to refute the Reduction argument in the next chapter. INTENTIONALITY AND THE EXERCISE OF CONTROL

Hobbes believed that any exercise of action control must be explained by the will. And he believed that, we have seen, because he believed that any action must be explained by the will - a theory of action which was in turn based on a reductive conception of the will as a form of desire. Now Hobbes' contemporary opponent Bramhall certainly did not share this reductive conception of the will; nor did he share the consequent assumption that action as such must be explained by the will. But Bramhall did nevertheless agree with Hobbes on one important point. Bramhall agreed that any actions over which we are exercising control - Tree acts' as he termed them - must be explained by the will: they must be 'elected upon deliberation', election being an act of the will.3 But since Bramhall denied any form of Will-Desire reduction, and did not at all accept Hobbes' psychology of the will, the grounds for his agreement on this point were quite different. They reflected an understanding of freedom and of psychology which were thoroughly scholastic. Indeed, we find similar views in Aquinas, to whose authority Bramhall made frequent appeal. In chapter 2 I argued that the source of an agent's control must be located in her agency. Our control of what happens derives from, and is exercised through, what we do. Bramhall, along with Aquinas, located the source of an agent's control much more specifically. They located the source of an agent's control in the agency of the agent's will. The will is, as Aquinas put it, 'the root of freedom, for that is where freedom lies'.4 Our control over our action derives from, and must be exercised through, the agency of the will in particular. Action over which an agent is exercising control must therefore be the effect of a prior willing of it. Now willing, for Bramhall - as for his scholastic predecessors -

3 4

See Bramhall 'A defence of true liberty', p. 49. Summa Theologiae, ia2ae, article 17, question 1, p. 185.

216

The psychology of freedom

certainly was not reducible to the desires that explain purposive action. As we saw in chapter i, Scholasticism held the will to be a distinctively rational appetite - a rational faculty which any free adult possesses, but which purposive though unfree agents such as children and animals lack. So there could perfectly well be purposive but unwilled action. Because unwilled, this would be action over which the agent failed to exercise control. Such action might include an action performed by a will-possessing adult in the heat of some non-rational passion - an action whose causes bypassed that agent's will; or it might include an action performed by an unfree will-less agent, such as a child or an animal. Such action, since not produced by the will, would be a product of the non-rational or sensitive appetite.5 There are modern philosophers who are committed to sharing the substance of BramhalPs and Aquinas' position. They would agree in opposing Hobbes' Will-Desire reductionism. Action as such need not be explained by any prior intention. But they would also agree - or at any rate are committed to agreeing - that action over which one is exercising control must be explained by prior intention. This commitment comes from a natural and popular account of action intentionality. We have seen that one condition of our exercising control over our actions is that our actions be deliberate or intentional. To be exercising control over how loudly I speak, I must be speaking at a given volume intentionally or deliberately. Now many modern philosophers hold that actual explanation by the will is a condition of action being deliberate or intentional - and so of the exercise of any control. We do A intentionally, on this view, only if we do A because of some intention - whether an intention to do A, or an intention directed at something appropriately related to doing A, such as trying to do A, or doing some B which one expects to lead one to do A. Bratman, to take one modern example, has espoused what he terms the Single Phenomenon View - the view that: 5

See Bramhall 'A defence of true liberty', pp. 37, 161; and Aquinas: 'There are degrees of being transformed by passion. It may go so far as to bind the reason completely . . . In this condition men become like the beasts, driven of necessity by passion; they are without the motion of reason, and, consequently, of will' (Summa Theologiae, question 10, article 3, PP- 9i-3)-

The Regress argument

217

intentional action and the state of intention both involve a certain common state, and it is the relation of an action to this state that makes that action intentional (Intention, Plans and Practical Reason, p. 112)

The state in question is, of course, that of intention - a state which explains the intentional action. There is some appeal to Bratman's position. For the intentionality of action certainly comes to more than mere purposiveness. There are what O'Shaughnessy has called sub-intentional actions: actions which the agent performs without doing anything deliberate or intentional - actions over which he is not therefore exercising control - but which, nevertheless are purposive.6 Nosepicking or scratching - these are actions which one can be performing purposively, in order to relieve discomfort for example, but without doing anything deliberate. One indication that the actions are wholly unintentional is precisely that they may be performed absent-mindedly. The agent need not even suspect that he is performing them. Sub-intentional actions are actions which, perhaps to one's embarrassment, one ca.nfind oneself vtriormmg. What would have to be added to make such actions intentional? It is extremely tempting to suggest that the required ingredient is explanation by intention. Only if the agent had been acting out of some intention — such as an intention to pick his nose — could he have been picking his nose deliberately. Perhaps, then, we have another explanation of why free action must be subject to, and so explicable by, the will. For us ever to be exercising control in its performance, our action must actually be explained by a prior intention. And that is because the exercise of action control must be intentional, and intentional action must be explained by a prior intention. But does intentional action really have to be explained by a prior intention? Let us now examine more closely the relation between intention and intentionality. INTENTION AND INTENTIONALITY

Not all philosophers claim that the intentionality or deliberateness of action implies explanation by intention. Davidson, for one, denies the implication: 6

See The Will, vol. 2, pp. 58-74.

218

The psychology of freedom

Of course, we perform many intentional actions without forming an intention to perform them, and often intentional action is not preceded by an intention. ('Intending', p. 88) Certainly, if decision-making is a case of agency - as we ordinarily suppose it is - then the intentionality of agency in general had better not imply its explanation by intention. For as second-order agency, decision-making can itself be intentional. Indeed, not only can it be intentional; it seems that it must be. For how can one take a particular decision to act, without doing so deliberately? Dennett, we have seen, has reported our natural intuition that decisions are the preeminently voluntary moves in our lives, the instants at which we exercise our agency to the fullest. And part at least of the truth behind that intuition is that decisionmaking is essentially deliberate. We cannot unintentionally take a particular decision to act. But we obviously cannot explain the intentionality of decisionmaking in terms of its causation by prior intention. Except where the product of decision-inducing action, our decisions are not explained by prior intentions to decide. Whatever form it takes, then, our account of intentionality should certainly make sense of something which very much militates against identifying the intentionality or deliberateness of what we do with its explanation by prior intention - the fact that decision-making itself seems to be essentially deliberate. If our first-order agency can easily take sub-intentional form, the same does not seem to be true of the second-order agency of the will. Any account of intentionality should account for why. On the other hand, there clearly is some connexion between the deliberateness or intentionality of our action, and its explanation by intention. It cannot be a coincidence that the term 'intention' is a constituent of another term applied to action and meaning deliberate. Moreover, whether or not explanation by intention is necessary to action intentionality, it is enough to guarantee it. If I perform a purposive action on the basis of an intention to perform it, then that makes the action's performance intentional. If I purposively twiddle my fingers out of an intention to do so, then I shall be twiddling my fingers intentionally. Why should there be these connexions between intention and

The Regress argument

219

action intentionality? The easy answer is precisely to follow Bratman, and identify an action's intentionality with its explanation by prior intention. But as we can now see, that answer may well be too easy. For as we have just noted, if taking a particular decision to act is itself something which we do deliberately, then we cannot generally identify the intentionality of agency with its explanation by prior intention. There certainly must be something about being explained by prior intention which can make agency intentional, and we need to discover what that is. But it may well turn out to be a factor that though it would be supplied by, does not also require, the presence of an agency-explanatory intention. Let us first look at the intentionality of action, and then turn to the intentionality of our second-order agency- the intentionality of decision-making and intention. The intentionality of action

What action's explanation by a prior intention does ordinarily guarantee, is that the action which I am performing is co-ordinated with the remainder of my agency. How does the intention ensure that? Intending (say) to twiddle my fingers of itself leaves me with a belief that I am or at least may well be twiddling them. And this belief allows me to base my remaining decision-making and action on the fact that my fingers are or may well be twiddled by me. In particular, the intention ensures that, at least in the typical case, I do not need to look, or otherwise perceptually attend to what I am doing and its effects, to find out about what it is that I am doing. Simply thanks to having intended to do it, I can know, or at least suspect, that twiddling my fingers is something that I am doing. Contrast sub-intentional action. It is characteristic of the subintentional, as we have seen, that through inattention the agent can easily be wholly unaware of it as he performs it. The psychological process which explains and gives rise to the action does not also explain and give rise to an attention-independent belief that one is or may well be performing the action. In a sub-intentional case of finger twiddling, there is no prior intention to twiddle one's fingers - and so no intention-based belief in their being twiddled. To be aware of what I am doing sub-intentionally, then, I shall need to be attending perceptually to what I am doing. And given the other demands on my attention at the time, that I may not happen to do.

220

The psychology of freedom

That is why sub-intentional finger twiddlings and scratchings are precisely things which I may only 'find myself doing' after I have been doing them for a time. And that means that, unless and until I do happen to notice myself doing it, my sub-intentional finger twiddling will not be co-ordinated with my remaining agency. Intentional action, then, is the effect of a process which, in generating the action, also generates a belief that one is or may well be performing it. Without having to attend perceptually to what one is doing, one is left aware that one is or might well be performing the action as one performs it. And this belief facilitates the co-ordination of the action with the rest of what one does. Now such awareness of one's action as one performs it, we have seen, is something which the action's explanation by a prior intention normally guarantees. As action co-ordinators, intentions must leave us believing that we will, or at least might well, act as intended. But though this attention-independent awareness of one's action as one performs it characteristically comes from a prior intention, there is no reason a priori why it need do so. In which case explanation by a prior intention to act, though characteristic of action intentionality, is not strictly necessary to it. For example, the requisite awareness could instead be generated just by one's trying to perform the action. Remember the general model of intention for which I have been arguing throughout the book. Forming an intention to act is a second-order, motivation-influencing action. It is an actiongenerating action by which one ensures a persisting motivation to act as intended. The intention formation stabilises and perpetuates one's present motivation into the future, thereby ensuring that one remains motivated to act as one now judges one should. And, of course, since forming an intention to act is essentially a way of influencing one's future motivation, it is something one will only do if it actually matters what one's motivation into the future will be. Intention formation or decision-making is something which one will only do in cases where one's performance of some desirable plan of action depends on one's future motivation. If this model of intention formation is correct, then there can perfectly well be cases of intentional action which are not preceded by prior intention. These are cases - albeit, I argued in chapter 3, very atypical cases - where it is desirable that one perform some action now, and where the desirability of the action is independent

The Regress argument

221

of any plan extending into the future. These are the relatively rare cases where the value of what one does now does not depend on one's future motivation. Here is an example. I am driving along and suddenly, rounding a bend, find a child crossing the road in front of me. As soon as I see the child, I judge that I should immediately step hard on the brake. Now let us suppose that stepping hard on the brake is an action which I shall perform just provided I now try to perform it. The action's performance does not depend on my motivation into the future. It will result directly, in a motivation-independent way, via nerve signals and muscle motions, from my now trying to perform it. And let us suppose that the value of stepping on the brake is pretty independent of any plan extending into the future. If I do not step on the brake now, I shall kill the child; if I do, then I will not and what matters is that I do not kill the child. Now if all this is the case - as it perfectly well might be - why, prior to my trying to step hard on the brake, need there be some additional, motivation-influencing agency which I have to perform first? Why, before I try to step on the brake, do I need first to decide or form any intention, be it an intention just to step on the brake, or to perform some wider plan which involves stepping on the brake as one stage? Since, ex hypothesis neither the action's performance nor its value depends on my future motivation, any such future motivation influencing will agency is entirely redundant. So it need not occur. But stepping on the brake can still be something which I do intentionally or deliberately. For my trying to step hard on the brake can perfectly well be what generates the awareness of what I am doing - that I am stepping hard on the brake - which action intentionality requires. Why have so many philosophers wanted to identify the intentionality of action with its explanation by prior intention? Part of the explanation is, I suspect, a common confusion - the confusion which, we saw at the end of chapter 2, gave rise to Locke's spurious category of mental volition. This is the confusion of trying with intention formation - the miscategorisation of tryings to act or conations as mental events of the same kind as motivationinfluencing intention formations or decisions to act. But tryings are not decidings. Deciding to step on the brake is the formation of an intention - a mental state of being decided. It leads to the action

222

The psychology of freedom

decided upon by way of ensuring a continuing motivation to perform that action. Whereas trying to step on a brake leads to the action attempted simply by causing the requisite nerve signals and muscle motions. Conations produce actions other than by causing the agent to be motivated to perform them. Once we remember to distinguish tryings from decisions, it is clearly going to be possible for some actions, and perfectly deliberate or intentional actions at that, to be the product of tryings without ipso facto being the product of decidings or intention formations. These will be actions resembling the action in our example: actions performed on the spur of the moment, and whose performance and value is independent of the agent's motivation into the future and of what that motivation might explain. Such actions may be atypical of our first-order agency. But they may still sometimes occur. And when they do, they may also occur in intentional or deliberate form. The intentionality of intention

The intentionality of action depends then on its being performed given an attention-independent awareness of its performance an awareness that a prior intention provides, but which, it seems, can also come from one's simply trying to act. What of the intentionality of intention formation or decision-making itself? Does the deliberateness of a decision also presuppose that one has some awareness of the decision as it is taken? I doubt in fact that such awareness is a requirement. We have already seen that not all conditions on the freedom of our action automatically translate into conditions on the freedom of our intention formation. And that is because of the differing ways in which practical reason is concerned with the two kinds of agency. Practical reason's concern with decision-making is secondary: it characteristically recommends decisions as action co-ordinators. There is a practical primacy of action. For this very same reason, as we shall now see, not all conditions on action intentionality automatically translate into conditions on intention intentionality either. Why does the intentionality of action require some intention- or attempt-based awareness of the action as one performs it? Such awareness, we have seen, is a belief condition on the co-ordinability

The Regress argument

223

of the action with the agent's remaining agency. It allows the agent performing the action to base his remaining agency on the fact of that action - thereby facilitating the agent's application of practical reason as it concerns his agency generally. Given that the awareness is directly intention- or attempt-based, it is no accident that, as the agent acts, he is left aware of what he is doing. The agent's awareness of what he is doing is not dependent on the chance of his happening to perceptually attend to and notice his action. So conditions on the intentionality of one's agency include conditions on one's beliefs about what one is doing - conditions on belief which have to be met if one is reliably to apply practical reason. Now, in the case of actions, it is a condition on their intentionality, that the agent be aware of them as he performs them. And that is because practical reason, as it governs action, is plan-centred. It requires the co-ordination of actions through time. And that, in turn, presupposes that the agent is reliably aware of his actions as he performs them. How can an agent co-ordinate his remaining actions with an action which at the time he is wholly unaware of performing? Now practical reason's concern with decision-making and intention, I have claimed, is secondary to its concern with action. Practical reason is characteristically concerned with intentions solely as action co-ordinators. And that makes a difference to conditions on the intentionality of decision-making and intention. The intentionality of our second-order agency depends, I suggest, only on its involving such awareness of what we do as facilitates action co-ordination. For that is the only awareness we need to apply practical reason. But what awareness is that? How does taking a decision to perform some action A facilitate action co-ordination? In two ways, as we have seen. The decision ensures, or at least raises the chance, that we shall do A. And it leaves us believing that we shall, or might well, do A. This belief then allows us to base our remaining agency on the fact that we shall, or might well, do A. We are then left in a position to perform those further actions which are justified given that we do A. The co-ordination-facilitatory awareness of our agency which decision-making provides, then, is an awareness of what we have decided to do. Provided deciding or intending to do A leaves the intender believing that he will or might well do A, it gives him

224

The psychology of freedom

precisely the awareness of his agency that action co-ordination requires. For, given that awareness, nothing more is needed for the agent to co-ordinate his remaining agency with the action decided upon. In particular, it is not necessary for the agent also to be left aware of his decision to perform that action. To base his remaining agency on the fact that he will or may well be doing A, the agent needs nothing more than an awareness that he will or may well be doing A. Action co-ordination only requires that agents have a decisionand intention-based, advance awareness of their actions. Provided they have that awareness, it does not strictly require that they be aware of the prior decisions and intentions to perform those actions as well. Intentional intention formation must, therefore, leave the agent aware of what is intended. But it is not necessary to the intentionality of the intention formation that it leave the intender aware of the intentions themselves. Of course, it may often be true that, unless one is aware of one's intentions as one forms them, one is unlikely to become aware of the actions intended. Often, perhaps even typically, I know that I shall or might perform some action which I have decided to perform only because I am aware that I have decided to perform it, and that this decision will or might lead me to perform that action. But that does not mean that awareness of one's intentions is itself a condition on their intentionality. One's awareness of one's intentions is merely something without which, at least in many cases, what is a condition on intention intentionality will not be met — that one's intentions give one advance awareness of what is intended. I noted earlier that although sub-intentional action seems an evident possibility, sub-intentional decision-making is much less obviously possible. I cannot take a decision, it seems, without doing so deliberately. We can now explain why. It is a condition on an agent's forming an intention to do A that he be left believing that he will, or at least might, do A. Intention formation has, at the very least, an action co-ordinatory function a function which it cannot fulfil unless it generates sufficient belief in the action intended. But meeting this condition on the intention's being formed at all, is also enough, as we have just seen, to guarantee the intentionality of the intention. Decision-making - indeed intention formation generally - so reliably counts as something deliberate simply because the very conditions that are

The Regress argument

225

required for intention formation to occur at all, are sufficient for that intention formation also to count as intentional. CONCLUSION

The Regress argument assumes that any free agency must be subject to the will - but that decision-making, in particular, is not subject to the will. The Regress argument is certainly correct on the second point. In this chapter, I have explained why decision-making is not subject to the will as action is. Indeed, I have explained why none of intentions, practical judgments and desires can be subject to the will; and why anything which is subject to the will must be agency something we might do or refrain from doing. To explain all this, I appealed to the idea that the will is an executive, reason-applying faculty - a faculty which serves to facilitate the rationality of those occurrences subject to it. The will's influence on what is subject to it must in particular be rationality-preserving. Where doing A is subject to the will - where deciding to do A directly causes us to be motivated to do A - the rationality of deciding to do A must guarantee the rationality of doing A. And that means that in respect of doing A, REASON-APPLY must hold. Any ends which justify deciding to do A must ipso facto justify doing A thereafter. We have seen that, in general, whatever doing A might be, the fact that doing A would further desirable ends can always justify deciding to do A. That performing an action, or taking a decision, or holding a desire, would bring huge rewards, can always justify deciding to perform that action, or to take that decision, or to form that desire. It follows that for doing A to be subject to the will, those ends which could always justify deciding to do A - any desirable ends which doing A would further - must ipso facto justify doing A thereafter. To be subject to the will, then, doing A must constitute means-end justifiable agency which is not special-purpose. Justification for doing A must be provided by any desirable end which doing A would further. Excluded from subjection to the will, then, are non-means-end justifiable pro attitudes, such as practical judgments and desires. But also excluded from will subjection are decisions. Decisions are agency-motivating occurrences; and, ex hypothesis ends can only

226

The psychology of freedom

justify our taking decisions if they would justify thereafter whatever agency it is which those decisions would directly motivate - whatever agency it is which is subject to those decisions. That means that even if decisions do constitute means-end justifiable agency, they must constitute a form of agency which is special-purpose. Decisions cannot be justified by any old desirable ends which they would themselves further - but only by those ends which would thereafter be furthered by, and so justify, whatever agency it is which those decisions directly motivate. Decisions in particular, then, cannot be subject to the will. This account of why decisions, along with pro attitudes, are not subject to the will is entirely built on the assumption that the will is indeed an executive faculty — that the will does serve to facilitate the rationality of any occurrence, such as action, that is subject to it. It is crucially assumed that doing A is subject to the will only if REASON-APPLY holds with respect to doing A - only if justifications for deciding to do A do ipsofacto justify doing A thereafter. The idea that the will has an executive, reason-applying function in relation to whatever is subject to it - an idea that, I have suggested, lies at the heart of the common-sense psychology of the will - turns out, then, to explain another important feature of common-sense psychology. It helps explains all our intuitions about what can or cannot be subject to the will. The Regress argument, then, is in my view right about the will's scope. Decision-making is not subject to the will. But the Regress argument goes wrong in what it infers from that fact. For though action's subjection to the will is a condition of freedom of action, decision-making's subjection to the will is not a condition of its freedom. And that is again because the will's function in relation to what is subject to it - in effect our first-order action - is executive and reason-applying. The reason why free action must be subject to the will lies in the fact that, as I argued in chapter 4, freedom involves a capacity for rational self-determination. The freedom of any given kind of agency depends on our having a capacity to exercise a reasonapplying, executive control over it. Now action is governed by reason in a plan-centred way. For this reason, our executive control over action must extend into the future — a control which we exercise through free, reason-applying decisions to which our future actions are subject. We need a purely

The Regress argument

227

will-based future action control to apply reason as it concerns our action. That is why free action must be subject to the will. But we do not need an analogously will-based future decision control to apply reason as it concerns our decision-making. So free decisions need not be subject to it. And that is precisely because decisionmaking is an executive activity, with the function of applying reason as it governs what is subject to the will - our subsequent action. Practical reason's concern with our decision-making, accordingly, is secondary to its concern with action. Practical reason characteristically recommends decisions, as action co-ordinators, to facilitate the application of reason as it concerns action. And using decisions as action-co-ordinators, I have argued, always involves taking decisions to act, to which our actions and not our decisions are subject. The executive, reason-applying function of the will in relation to what it explains and motivates is quite central, then, to a proper understanding of the Regress argument. At one point, in rightly denying that decisions and intentions themselves are subject to the will, the argument exploits a genuine implication of the will's executive function - that the will's influence on what is subject to it must be rationality-preserving. But then at another crucial point in the Regress argument, the will's executive function is simply ignored. The argument crucially assumes that conditions on freedom of decision are closely analogous to conditions on freedom of action. But they are not; and once more this is so precisely because decision-making is an executive, reason-applying activity; an activity which serves the rationality of what is subject to it - namely, our first-order action.

CHAPTER 8

In defence of the Action model

DECISIONS AS REASON-APPLYING ACTIONS

Can decision-making really be what, as users of common-sense psychology, we ordinarily conceive it to be - a reason-applying, second-order action? As chapter 5 showed, there is a tension between the idea that the will serves an executive function in relation to first-order action, and the idea that the will is itself a locus of agency. If decisions serve to apply reason as it concerns subsequent action, can decisions to act be actions themselves? If we adopt the Action model of decision rationality, we can plausibly establish that decision-making is a form of agency. The Action model says, after all, that decisions are means-end justifiable, just like the actions which they explain. Decisions to act are justified in terms of desirable ends which taking them would further - just as actions are justified in terms of desirable ends which performing them would further. And that means that agents can rationally take decisions as means to attaining desirable ends, in the same way that they can rationally perform actions as means to desirable ends. The Action model includes decisions within the same important class as the actions which decisions explain the class of events governed by what looks very much like reason in its distinctively practical form. But then, as we have seen, we face the problem of making sense of decision-making as a reason-applying activity. Rationally taken decisions to act must leave us disposed to act rationally thereafter. So we have to explain why, though decisions to act are means—end justifiable, there should be a special restriction, imposed by REASON-APPLY, on the ends that could justify taking them — a restriction which, it seems, does not apply to our means-end justifiable first-order actions. Decisions to act can be justified only 228

In defence of the Action model

229

by ends which would thereafter justify acting as decided. So decisions cannot be justified by prizes offered simply for taking them. Whereas prizes offered just for performing a first-order action, such as taking a decision drug, could perfectly well justify performing that action. If means-end justifiable at all, decisions to act must be distinguished from the actions which they explain by being what I have termed a form of specialpurpose agency. We need to show how the Action model is fully consistent with REASON-APPLY - why the addition of REASONAPPLY to the model is not simply ad hoc - otherwise we cannot do justice to the executive function of the will within the Action model's terms. The alternative Pro Attitude model of decision rationality looks far more consistent with the will's executive function. It seems to have the truth of REASON-APPLY built in. For the Pro Attitude model says that our justifications for taking a decision to do A are provided, purely and simply, by our justifications for doing A. Decisions can only be justified, therefore, by ends which already justify performing the action decided upon. The more obvious problem, with the Pro Attitude model, is making sense of the agency of the will. We have to show how, consistently with the Pro Attitude model, taking a particular decision to act is something that we deliberately do. And that means finding a plausible replacement for the highly intuitive Means-End model of agency. Means-end justifiability can no longer constitute reason in its distinctively practical form. We shall have to appeal instead to other accounts of how reason is distinctively practical, such as the Autonomy-within-Reason model or the Plural-Justification model - or some further model of practical reason and agency as yet unformulated. ARGUING FOR THE ACTION MODEL

In this chapter, I shall prove that the Action model of decision rationality is true. In so doing, I shall prove that there is a secondorder agency of the will, refuting the remaining arguments for scepticism about second-order agency. It is false that justifications for deciding to do A are, simply, justifications for doing A, as the Pro Attitude model claims them to be. Decisions to act are means-end justifiable occurrences, and so they are actions themselves. My

230

The psychology of freedom

method of proof is simple - and of a kind absolutely fatal to the Pro Attitude model. The Pro Attitude model looks as plausible as it does precisely because it appears straightforwardly to accommodate our natural belief that decision-making is a reason-applying activity. Why else believe the Pro Attitude model, which explains the rationality of decisions to act directly in terms of the rationality of the actions decided upon, than because decisions to act have the function of applying reason as it governs our subsequent action? It is very difficult to give up the idea that decision-making is indeed about applying reason as it concerns subsequent action that the point of taking decisions about which actions to perform is precisely to ensure that we end up performing the right actions. So arguments for the Action model which deny or ignore decisionmaking's reason-applying function will continue to appear unconvincing to many of us - rightly enough, in my view. What really must discredit the Pro Attitude model, though, are arguments for the Action model which actually assume and build on decisionmaking's reason-applying function. And these are the arguments which I shall be producing. The Pro Attitude model, this chapter will show, is actually inconsistent with our belief that decision-making serves to apply reason as it concerns subsequent action. And that is because the Pro Attitude model cannot accommodate the rational use of decisions to co-ordinate subsequent action. The action co-ordinatory function of the will, I shall argue, implies that decisions to act are means-end justifiable occurrences themselves, just as are the actions which they explain. The reason-applying use of decision-making, then, presupposes the Action model of decision rationality. Proving the Action model still leaves work for us to do. For if decision-making is a reason-applying activity, as we naturally suppose it to be, REASON-APPLY must be true. And we have yet to show that REASON-APPLY really is consistent with the Action model of decision rationality. Proving that the Action model and REASON-APPLY are fully consistent is the task for the last chapter of the book. With that we shall have established the coherence of the common-sense psychology of the will in its entirety. We shall have vindicated the psychology of our freedom.

In defence of the Action model

231

DECISION-MAKING AND EFFECTIVE ACTION CO-ORDINATION

The Pro Attitude model contains an Identity thesis explaining justifications for deciding to do A in terms of justifications for later doing A: J-IDENTITY

our justifications at t for then deciding or intending to do A are identical with our justifications at t for later doing A.

What are our justifications at t for later doing A? The Pro Attitude model, quite rightly, assumes that JUSTIFY is true: JUSTIFY

Any justification at t for later doing A consists in the likelihood at t that later doing A would further a desirable end E.

It follows that, according to the Pro Attitude model, any justification at t for then deciding or intending to do A consists in the likelihood at t that later doing A would further a desirable end E. By contrast, the Action model includes decision-making along with action in the range of agency governed by JUSTIFY. It claims that ACTION is true: ACTION

Any justification at t for then deciding or intending to do A consists in the likelihood at t that then deciding or intending to do A would further a desirable end E.

To falsify the Pro Attitude model and help establish the Action model, we have to find a desirable end E which clearly justifies now deciding to do A - but which, though consistent with ACTION, provides a counter-example to J-IDENTITY. SO the end E which justifies now deciding to do A, though likely to be furthered by that decision, must not, at least prior to a decision to do A being taken, be an end which doing A would be likely to further. The end must justify now deciding to do A - but without yet providing justification for the doing of A. Decision-making is an agency with an action co-ordinatory function. And it is when we consider action co-ordination that we find clear counter-examples to the Pro Attitude model. These are precisely the cases which we need - cases where the agent has a justification for deciding to do A which, prior to his taking that

232

The psychology of freedom

decision, is not a justification for doing A. The decision-justifying end in question is action co-ordinatory. It is the end that the actions which one performs at one time be justified given the actions which one performs at other times. It is the end that practical reason be consistently applied in one's actions through time. It might be thought that the mere fact that decision-making has an action co-ordinatory function immediately commits us to the Action model of decision rationality. Decisions to act settle and warn us in advance exactly how we shall be acting - a desirable end which, obviously enough, would not be furthered by performing the actions decided upon alone. If deciding on our actions in advance benefits us in this way, thus enabling us to co-ordinate our actions through time, then will not those benefits serve to justify the decisions to act which bring them? In which case it follows that decisions are indeed means-end justifiable occurrences - that our decisions to act are justified by the action co-ordination which they facilitate. But this argument is far too quick, and defenders of the Pro Attitude model have an obvious and immediate reply to it. The Action and Pro Attitude models are theories of what justifies taking particular decisions to act - of what justifies deciding specifically to do A. Now it may be that making up our minds in advance about whether to do A — bothering to take a decision at all one way or the other - is justified, in part at least, by the action co-ordinatory benefits it brings. But it does not follow that deciding specifically to do A is means-end justifiable too. And supporters of the Pro Attitude model will claim that it is not. Given that we make our minds up at all, what justifies deciding to do A in particular is, simply, what justifies doing A — the desirable ends which doing A would further. Our Psychologising conception of our freedom assumes that the taking of particular decisions to act - decisions specifically to do A rather than B - is itself a deliberate action. It is up to us which actions we perform, we think, only because it is up to us which actions we decide to perform. But it can be up to us which actions we decide to perform only if the taking of particular decisions to act is something that we can deliberately do. To establish that, we need to show that the taking of a particular decision to act - of a decision specifically to do A - is itself governed by reason in its practical or agency-governing form. We need to show that the taking of a

In defence of the Action model

233

particular decision to act is a means-end justifiable occurrence, just as the Action model teaches. To defend the Action model, then, I need to show that we can have an action co-ordinatory justification, not just for making up our minds one way or the other, but for taking a specific decision for taking a decision to do A in particular. So how is it possible to have action co-ordinatory justification for taking a particular decision to act? It is easy enough to show how. Decisions fulfil their action co-ordinatory function by causing the agent to act as decided, thereby settling in advance how he will act. The agent is left knowing how he will act in the future, so allowing him to coordinate his actions in the present with that future action decided upon. As a result, the agent's actions in the present are reliably those that are justified given his actions in the future - and vice versa. But not all decisions are equally likely to cause the action decided upon - or even an attempt at its performance. For decisions affect future action by affecting intervening action motivation. And some decisions are much more likely to cause a persisting motivation to act as decided than are others. That means that those decisions are much more likely to settle in advance what we shall be doing, than are the others. And this fact is very relevant to the justification we have for taking the decisions. That some decisions are more likely than their alternatives to settle in advance how we shall be acting may provide us with some justification for taking those decisions in particular. It may be important that whatever action we decide on be prepared for in advance; and it may also be important that our preparations not be wasted - that our preparations be followed by the action decided upon and prepared for. If so, then it will correspondingly matter that any decision we now take settle firmly how we shall act - that, in particular, there be no risk of the decision's being abandoned through a subsequent change of mind before we can perform the action decided upon. In which case, it may be a significant justification for taking a given decision, that that decision would be the decision most likely to prevent a subsequent change of mind - that that decision, accordingly, would settle how we will be acting in the future more firmly than would any alternative. Depending on how important it is to settle definitely how we shall act, and depending how much more firmly than its

234

The psychology of freedom

alternatives the decision would settle how we shall act, that justification may even override strong justifications for not performing the action decided upon. Action co-ordinatory considerations may lead us, quite rationally, to decide to perform actions which, prior to our decision, we had insufficient justification for performing. The Pro Attitude model explains decision justifications by simple reference to ends furthered by the action decided upon. Hence the Pro Attitude model entirely ignores the possibility that a decision, once taken, might be abandoned before that action could occur. It entirely ignores the likely co-ordinatory effectiveness of decisions in its account of the justifications which we have for deciding. And this, in a model of the rationality of an action co-ordinatory agency, is a fatal omission.1 DANS DECISION

Consider the case of Dan, the risk-averse stuntman. Dan is a stuntman who needs to decide now whether in six months' time he will attempt a particular stunt or not. This is because he has to organise any publicity for the stunt soon. In order to decide about publicity he needs to decide about the stunt, since he wants to avoid attempting the stunt without having organised publicity (a wasted risk) and not attempting the stunt having organised publicity (which would make him look very foolish). Either of these two most disliked eventualities would constitute a mismatch or miscoordinated pair of actions - a pair of actions at least one member of which is not justified given performance of the other. There are only two plans of action which it is worth Dan's while considering. Attempting the stunt having first organised publicity for it, and refraining from such an attempt having first refrained from organising publicity. We need to determine, in what follows, what justifications Dan might have for and against now deciding to attempt the stunt. 1

As I have presented it, Gauthier's indirectly Teleological theory of decision rationality faces exactly the same objection. This theory justifies decisions in terms of the ends furthered by whole Gauthier-plans - in terms of the ends furthered by the combination of the decision to act with the action decided upon. The theory, then, ignores the crucial issue of whether, if a given decision were taken, the Gauthier-plan that contains it would actually occur. And so the theory says nothing at all about how the rationality of taking a decision might depend on the decision's likely effectiveness as an action co-ordinator.

In defence of the Action model

235

We can assume that Dan has already decided not to attempt the stunt without organising publicity, and not to organise publicity without attempting the stunt. Each of attempting the stunt and organising publicity is to be valued only as a part of one and the same plan. So if Dan is rational, he will not decide on either one of the actions without deciding on the entire plan. He will not decide to attempt the stunt without at the same time also deciding to organise publicity - and will not decide to organise publicity without at the same time also deciding to attempt the stunt. The stunt will be risky. Because of the risks, it may well be that there is more justification for not attempting the stunt and not organising publicity for it, than there is for attempting the stunt having organised publicity for it. Let us suppose that this is the case. Does it follow that Dan has more justification for deciding not to attempt the stunt? It does not follow. We have to consider the likely efficacy of the decision as an action co-ordinator. Dan has just had a nasty accident, and he is at the moment riskaverse. Were he sure that whatever he now decided to do he would do, he would now decide as his justifications for action dictate. Dan would decide not to attempt the stunt. For Dan quite rationally holds an overall preference not to attempt the stunt and not to organise publicity for its performance. Not attempting the stunt and not organising publicity for a stunt is the preferred as well as the preferable or more justified plan. Dan is well aware, however, that within the next six months his risk aversion may wear off. There is a significant possibility (though no more) that even if he initially decided not to attempt the stunt, so that no publicity was organised, Dan might still end up preferring to attempt the stunt and attempting it. Now whatever decision concerning the stunt Dan takes today, he wants it to be one that will stick. In particular, he does not want a change of mind to occur after the opportunity to organise - or prevent - publicity has passed. Such a change of mind would just produce what Dan particularly wants to avoid - a mismatch. So there is one justification for now deciding to attempt the stunt. This is the desirability of avoiding a mismatch - a pair of misco-ordinated actions. Dan believes that a decision to attempt the stunt is less likely to be abandoned through change of mind than a decision not to attempt the stunt. After all, Dan believes, he is going to become less rather than more careful. Hence, Dan

236

The psychology of freedom

believes, deciding to attempt the stunt is more likely to avoid a mismatch than deciding not to attempt the stunt. Notice that, prior to his decision, the desirability of avoiding a mismatch does not provide Dan with any justification for attempting the stunt. For, until Dan decides about whether to attempt the stunt, and so about whether he will organise publicity, neither of attempting the stunt or not attempting the stunt need be the more likely to avoid a mismatch. Which action would avoid a mismatch plainly depends on whether Dan first organises publicity for a stunt; and that in turn depends on whether Dan now decides to attempt the stunt. But prior to actually making up his mind, Dan need not be in any position to know which of the two decisions about the stunt he is now most likely to take. It may be a nice matter, whether the risk of misco-ordinated action arising from now deciding not to attempt the stunt is large enough to offset the dangers of attempting the stunt - dangers which are risked by both decisions but which, obviously enough, are more likely to arise from a decision to attempt the stunt. Prior to making up his mind Dan may, as a rational agent, feel evenly torn between both decisions. He will not know or be in any position to know which decision he is more likely to take. It is clear that J-IDENTITY must be false. By deciding to attempt the stunt Dan would assure himself of the performance of a consistent plan - a publicised attempt at the stunt - and so be sure to avoid the mismatched outcomes. The desirability of avoiding a mismatch therefore constitutes a justification for Dan's now deciding to attempt the stunt - a justification which, however, does not at the same time justify later attempting the stunt. Of course, once the decision to attempt the stunt is taken, Dan's justifications for attempting the stunt change. Once Dan decides to attempt the stunt, he is in a position to know that he will be organising publicity for it. So it is now true, when it wasn't true before, that attempting the stunt would be most likely to prevent a mismatch. The action co-ordinatory end which, prior to Dan's decision, justified his now deciding to attempt the stunt provides once that decision has been taken — at least as much justification for performing the action decided upon. In chapter 5 I argued that with regard to any action A, an acceptable theory of decision-making must entail:

In defence of the Action model REASON-APPLY

237

any end E that justifies deciding to do A must, supposing that decision is taken, also provide at least as much justification for doing A.

For, unless REASON-APPLY is true, decision-making will not exercise a rationality-preserving influence on future action. Unless REASON-APPLY is true, rational decisions to do A will not be guaranteed to leave us disposed to do A rationally thereafter. Notice that action co-ordinatory ends conform to this condition. Action co-ordinatory ends provide justifications for a decision to act which need not, prior to that decision being taken, be justifications for acting as decided. However, once a decision to act has been taken, action co-ordinatory ends will only be furthered if the agent does go on to act as decided. Action co-ordinatory ends are decision-justifying ends which conform to REASON-APPLY.

A FURTHER VERSION OF DANS DECISION

In the example just given, Dan's decision to attempt the stunt is taken at the same time as another, co-ordinated decision - a decision to organise publicity. Notice that this second, contemporaneous decision is not at all necessary to the case. Dan's co-ordination problem can easily arise without it; and taking a decision to attempt the stunt can still provide exactly the same rational solution. Dan may need to settle now whether he will attempt the stunt later in case he soon has to take further decisions which depend on whether he attempts the stunt. Dan will still need to avoid mismatches - attempting the stunt having acted in a way only justified given that he does not attempt the stunt, or not attempting the stunt having acted in a way only justified given that he does attempt the stunt. It can still be important, therefore, that any decision Dan now takes about attempting the stunt should stick. For what Dan now decides about the stunt might well influence what further actions he performs in the nearer future. Any change of mind about the stunt after these further actions had been performed would, therefore, again produce the sort of mismatch Dan needs to avoid. So in this case too, if a decision to attempt the stunt is the

238

The psychology of freedom

decision least likely to be abandoned in a subsequent change of mind, the need to co-ordinate actions through time can still justify taking it rather than any other. It does not matter that Dan has not actually yet co-ordinated whether he attempts the stunt with doing anything else. He may soon need to do so. If Dan does take a decision to attempt the stunt, he will be quite rationally using the second-order agency of his will to settle, as firmly as he can, whether or not he will be performing some given first-order action in the future. The matter settled, he will then have a knowledge of his future action which he can deploy as input to practical deliberation about further matters if and when he needs to - a knowledge of what he will be doing which facilitates action co-ordination, and which he would not have possessed had he not taken whichever particular decision to act was most likely to leave him with a persisting motivation to execute it. As before - by deciding to attempt the stunt Dan will acquire the same action co-ordinatory justification for attempting the stunt as beforehand he had for deciding to attempt it. Once Dan takes a decision to attempt the stunt, it will come to be true, when it was not true before, that attempting the stunt would be most likely to avoid a mismatch. GO-ORDINATION AND JUSTIFICATION CHANGE

Dan's decisions are counter-examples to J-IDENTITY. What makes them counter-examples is the fact that an end justifies deciding to perform an action - to attempt the stunt or do A - which does not initially justify the action - doing A. However, this end is a decisionjustifying end which, as decision justifications must, conforms to REASON-APPLY. So, once a decision to do A is taken, the end then comes to provide the same justification for doing A which, beforehand, it provided for deciding to do A. It follows that deciding to do A increases Dan's justification for doing A: thanks to his decision, the agent comes to have a justification for doing A which he didn't have previously. But this is not because there is anything intrinsically desirable about executing a decision which one has taken. It is no more intrinsically desirable to execute a decision than it is to execute a decision drug. The increase in Dan's justification for acting as decided is

In defence of the Action model

239

to be explained entirely in terms of the end which justified his original decision. This is the desirable end that Dan's actions be co-ordinated through time. What constitutes the attainment of this end depends on which actions co-ordinate with which - a matter which is quite independent of what decisions Dan actually takes, and so has nothing to do with the supposed intrinsic desirability of decision execution. Dan's decision to do A simply makes it true, when it was not true before, that doing A would be likely to produce a co-ordinated outcome. We have already seen that Dan's decisions falsify J-IDENTITY because ends can justify deciding to do A without at the same time justifying doing A. But if deciding to do A increases an agent's justification for doing A, it follows that there is a second way in which Dan's decisions are counter-examples to J-IDENTITY. J-IDENTITY explains justifications for now deciding or intending to do A in terms of justifications for doing A. So it predicts that any increase in Dan's justification for doing A ipso facto counts as increase in his justification for now intending to do A. The increased justification which Dan's decision gives him for attempting the stunt ipso facto counts also as increased justification for his now intending to attempt it. But what J-IDENTITY entails is false. Dan's decision is not a magically self-justifying activity. When Dan forms his intention to attempt the stunt, he certainly adds to his justification for attempting it. We have seen exactly why. Attempting the stunt is more likely to ensure the co-ordination of Dan's agency over time. But that is only because, and for as long as, Dan continues to intend to attempt the stunt. So the added justification provided for the action is entirely conditional on, and presupposes, the fact that Dan now holds the intention. It certainly does not count as added justification for the intention as well. Hence we should reject J-IDENTITY.

In fact it is true in most cases that deciding to do A will add to our justifications for doing A - but without thereby adding to our justification for deciding or intending to do A. This means that counter-examples to the Pro Attitude model are not limited to cases like Dan's. We do not need to consider only those cases where an agent is unsure of the effect of his decisions on his future motivation, and so of the co-ordinatory efficacy of his decisions.

240

The psychology of freedom

We saw in chapter 4 why deciding to do A typically adds to our justifications for doing A. As I argued there, one way of looking at decision-making in advance of action is in terms of betting. By deciding on my actions in advance, I effectively increase the likely stake on a bet that I shall perform the act decided on. Deciding to do A in advance of doing A increases the likely stake on A's being done: it increases the likely benefit from doing A, and it increases the likely loss from not doing A. And this is because decisions have an action co-ordinatory effect. They make it more likely that before doing A, the agent will have performed other actions on the assumption that he may or will do A — actions which tend to increase the benefit the agent derives from doing A. On the other hand, such actions tend to be ones which the agent will regret performing unless, in the end, he does A. Deciding to go on holiday in France makes me devote time and money to booking hotel rooms in France - time and money which will add to the enjoyment of holidaying in France, but which will be quite wasted if I do not actually go. So deciding now, rather than later or not at all, to do A typically increases what an agent is likely to gain were he to do A - and to increase what the agent is likely to lose were he not to do A. So (i) is often true: (i) Now intending to do A later increases an agent's justification for doing A later. J-IDENTITY identifies justifications for now intending to do A with justifications for later doing A. So J-IDENTITY predicts that whenever (i) is true, (ii) is true too:

(ii) Now intending to do A later increases an agent's justification for now intending to do A later. But it cannot be right to claim that (ii) is true whenever (i) is true. The claim was obviously false in Dan's case and it is no more plausible for other cases. We already have an explanation of how (ii) can be false when (i) is true. Consider again why (i) is so often true. It is so often true because deciding to do A changes what ends later doing A would be likely to further. The decision changes the likely stake on the agent's doing A. Deciding to do A changes the likely effects of doing A or of not doing A. For example, the decision makes it

In defence of the Action model

241

more likely that not doing A later would frustrate actions which the agent had performed earlier on the assumption that he would do A. So why do we suppose in such cases that (ii) may be false - that the agent's justifications for now intending to do A will not have changed just because his justifications for doing A have changed? The reason is clear. Forming the intention to do A changes the likely effects of later doing A or not. But it will not change the likely effects of the agent's now intending to do A or not. So, in particular, it will not change the likelihood that now intending to do A would get him subsequently to decide and act on the assumption that he will do A. It is, after all, this unchanged likelihood which, given a decision to do A, explains the change in the likely effects of doing A or not - the increase in what the agent expects to have staked on his doing A. The likely effects of my now intending or not to go on holiday do not change just because I form that intention. What changes are the likely effects of my later going or not going on holiday. Once I have formed an intention to go on holiday, it becomes likely that I shall act in the near future on the assumption that I shall go - so that I shall face possible financial loss or similar bad consequences if I do not in fact go. Likewise the likely effects of Dan's now intending or not to attempt the stunt do not change just because he forms that intention. What changes are the likely effects of Dan's attempting or not attempting the stunt. Once Dan has decided to attempt the stunt, it becomes more likely that he will act in the near future on the assumption that he will attempt the stunt - so that Dan will suffer possible loss of face or similar bad consequences if he does not in fact make the attempt. So we can see what justifications for deciding or intending to do A are not - they are not desirable ends which doing A would be likely to further. The ends which doing A would be likely to further can change, with concomitant change in justifications for doing A, without any change in our justifications for deciding or intending to do A. At the same time we can also see what justifications for deciding to do A are. They are what need not change simply because our justifications for doing A have changed: they are desirable ends which deciding or intending to do A would be likely to further. Decision-making is a means-end justifiable activity, ACTION and the Action model are true.

242

The psychology of freedom IS R-IDENTITY TRUE?

As well as explaining decision justifications in terms of action justifications, the Pro Attitude model explains reasons for deciding in terms of reasons for action. The Pro Attitude model entails: R-IDENTITY

our reasons at t for then deciding or intending to do A are identical with our reasons at t for later doing A.

Reasons for action are explained by REASON: REASON

Any reason at t for later doing A consists in a desire for an end E combined with belief that doing A would further E.

Hence the Pro Attitude model claims that the attitudes which motivate us to decide to do A are desires for ends E combined with beliefs that doing A would further E. Reasons for deciding to do A are purposes for doing A — not purposes for deciding to do A. Whereas the Action model claims that decision-making is a means-end justifiable agency. Now, as we saw in chapter 5, any means-end justifiable agency may be done purposively: so, the Action model predicts, decision-making may be purposive. What the Action model predicts, Dan's decision shows. If Dan does decide to attempt the stunt, that will be because of his concern to co-ordinate his actions over time — his desire to avoid mismatched combinations of action. But until he has already taken that decision, Dan certainly will not yet believe that attempting the stunt would further that action co-ordinatory end. As we have seen, whether eventually attempting the stunt would minimise the risk of mismatched combinations of action, very much depends on whether Dan now takes a decision to attempt the stunt. So, prior to actually taking that decision, it is deciding to attempt the stunt, and not attempting the stunt, which Dan believes to be the means to co-ordinating his actions through time. Dan takes a decision to attempt the stunt, if he does so, because he believes that taking that decision minimises the risk of mismatched combinations of action. So R-IDENTITY can be false. The reasons which motivate Dan's decision to attempt the stunt are not reasons for performing the action decided upon. Dan's reasons for his decision are desires for ends combined with beliefs about how that decision would further those ends. His reasons for deciding to attempt the stunt are purposes for taking that decision.

In defence of the Action model

243

Dan's decision to attempt the stunt is motivated by a desire to co-ordinate his actions - a desire which, prior to his decision, does not provide him with any motive for attempting the stunt. But notice that Dan's decision is still motivation-perpetuating. Once Dan decides to attempt the stunt, the desire to co-ordinate his actions which motivated the decision will thereafter motivate him to act as decided. Given that Dan has decided to attempt the stunt and will thereafter be organising publicity and co-ordinating his remaining action with that decision, Dan's actions will only be co-ordinated if he does eventually attempt the stunt. So when Dan does attempt the stunt, his action will be motivated by the same desire that motivated the original decision - by a desire to avoid mismatched combinations of action. Decisions of which R-IDENTITY is false can still be decisions which exercise a motivation-perpetuating influence on subsequent action. Dan's decision, we have seen, falsifies R-IDENTITY. This is because a decision specifically to do A would settle more firmly than would any alternative exactly how Dan will eventually act. So a desire to coordinate action over time motivates Dan to take that specific decision to do A — to take that decision rather than another. Yet, until a decision about his future action is taken by Dan, it will not be clear to him which future action would best ensure the co-ordination of his actions. Dan has a reason for deciding which, at least initially, is not a reason for acting as decided. But Dan's case is unusual. Ordinarily R-IDENTITY turns out to be true. For we rarely have reasons for deciding to do A which are not pretty straightforwardly reasons for doing A. Ordinarily, each of a decision to do A or a decision not to do A would equally well settle our future action. So a concern to co-ordinate action by settling how we shall act in advance will not incline us towards taking either decision in particular. Indeed, in much everyday decision-making we do not consider the possible consequences of our decisions at all. We consider merely the actions decided upon and the ends which those actions would further. So we are moved to take decisions as the Pro Attitude model predicts - simply by our desires for ends and our beliefs about which actions might further these ends. The deliberation leading to decision is very often as the Pro Attitude model's supporters suppose - deliberation about how to act and not about how to decide. So if in our theory of the will we ignore questions of justification,

244

The psychology of freedom

and attend only to what attitudes actually motivate our decisions and to what, prior to deciding, we actually deliberate about, the Pro Attitude model may very well appear to us to be generally true.2 But remember that the Pro Attitude model is in every case false. For the story it tells about decision justification is in no case true. In chapter 2 I drew a parallel between deciding and trying. Trying to do A is in every case means-end justifiable: it is always justified by reference to the ends which trying to do A would further, and not by reference to the ends which doing A would further. That is why, no matter how desirable the ends which doing A would further, the fact that trying to do A might end in disastrous failure always serves as some justification for not bothering to try. But it is quite consistent with the means-end justifiability of trying, that our reasons for trying to do A often be simply our reasons for doing A. Unless we have been specially led to consider the possibility of trying and failing, we shall omit to consider the possible consequences of our attempts at action. We shall worry only about the possible consequences of the actions themselves. In such cases our deliberation will be about what action to perform, and not about what action to attempt. So we shall be led to try to do A by our desires for ends and our beliefs that doing A would further those ends. An analogue of R-IDENTITY is often true of our reasons for trying. But the analogue for trying of J-IDENTITY is, nevertheless, in every case false. As for trying, so too for deciding. ANSWERING SCEPTICISM ABOUT SECOND-ORDER AGENCY

In chapter 21 listed three sceptical arguments against the existence of an agency of the will. The last chapter answered the Regress argument. This chapter has answered the Reduction argument and the argument from Non-Purposiveness. The Reduction argument was simple. It assumed that intention 2

It is significant that supporters of the Pro Attitude model such as Bratman and Davidson tend to restrict their discussions to questions of motivation - to discussion of the belief-desire reasons we have for our decisions and intentions. They neglect explicit discussion of decision justifications. Of course, even if such neglect were warranted, it would not save the Pro Attitude model. Dan's decision falsifies R-IDENTITY as well as J-IDENTITY. But discussing decision rationality entirely in terms of motivating belief-desire reasons disguises how generally wrong the Pro Attitude model is about decision rationality. As we have just seen, counterexamples are not restricted to the Dan-style cases which falsify R-IDENTITY.

In defence of the Action model

245

is a kind of desire, and decision-making a kind of desire formation. It also assumed that desire formation is passive. From these assumptions it followed that decision-making is no agency. But the Reduction argument is not sound. For its soundness depends on the truth of the Pro Attitude model of decision and intention rationality. It is clear enough why. Desires to act are pro attitudes. They are justified by reference simply to the ends furthered by their objects - the actions desired. Hence, an intention to act cannot be a kind of desire, and a decision to act cannot be a kind of desire formation, unless intentions are pro attitudes too - unless, as the Pro Attitude model teaches, decisions and intentions to act are likewise justified by reference simply to the ends furthered by their objects. But we have now established that the Pro Attitude model is false. It is the Action model of decision and intention rationality which is true. Because the Action model is true, intention not only cannot be a kind of desire; it cannot be a kind of practical judgment either. Pace Davidson, intentions cannot be judgments that the actions intended are desirable. For, like desires, such judgments are pro attitudes too. Like desires but unlike intentions, judgments that actions are desirable are justified in terms of the likely desirability of their objects - the actions judged desirable. Since intentions are distinct from practical judgments, the will cannot be a deliberative faculty. The will must be precisely what, all along, I have suggested it to be - the locus of an executive, rather than a deliberative agency. Decision-making is a second-order executive agency by which we apply our practical deliberation, by causing ourselves to perform the first-order actions which those deliberations recommend. Decision-making really is precisely what, according to Bernard Williams, only 'bad philosophy' imagines it to be - 'another mental action' which lies 'between coming to a conclusion and acting on it'. The argument from the Non-Purposiveness of decision-making has also been answered. Decision-making counts as agency on the same terms as trying - as a form of means-end justifiable agency which may take purposive form, but which is characteristically performed non-purposively. But that does not mean that decision-making is like trying in all other respects. In the next chapter, we will show how the status of decision-making as a form of agency is fully consistent with its

246

The psychology of freedom

being reason-applying in its effects on subsequent action. We must show how the Action model of decision rationality is fully consistent with REASON-APPLY. And to show that, we shall have to exploit some crucial dissimilarities between decision-making and trying.

CHAPTER 9

The special-purpose agency of the will

INTRODUCTION

The last chapter argued that decision-making is a form of means-end justifiable agency. The Action model of decision rationality is true: justifications for deciding to do A consist in the likelihood of that decision furthering desirable ends. But chapters 4 and 5 also argued that, as agency, decisionmaking is reason-applying in its effects on future action. In particular, decision-making is a rationality-preserving cause of action. And that means, I have argued, that there has to be a special restriction on the ends which can justify deciding to perform an action A. Such decision-justifying ends must conform to REASONAPPLY: REASON-APPLY

any end E that justifies deciding to do A must, supposing that decision is taken, also provide at least as much justification for doing A.

Now, as we have seen, decisions can perfectly well further desirable ends which do not conform to REASON-APPLY. There could be a prize offered simply for taking a decision to do A. The winning of such a prize might then be a desirable end which deciding to do A was likely to further. But such a prize could not justify thereafter doing A as decided - and so, according to REASON-APPLY, the likelihood of the decision's winning the prize could not provide any justification for taking the decision. It may be true that any justification for deciding to do A must be a desirable end which taking that decision is likely to further. But the converse is false. Not every desirable end which taking that decision is likely to further, can be a justification for taking it. 247

248

The psychology of freedom

Decision-making, though means-end justifiable, must be a form of special-purpose agency.

We need now to explain why decision-justifying ends should conform to REASON-APPLY. And we need an explanation framed in terms which apply to means-end justifiable agency generally - which apply to actions as well as to decisions. We do not want REASON-APPLY to be an arbitrary constraint on decision justifications without any parallel in the rest of practical reason. Our theory of decision rationality must not be ad hoc. It must not rely on principles of practical reason which are mysteriously specific to the will. We have, therefore, to consider whether there is any such thing as special-purpose action - action which is justified only by ends which conform to an analogue of REASON-APPLY. If decisionmaking turns out to be an instance of a more general phenomenon - a phenomenon also found in our means-end justifiable action we can then obtain the principled theory of decision rationality which we want. For we shall then be able to derive our account of decision rationality from an account of how reason governs means-end justifiable agency generally.

END-SPECIFIC AGENCY

Let me now introduce the notion of end-specific agency. By end-specific agency I mean any means—end justifiable agency which must, when performed, be motivated by the desire to attain a specific end. Endspecific agency is nothing recondite or mysterious. We can find cases of it in plenty in our everyday action. Let me begin with an example of action which is not end-specific, and then contrast it with action which is. Playing loud music is not an end-specific action. And that is because, when I play music loudly, there is no specific end which need be motivating me so to act. I may have no motive at all for playing loud music. I could be playing loud music because, although I prefer to play it quietly, my volume control is unreliable. Or, if I have a motive, it could be any one of a variety of possible motives. It may be that I want to win the prize you have offered for successful loud music playing. Or I could be playing loud music because I want to annoy my neighbours.

The special-purpose agency of the will

249

By contrast, playing loud music in order to annoy the neighbours is an end-specific action - as is any action performed for a specific purpose. I am performing an action A in order to attain a specific end G - in this case, the end that the neighbours be annoyed. That means that a desire for that specific end G must be motivating what I do. I must be acting as I do because the attainment of G is something which, overall or on balance, I want. Where I am playing loud music in order to annoy the neighbours, annoying my neighbours must be something which, overall or on balance, I desire to do; and that desire must be motivating me to play music loudly. What more precisely must the overall desire for G's attainment come to? Often the desire that motivates end-specific action is an outright and overall preference that the end G be attained. If I am playing loud music in order that my neighbours be annoyed, then that is probably going to be because on balance I actually prefer that my neighbours be annoyed. But the desire for G need not be an outright preference. When I reach for a packet of biscuits in the supermarket, my action is performed in order to pick out a particular packet - that B-brand packet, for example. But, as we have already seen, I might none the less be on balance indifferent to whether I end up picking out that particular packet or another one of the many on display. If I discover at the check-out counter that I picked out another, brand-A packet instead, I need not have any regret. To be reaching out in order to pick out that brand-B packet, then, I need be motivated by nothing more than a desire to pick out that packet which is at least as strong as my desire to pick out any other. I need not hold a preference for that packet in particular. Again, the motivating desire for G may be unconditional on whether I do A; but it need not be. I may desire the attainment of G, but only on condition that I also do A. Here is an example. Agents nowadays still trek dangerous miles, rather than fly them by helicopter, in order to reach the North Pole. They clearly desire to reach the North Pole supposing that they do trek dangerous miles; but by not taking nowadays easily obtainable helicopters, they equally clearly show that they desire not to reach the North Pole unless they have first trekked dangerous miles. The end is not valued apart from the means. They do not hold a desire to attain the end which is unconditional on whether they employ the means.

250

The psychology of freedom THE RATIONALITY OF END-SPECIFIC ACTION

Now that we have an idea of what counts as end-specific action, we need to develop a theory of its rationality. What justifies performing end-specific action? What justifies doing A in order to attain some specific end G? A natural thought is that justification is provided by desirable ends which, once A is done, would thereafter be furthered by the attainment of G. What justifies playing loud music in order to annoy the neighbours, are desirable ends which, once loud music has been played, would be furthered by the neighbours getting annoyed. Such ends might be getting one's own back on the neighbours, or forcing the neighbours to leave the street and move elsewhere. Now this natural view places a special restriction on the ends which justify end-specific action. End-specific action comes out as special-purpose action. Not every desirable end which end-specific action is likely to further justifies performing that end-specific action. It is easy enough to see this by adapting a case already familiar from our discussion of decision-making. Suppose that I offered you a prize - the usual £1 m - simply for playing loud music in order to annoy your neighbours. You win the prize irrespective of whether, in fact, your neighbours are successfully annoyed. The prize rewards you for performing an action for a certain purpose - but not for the achievement of that purpose. Then, on the natural view which I am proposing, the prize which I have offered you does not actuallyjustify your playing loud music in order to annoy your neighbours. It does not matter that winning the prize is a highly desirable end which performing the end-specific action would further. The point is that winning the prize is not an end which, once the action was performed, would also be furthered by annoying your neighbours. So playing loud music in order to annoy your neighbours is a case of special-purpose agency just as — or so I have been claiming — taking a decision is a case of specialpurpose agency. Some philosophers might deny that end-specific action is specialpurpose. They might think that the huge reward of £1 m just for playing loud music in order to annoy one's neighbours, provides ample justification for that action. It is, after all, a highly desirable end which performing the action would further. What can be said to refute these philosophers?

The special-purpose agency of the will

251

In order to refute them, I must establish a principle about how practical reason governs end-specific action generally. Let doing A in order that G be a case of end-specific action - action which must be motivated by a desire that G be attained. Then, I claim, the following principle END-SPEC is true: END-SPEC

any end E that justifies doing A in order that G must provide at least as much justification for holding an overall desire that G be attained supposing that A is done.

So, I claim, justification for doing A in order that G is not provided by any old desirable end which that action would further. The end must conform to END-SPEC. Why should we believe END-SPEC? The argument we need exploits a fundamental but very simple truth about the connexion between justifications and rationality. If certain considerations count as justifying the performance of a given action, then it must also be true that those considerations would move a rational agent to perform that action. Agents who are rational, are precisely those agents who are moved by the available justifications for action, to do what those justifications justify doing. Justifications for action, then, are such as move a rational agent to perform the actions justified. But that they can do only if, in justifying those actions, they ipso facto provide at least as much justification for the motivation which those actions require. Why else should a rational agent develop the required motivation to act, unless the relevant justifications for action ipso facto make it rational for him to develop it? What motivation do I have to have if I am to do A in order that G? As we have already seen, if I am doing A in order that G, what motivates me must be an overall desire to attain G which, if not unconditional, must at least be conditional on my now doing A. So that is the motivation which any justification for doing A in order that G must also justify. Any end E which provides justification sufficient to move a rational agent to do A in order that G, must provide corresponding justification for an overall desire for G's attainment supposing that A is done. But that is just what END-SPEC claims. So END-SPEC is true. When does an end provide justification for holding a desire? We already know the answer to this question. Desires are pro attitudes.

252

The psychology of freedom

They are justified by the likely desirability of their objects. An overall desire, in particular, is justified by the desirability, overall or on balance, of its object. And that overall desirability of the object depends, in turn, on the various ends which that object is likely to further, and their desirability. An end E provides some justification for an overall desire to do A if, and only if, E is a desirable end which doing A is likely to further. As for overall desires in general, so too for overall desires to attain G once A is done. An end E justifies such a desire only if E is a desirable end which would likely be furthered by G's attainment supposing A is done. Now, as we have seen already, an end E can both be desirable and be furthered by doing A in order that G, without its also being true that it would be furthered by G's attainment supposing A were done. The end could be a prize offered, as in our example of loud music playing, simply for doing A in order that G - a prize that would be won whether or not G was attained, END-SPEC therefore entails that an end E can be desirable and likely to be furthered by doing A in order that G, without however providing any justification whatsoever for doing A in order that G. END-SPEC entails that endspecific agency is special-purpose agency. DECISION-MAKING AS END-SPECIFIC AGENCY

I claim that taking a decision is another case of end-specific agency. Deciding to do A, I claim, is an activity - a second-order action which, whenever it occurs, must be motivated by an overall desire for a specific end. The end in question is simply that the agent does A. A decision to do A implies an exercise of the will which is motivated by a desire that one subsequently does A. As with end-specific first-order action, the motivating desire to do A may be an outright preference. But it need not be. We have already considered the example which shows this. In a supermarket, I can decide to pick out a particular B-brand biscuit packet without having to be motivated to do so by an outright preference for picking out that packet in particular. My decision may be motivated by a desire to pick out that biscuit packet which is at least as strong as, but no stronger than, my desire to pick out any other. Again, the desire to do A may be unconditional. But, as with endspecific action, it need not be. The motivating desire may easily be wholly conditional on a decision to do A being taken first. And we

The special-purpose agency of the will

253

have already seen why. Many actions are worth performing only if one has decided to perform them in advance. Performing them can be extremely bad news otherwise. In Dan's case, for example, it is worth his attempting the stunt only if he takes a decision to attempt it long enough beforehand, and so can organise rewarding publicity for his attempt. Otherwise it is very much preferable not to attempt the stunt at all. If deciding to do A is a form of end-specific agency, and the relevant, specified end is that one does A, then deciding to do A is governed by END-SPEC SO that: any end E that justifies deciding to do A must, supposing that decision is taken, also provide at least as much justification for holding an overall desire to do A. This claim is close to REASON-APPLY: any end E that justifies deciding to do A must, supposing that decision is taken, also provide at least as much justification for doing A. How can we derive REASON-APPLY from END-SPEC? TO derive REASON-APPLY, we need to assume that any ends which justify holding an overall desire to do A are going, ipso facto, to provide at least as much justification for doing A. It must be true that whatever makes it rational to desire overall to do A is also going, ipso facto, to make it rational to do A. To derive REASON-APPLY, therefore, we must first eliminate those cases where the rationality of desiring overall to do A does not guarantee the rationality of doing A. What makes it rational to desire overall to do A? As we have agreed, desires are pro attitudes: they are justified by the likely desirability of their objects. What makes an overall desire to do A rational is the likely overall desirability of doing A - and that, in turn, depends on the desirability of the ends which doing A would be likely to further. So there is one case in particular where the rationality of desiring overall to do A clearly will not guarantee the rationality of doing A. This is a case where the rationality of doing A does not depend at all on the desirability of the ends which doing A would be likely to further. In other words, this is a case where doing A is not means—end justifiable.

254

The psychology of freedom

Another case is where, though doing A is means-end justifiable, the fact that doing A might further a desirable end does not necessarily count as a justification for doing A. That is a case where doing A, though means-end justifiable, is also special-purpose. In the case of special-purpose agency, as we have seen, the fact that the agency would further highly desirable ends - such as the gaining of huge cash prizes - might make us quite rationally desire to perform it. That we perform the agency in question is, after all, highly desirable. We might even be led by our desire rationally to try to manipulate ourselves into performing the agency. But the agency's furtherance of these desirable ends still need not provide any justification whatsoever for the agency's performance itself. Suppose, though, that doing A is means-end justifiable without being special-purpose. Doing A, that is, is justified by the desirability of the ends which it is likely to further - and there is no special restriction on which such ends can justify doing A. In other words, suppose doing A is justified by whatever makes doing A desirable. Then in such a case I see no reason why the rationality of desiring to do A and of doing A should fall apart. If it is rational to desire overall to do A, it is also going to be rational to do A. So where doing A is a means-end justifiable activity which is not special-purpose, we can derive REASON-APPLY from END-SPEC. In this case, REASON-APPLY is true. Anyjustification for deciding to do A must provide corresponding justification for doing A, supposing that decision is taken. But then, in any case, REASON-APPLY does only hold good for those activities which are means-end justifiable without being special purpose. Where doing A is not means-end justifiable, or is a case of special-purpose agency then, of course, REASON-APPLY is anyway false, and the activity in question is not subject to the will — as we saw in chapter 7. Ends can easily justify deciding to perform a special-purpose agency such as decisionmaking, without thereafter providing at least as much justification for performing that agency decided upon. My argument, then, is that the restrictions on eligible decision justifications placed by REASON-APPLY, are just a special case of the restrictions on justifications for a more general kind of agency - a kind of agency which includes action as well as decision. They are a special case of the restrictions placed by END-SPEC on our justifications for performing end-specific agency generally.

The special-purpose agency of the will

255

HOW DOES DECISION-MAKING CAUSE END-SPECIFIC ACTION?

In chapter 7 I argued that an activity A is subject to the will, so that we can do A simply on the basis of deciding to do A, only if doing A is a form of means-end justifiable agency which is not specialpurpose. Since decision-making is a case of special-purpose agency, decision-making is not subject to the will. Now we can see that some of our action is not subject to the will either. Action which is end-specific is a form of special-purpose agencyso it too is not subject to the will. I cannot play loud music in order to annoy my neighbours simply on the basis of deciding to. You can offer me £1 m if I play loud music in order to annoy my neighbours. And that prize may fully justify and motivate my deciding to perform that end-specific action. The prize is, after all, a highly desirable end which performing the end-specific action would further. But that prize will not justify performing the end-specific action itself. So the only way that my prize-driven decision will give rise to the action decided upon is by way of causing some intervening and motivation-influencing action. I shall first have to perform some action which leaves me with the required motivation for annoying my neighbours. Perhaps I will take an animosity drug which leaves me hating my neighbours; or perhaps I will place a bet on my managing to annoy them. End-specific action is not subject to the will. But, even so, endspecific action can still arise from the will. For all action is purposive. All action is performed in order to attain some specific end or other. And that means that all action is, under some description or other, a case of end-specific agency. Whenever I do A purposively, I do it in order to further some specific G. My non-endspecific A-ing is thus also an end-specific A-ing in order that G. How, in the case of end-specific action, does decision-making give rise to a form of agency which is not subject to the will? The answer is simple. Qua end-specific agency, action is not subject to the will; but qua non-end-specific agency it is. Doing A in order that G is end-specific, and so not subject to the will. I cannot do A in order that G at will. But doing A is non-end-specific and is subject to the will. I can do A at will. My decision-making leads me to do A in order that G when, motivated by a desire for G, I decide to do A. This decision then perpetuates its motivation, and I am caused to do A out of the very same desire that motivated my

256

The psychology of freedom

decision - the desire for G. I cannot play loud music in order to annoy my neighbours at will. But what I can do at will is play loud music. So I can decide to play loud music. And if my decision to play loud music has been motivated by a desire to annoy my neighbours, then - since decisions perpetuate their own motivation when they give rise to action - so too will be the action decided upon. My decision will lead me to play loud music in order to annoy my neighbours. Whether actions are end-specific or not is relative to a description, the same actions can be described both in end-specific and in non-end-specific terms. So whenever I perform some end-specific action — whenever I do A in order that G — there is something nonend-specific which I do - doing A - which is not special-purpose, and which therefore is both subject to the will and an action which I could have performed for quite different purposes. I can do what I do when I perform some end-specific action - that is, I can perform an action which produces the same outcomes or effects in the same way - even if I lack the motivation which that end-specific action requires. I can do what I do when I play my music loudly in order to annoy the neighbours - but without doing it in order to annoy the neighbours. I may perform the action in order to win a prize instead. CAN WE EXERCISE THE WILL WITHOUT TAKING DECISIONS TO ACT?

What about decisions? Decisions to act, I have claimed, are cases of end-specific agency too. They too must be motivated by a desire for a specific end. An exercise of the will only counts as a decision to do A if it is motivated by a desire to do A. Does that mean that what was true of end-specific action is going to be true of end-specific decision-making too — that it can occur without the motivation which its end-specific form requires? We have seen that I can do what I do when I perform some end-specific action - that is, perform an action which produces the same outcomes or effects in the same way- even if I lack the motivation which that end-specific action requires. So why cannot I do what I do when I take a decision to do A - that is perform will agency which produces the same outcome in the same way - though without what I do having the motivation which a decision to do A requires?

The special-purpose agency of the will

257

It is clear what outcome a decision to do A produces. It leaves one motivated to do A. A decision to do A causes one to do A by leaving one with a continuing desire to do A — a desire which, given belief that the time for action had arrived, would move one to do A. Why, then, could not there be an exercise of the will which produced precisely that same outcome in exactly the same way, though without itself being motivated by a desire to do A? Cannot I do what I do when I take a decision to do A - but do it simply in order to win a prize for doing it, a prize won whether or not I subsequently did A? Perhaps what I would then be doing would not be deciding to do A. But could not it still be an exercise of the will with the same effect on subsequent action motivation and action? Let a *decision to do Abe an exercise of the will which produces the same outcome in the same way as a decision to do A - it causes a motivation to do A in the same way as that decision would - but which, unlike the decision, is not itself motivated by a desire to do A. Why cannot we take *decisions as well as decisions? Compare decision drugs. Taking the decision drug for an action A in order to ensure that I do A, is an end-specific action. It shares its motivation - a desire that I do A - with deciding to do A. But I can do what I do when I perform that end-specific action - 1 can take the drug, thereby leaving myself motivated to do A - without being motivated by that desire. I can take the drug simply in order to win a prize for taking it. Why cannot our second-order will agency be like our first-order decision drug taking? If we could take *decisions as well as decisions, then we could rationally exercise the will in ways which directly subverted our subsequent rationality. The will would no longer be a faculty for rationality-preserving agency - and it is easy enough to see why. As *decision takers, we could exercise the will exactly as we would when taking a given decision to do A — producing a motivation to do A in the same way as the decision — but be thus exercising the will simply out of a desire to win a prize offered just for so doing. In which case REASON-APPLY would not rule out the prize offer as sufficient justification for that exercise of the will. For REASON-APPLY, remember, only restricts what can justify our taking decisions to act. It places no restrictions on what can justify taking *decisions. REASON-APPLY, after all, is a constraint only on

258

The psychology of freedom

the rationality of will agency in its end-specific form - in the form in which it must be motivated by a desire to perform the action which it produces. And *decisions to do A, by definition, are not so motivated. So if *decisions were possible at all, there is no reason why purely prize-driven *decision-taking could not be as rational as purely prize-driven decision drug taking. And no more in the case of purely prize-driven *decision-taking than in the case of purely prize-driven decision drug-taking, would the rationality of the agency which won the prize, guarantee the rationality of whatever further action it was which that agency left one motivated to perform. As by decision drug taking, so by *decision-taking, we could rationally leave ourselves motivated to act irrationally. But the will, I take it, is a faculty for rationality-preserving activity. The action which any rational exercise of the will leaves us motivated to perform, must be rational in turn, REASON-APPLY constrains any rational exercise of the will - and not just such exercise of the will as takes the form of decision-making. It must then be true that we cannot take *decisions to do A. We cannot do what we do when we take a decision to do A - exercise the will to produce a motivation to do A in the same way - without what we do itself being motivated by a desire to do A. We cannot do what we do when we take a decision to do A, without ipsofacto taking such a decision. Any exercise of the will is a decision to act — a decision to act which, like all such, is governed by REASON-APPLY. And it is clear enough why there cannot be *decisions as well as decisions. The explanation has formed a prominent part of our theory of the will all along. All along I have been claiming that the will is a motivation-perpetuating cause of action, and that this feature of the will is essential to it. Any exercise of the will affects subsequent action by working off and perpetuating its own motivation. Any exercise of the will works by ensuring that the desires which motivated it are retained and thereafter motivate the agent's action too. Consider how decisions cause the action decided upon. As chapter 4 explained, a decision to act functions as a desire-stabiliser and tie-breaker in relation to its own motivation. A decision to act ensures that the desires to act which have motivated it are retained to the time of action, and override any equally strong desires to act otherwise. That is always how a decision to act causes the action decided upon. Compared to desires, the intentions which our

The special-purpose agency of the will

259

decisions form possess, as we have noted, inertia. And that is because of the tie-breaking and stabilising effect which any decision has on its own motivation. In characteristic cases - cases where the agent's reasons for his decision are simply reasons for acting as decided - that is all a decision does to generate action. The decision is motivated by what is already an unconditional desire to act as decided; and the decision simply ensures that the desire is retained up to the time of action and motivates the agent to act as decided. But sometimes the desire to act as decided which motivates a decision is very much conditional on that decision being taken. In which case, besides acting as a desire-stabiliser and tie-breaker, the decision has to turn the desire which motivates it from a merely conditional desire into a non-conditional or outright desire. And that the decision does by causing the decision maker to believe that he has taken it. Dan's decision in the last chapter is a clear example. His decision to attempt the stunt is motivated by a desire to co-ordinate his actions over time. But until Dan actually takes that decision to attempt the stunt, he does not yet believe that acting as decided — attempting the stunt - would co-ordinate his actions over time. Dan's reason for taking his decision would provide a reason for acting as decided - but only supposing the decision has been taken. Dan's decision to attempt the stunt is motivated by a desire to act as decided - but one which is very much conditional on his taking that decision. So if Dan's decision is to leave him with an outright motivation to attempt the stunt as decided, it must cause him to believe that he has actually decided to attempt the stunt. But that the decision can easily do. Often, perhaps typically, taking a decision to do A does get us to believe that we have so decided. For deciding to do A has an action co-ordinatory function - to fulfil which function, the decision has to get us to believe that we will or might well do A. And, as I have already observed, that the decision may well do by getting us to believe in it - by getting us to believe that we have decided to do A. So will agency is motivation-perpetuating. It generates action by working off its own motivation - by stabilising the desires that motivated it, by tie-breaking or ensuring that those desires override any contrary desires that are equally strong; and also, where necessary, by leading the agent to believe that the will agency has been performed, thus converting its own motivation into the non-

260

The psychology of freedom

conditional desire to act which the action willed requires. That is, at any rate, how the will agency which takes the form of a decision to do A generates action. Any exercise of the will which is to have the same outcome as a decision to do A-which is similarly to leave the agent motivated to do A - must, as an exercise of the will, work in precisely the same motivation-perpetuating way too. It must likewise cause the agent to do A by working off its own motivation in the same way. It must likewise operate by stabilising the desires that have motivated it and ensuring that they continue thereafter to motivate the agent to do A — where necessary, additionally transforming those desires into a non-conditional desire to do A by causing belief in its own performance. But then if the requisite motivation for doing A is to be thus produced, the desires which motivated that exercise of the will must already have constituted an overall desire to do A - a desire to do A which was either unconditional or at least conditional on that exercise of the will. Only then can the will's exercise leave the agent motivated to do A by simply perpetuating its own motivation. But then such an exercise of the will constitutes a decision to do A. A decision to do A just is will agency which disposes one to do A, and which is itself motivated by an overall desire to do A. So no ends can move me to do what I do when I take a decision to do A, without first moving me to form a desire to do A. For, unless I already hold such a desire, and unless this desire is already motivating my exercise of the will, there will be nothing recognisable as the same will activity as occurs when I take that decision: there will be no will activity which produces the same effect on subsequent action-motivation and action in the same way. Any exercise of the will which affects subsequent motivation and action in the same way as a decision to do A must itself be so motivated that it counts as a decision to do A. The end-specificity of secondorder agency is an essential feature of it. REASON-APPLY is a constraint on the rationality of any exercise of the will which counts as a decision to perform first-order agency. But the will's operation is essentially by motivation-perpetuation. So any exercise of the will with the same outcome as a given decision to act, must anyway constitute the taking of that decision. It is essential to will agency that it take the form of decisionmaking. Hence REASON-APPLY is a constraint on the rationality of our will agency generally.

The special-purpose agency of the will

261

DECIDING AND TRYING

Let us now compare decision-making with another kind of agencya kind of agency which we have already seen to be importantly similar. So similar, indeed, that Locke treated both kinds of agency as one and the same - as instances of an undifferentiated category of volition. Let us now compare decision-making with the agency of conation or trying. Deciding to act and trying to act are indeed very alike. Consider, again, the resemblance between, say, deciding to raise one's arm and trying to raise one's arm. Each is a kind of action-generating agency. Each can cause one's arm to rise, thereby explaining one's successful performance of the action of raising one's arm. Indeed, each of these action-generating activities is identified, whether as a decision to raise one's arm, or as an attempt at raising one's arm, in terms of the action which it generates. Moreover, when one tries to raise one's arm, what one does is characteristically explained by just the same desire as would explain a decision to raise it — by an overall desire to raise one's arm. That is characteristically exactly why one decides or why one tries to raise one's arm. In each case one usually does so because one wants to raise one's arm. But notice now a crucial difference between trying and deciding. I have been claiming that I cannot do what I do when I take a decision to raise my arm - I cannot exercise my will in a way that has the same outcome - without what I do being motivated by a desire to raise my arm. But I plainly can do what I do when I try to raise my arm — I can exercise my capacity for conative activity in a way that has the same outcome -without what I do being motivated by a desire to raise my arm. Recall the conation experiment I mentioned in chapter 2. Suppose I am carrying out such an experiment. I am investigating what happens in the arm muscles when people try to raise their arms, but then fail because their arms are restrained. So I have had my arm bound as firmly as possible to a table. Now I do not desire, overall or on balance, that as a result of what I am about to do I raise my arm. Indeed, I definitely prefer that I do not raise my arm. For if I did raise my arm, the experiment would have failed, and I should only have to restart it with my arm bound yet more firmly. But I can still proceed to do exactly what I would do if I were trying to raise my arm out of a normal motivation - out of a genuine

262

The psychology of freedom

desire to raise it. That is, what I do can have the very same outcome as a normally motivated trying performed in the same circumstances, with my arm bound down. What I do can still cause the very same nerve signals and muscle motions - the very nerve signals and muscle motions which usually cause my arm to rise. Just as in the case of a normally motivated trying, what I do would cause my arm to rise: unless my arm is bound tightly enough, I shall raise it. Indeed, it is very natural to describe what I do as an actual case of my trying to raise my arm - albeit a trying performed, very unusually, given a desire that my attempt in fact fail. Someone might object to the last claim. Someone might insist that trying is another case of end-specific agency- that any attempt at action must be motivated by a desire to perform the action attempted. In which case, given the lack of a motivating desire to raise my arm, my conation in the conation experiment cannot count as a genuine attempt at raising my arm. We should instead describe it as a *trying to raise my arm, to mark the lack of the motivation which, on this view, a genuine attempt at raising my arm would require. Now I doubt that this objection is sound. I doubt that we do so tie the identity of trying to its motivation. 1 When the lack of a genuine desire to succeed does lead us to deny that someone is genuinely trying, that is always because their lack of the desire is affecting what they do in other respects. They are not pushing so much, or flexing their muscles quite as they would if they really wanted to succeed. But ex hypothesis such considerations do not apply in this case. When in the conation experiment I perform my conation, aside from its motivation there is absolutely no difference between what I actually do, and what I would be doing if I did want to raise my arm. As we might naturally put it: it can be part of the conation experiment, and essential to its success, that I be trying as hard to raise my arm as someone would who really did want to raise it. In any case, it is clear that the issue is largely verbal. Whether, as I recommend, we describe what I do in the conation experiment as trying to raise my arm, or whether we describe it instead as *trying, the main point that I made still stands. The outcome of what I do when I perform my conation in the conation experiment, is still 1

My thanks to Mike Martin and Alfred Mele for encouraging this doubt. See also on this point Alfred Mele's 'He wants to try*, Analysis, 1990.

The special-purpose agency of the will

263

independent of the conation's unusual motivation. What I do in the conation experiment can have exactly the outcome that an ordinarily motivated attempt at raising one's arm would have, when carried out in the same circumstances. Trying to raise one's arm is normally motivated just like a decision to raise one's arm - by a desire to raise one's arm. But in the case of the trying, I can do what I do when I try - perform conative activity which produces exactly the same outcome in the same way - without what I do having that normal motivation. I can even do what I do when I try to raise my arm despite holding an outright preference that my arm not rise. Things are quite different when we come to the will. We need only compare the conative agency that occurs in the conation experiment with the prior agency of the will which gives rise to it the prior decision to try to raise my hand. For conation, as I observed in chapter 2, is subject to the will. Unlike deciding, trying is something which we can do on the basis of a prior decision to do it. So consider the will agency in the conation experiment which explains my conation. What motivates this will agency is exactly what thereafter motivates the conation which it explains. What motivates will and conative agency alike is the same rather unusual desire - a desire that I try to raise my arm without raising my arm. But, in the case of the will agency, this unusual motivation very clearly makes a difference to the outcome of what I do. The outcome of my will agency is certainly not what it would have been had it been motivated by a desire to raise my arm. For I am not left with a persisting desire to raise my arm. Once my will agency has occurred, raising my arm is clearly something I prefer not to do. The persisting desire with which my will agency leaves me, is a desire to try to raise my arm without actually raising it - the very desire which eventually motivates my conation. Moreover, we have no inclination to describe my will agency in the terms which we would have employed had it been motivated by a desire to raise my arm. My will agency in the conation experiment is clearly not a decision to raise my arm. Given the desire that does motivate it, the decision I take is, very clearly, a decision to try to raise my arm without actually raising it. What I do when I exercise my will very much depends, then, on what desires have motivated its exercise. Whereas what I do when I perform a conation — when I try to do something — is independent

264

The psychology of freedom

of its motivation. And that is exactly what we should expect on the account of the will and of conations or tryings defended so far. Remember the key distinction between trying and deciding - a distinction I drew in chapter 2. When a decision to raise my arm or an attempt at raising my arm cause my arm to rise, they do so in quite different ways. Deciding to raise my arm causes my arm to rise by ensuring that thereafter I am motivated to raise my arm. The function of decision-making is to cause the action decided on by way of influencing my motivation for that action. And this the decision does by perpetuating its own motivation. Whereas trying is not even by function a motivation-influencing agency, let alone a motivation-perpetuating one. Trying to raise my arm causes my arm to rise in a motivation-independent way. The trying causes my arm to rise simply byway of causing nerve impulses and muscle motions. Any motivation for the action of raising my arm will already have caused the trying - it will not itself be caused by it. And that is why classifying will activity and trying together as uniformly cases of volition is such a mistake. The agency of the will is a second-order agency which is essentially motivationperpetuating in its effects. Whereas trying is an exercise of a quite distinct capacity for first-order agency which is motivationindependent in its effects. It should be no surprise, then, that the identity of agency which is motivation-perpetuating in its effects, depends precisely on its own motivation — whereas the identity of agency which is motivation-independent in its effects, does not. The will is a faculty for second-order agency whose function is to generate first-order agency - both the first-order agency of our conation and the first-order agency of the actions which our conations explain. This the will does through motivation-perpetuation. And, as we have seen, it is because the will is motivation-perpetuating in its effects, that the rational exercise of the will is constrained by REASON-APPLY. Justifications for doing what we do when we decide to perform some given first-order agency must always be justifications for performing that first-order agency thereafter. But, as the conation experiment shows, there is no analogue of REASON-APPLY constraining rational conation. Like the action which it explains, conation is means-end justifiable without being special-purpose. There is no special restriction on the ends which can justify performing conations. The ends which justify doing what we do when we try to perform some action, are not restricted to

The special-purpose agency of the will those which justify performing that action thereafter. And that is precisely because, although the function of conation may also be to generate action, conation does not operate through motivationperpetuation. So, in the conation experiment, I do exactly what I do when I try to raise my arm — indeed, surely it is straightforwardly true that I try to raise my arm - and can perfectly well have ample justification for doing it. After all, it may be very important to discover what happens when people try to raise their arms but fail. But what justifies my conation in the conation experiment, precisely does not justify raising my arm thereafter. Our conative capacity, then, is a capacity for agency-generating first-order agency - a capacity which, unlike the will, is not restricted to generating agency in a reason-applying way. Since conation, like action but unlike decision, is means-end justifiable without being special-purpose, it is subject to the will. We can make attempts at action — as we can perform actions, but as we cannot take decisions to act —just on the basis of prior decisions to do so. And, while we cannot exercise the will in order to win a prize just for so doing, we can perfectly well perform conations in order to win a prize just for so doing. I can perfectly well do what I do when I try to raise my arm, and do it in order to win a prize just for doing it — a prize which I win whether or not I subsequently raise my arm. The difference between conation and decision also shows up when we consider what we can sensibly be commanded to do. Remember, as O'Shaughnessy has observed, that we cannot sensibly be commanded to take a particular decision: there is no order: 'Decide to raise your arm'. (The Will, vol. 2, p. 300) And the reason for this, I argued, is that decision-making is governed by REASON-APPLY. Commanding makes sense only where it is possible, at least in principle, for the command to justify doing what is commanded. That what is commanded would comply with the command must be the sort of consideration which - if the command were authoritative enough - could justify doing it. Now, no matter how authoritative, the need to comply with a command simply to decide to raise one's arm could not possibly justify taking the decision commanded. And that is because the need to comply with the command, considered as a decision justifier, violates REASON-APPLY. It violates REASON-APPLY because, once the

265

266

The psychology of freedom

decision is taken, the need to comply with the command does not justify thereafter acting as decided. Conation or trying is quite different. As someone carrying out a conation experiment, or as your doctor, I might very well instruct you: Try to raise your arm.' Such commands make perfect sense. And that is because there is no analogue of REASON-APPLY to constrain rational conation. So a command simply to try, when given with sufficient authority, could justify doing what was commanded. It could justify and motivate the appropriate conation. Just as our capacity for will agency, in being subject to REASONAPPLY, differs from our capacity for conative agency, so too it differs from any other capacity for action-generating first-order agency which we might possess. Suppose, for example, that we were able to take decision drugs. Decision drugs, like decisions but unlike conations or tryings, are motivation-influencing. But like conations and unlike decisions, decision drugs are not motivation-perpetuating. They do not work off their own motivation. Taking the decision drug for doing A is bound, whatever one's prior motives for taking it, to introduce a motivation to do A - a motivation which will persist for as long as one retains the beliefs that originally motivated taking the drug. So the decision drug for doing A can be taken - and quite rationally taken - in a way which would cause A's performance even when taken for purposes that do not require A's performance. Hence, our capacity to influence future action by taking decision drugs — should we ever have such a capacity — would not be a capacity for agency which could only be exercised in a reason-applying way. It would not be the capacity for reasonapplying future action control which we possess in the will. As we have already noted, all too many theorists of the will would deny that REASON-APPLY is a constraint on rational will agency. Now we can see that not only is this denial a mistake. It is also one manifestation of a yet more fundamental error — an error of a kind which up till now has quite generally bedevilled those theories of the will which are not Hobbesian, and so which rightly admit that the will is a capacity for agency. This is the error of failing fully to distinguish the will, as a capacity for action-generating second-order agency — a capacity which generates action through motivationperpetuation — from capacities for action-generating first-order agency which, though superficially similar, generate action in fundamentally different ways. This is, in particular, the error of

The special-purpose agency of the will

267

failing adequately to distinguish the agency of the will from the agency of conation. From John Locke onwards, too many believers in an agency of the will have made out tryings to work like decidings - or, as in the case of those like Lewis and Kavka who deny REASONAPPLY, have made out decidings to work like tryings. CONCLUSION

In the last two chapters, I have explained how the will can be what common-sense psychology ordinarily supposes it to be: a faculty for reason-applying agency. The function of the will is to apply reason as it concerns our subsequent action. And, quite consistently with that fact, the will is also a locus of action itself- action-generating action of a second-order kind. What vindicates the common-sense psychology of the will is the truth of a particular model of its rationality - a model which combines the Action model with REASON-APPLY.

In the last chapter, I showed that the will is a capacity for agency. And that is because the exercise of the will is means-end justifiable. The Action model of decision rationality is true. Then in this chapter, I have also explained the restriction placed by REASON-APPLY on the justifications which we can have for our decisions - a restriction which must apply if decision-making is to be reason-applying in its effects on subsequent first-order agency. And — as we wanted — my explanation has been provided by a general theory of how practical reason governs means-end justifiable agency of all kinds, whether second or first-order. I have avoided introducing ad hoc principles of practical rationality that are specific to the will. I have based my explanation on the claim that the will is a faculty for a kind of agency - for decision-making - which, as a specialpurpose agency, is an example of a more general kind of agency found in our first-order action also. Decision-making is a specialpurpose agency because it is a case of end-specific agency. That is to say, decision-making is an agency which, whenever it is performed, must be motivated by a desire for a specific end. But actions may be end-specific too. Any action will fall under certain descriptions — end-specific descriptions — such that so described it must be motivated by a desire for a specific end. And, as with decisions, there are special restrictions on what ends can justify performing

268

The psychology of freedom

actions thus described. The restrictions on what ends can justify our decisions are only a particular case of more general restrictions on what can justify end-specific agency generally. What is distinctive about the agency of our decision-making, and distinguishes it from any first-order agency, is that its end-specific form is essential to it. Unlike the first-order agency of our action, the second-order agency of the will can only be motivated as decisionmaking. And this is because the will's operation is essentially motivation-perpetuating. The will affects subsequent action by influencing what actions we are motivated to perform; and the will's effect on our subsequent action motivation is produced by perpetuating its own motivation. We can only do what we do when we take a given decision to act — perform agency which produces the same outcome in the same way- if our will agency has anyway been motivated in the same way as that decision to act. In other words, we can only do what we do when we take a given decision to act by taking that very decision. So the will is governed by reason in the same general way as our means-end justifiable agency generally. What distinguishes the second-order agency of the will from our first-order agency, is the will's mode of operation - the motivation-perpetuating way in which the will determines subsequent action.

Conclusion

In this book I have provided a new account of freedom and of the psychology which freedom requires. The account of freedom is Psychologising. Our freedom of action depends on our also possessing a prior psychological analogue of that freedom - a freedom that is located among the mental states which explain our action. Our freedom of action depends on our possessing a further free, reason-applying agency of the will. And this is because freedom involves a capacity for rational self-determination — a capacity to exercise control over our actions in a reason-applying way. Given the plan-centred nature of practical reason, to be free we need to be able to exercise a reason-applying control over whole plans of action extending into the future. And that we do through taking free, plan-controlling decisions. By taking decisions about how to act, we both exercise control over our actions in the form in which they matter - in the form of whole plans extending into the future — while at the same time ensuring those actions' continued rationality. For, by rationally deciding to perform actions in the future, we leave ourselves disposed to act rationally thereafter. And decisions are rationalitypreserving in this way, because they perpetuate their own motivation into the future. Rationally deciding to do A in the future leaves us disposed to do A equally rationally thereafter, since the decision leaves us disposed to act as decided for the same good reasons as motivated it. And because decisions determine action by perpetuating their own motivation, the influence of our decisionmaking on our subsequent action is essentially non-manipulative. When we take a decision to act, we are not twisting our future self s arm to get it to act as we want it to. We are not goading our future self into action, or binding our future self to performing the actions we favour. We are rather ensuring that our future self acts in the 269

270

The psychology of freedom

ways we favour out of a common rationality - for reasons which we share with it. Since the influence of our decision-making is essentially rationality-preserving and non-manipulative, it can never of itself remove our freedom. Deciding to perform an action always leaves us with a continuing freedom to change our minds and act otherwise than as we have decided. Our capacity for free decisionmaking both makes us free and, in its exercise, preserves our freedom. Notice that my account of why freedom of action depends on freedom of will, is economical in its assumptions both about the nature of freedom and about the nature of practical reason. The account is largely neutral in the dispute between Compatibilists and Incompatibilists. At any rate, it is entirely consistent with Incompatibilism in a form which is autonomy- rather than timecentred. Nor is my account committed to any controversial theory of practical reason, such as the radically Kantian theory that the requirements of practical reason are products of our legislation. What generates the Psychologising conception is something far less debatable. It is simply, first, the idea that to be free agents we must be able to exercise control over our actions in the form in which they matter to us, and in a way consistent with our continuing rationality; and, secondly, the plan-constituted form of almost all the action which matters to us — even of such everyday actions as curing tooth-ache, or getting a packet of biscuits. Besides defending a new, Psychologising account of freedom, the book also defends new accounts of agency and of the will (see Table 2, The psychology of agency - a final sketch). These theories are

linked to a new account of rational choice - of how the rationality of our decisions to act is related to the rationality of the actions which our decisions explain. These theories of agency, the will and of rational choice are all needed to show that we do actually possess the kind of will - the decision-making capacity - which our freedom requires. To be free agents we need in the will a special capacity for reason-applying, second-order action. I show that the decisionmaking by which we actually determine how we act meets this description. Our decisions to act both serve to apply reason as it governs our subsequent action, and also constitute a further form of deliberate action themselves. What makes decisions to act deliberate actions themselves, is a

Conclusion

271

Table 2. The psychology of agency - a final sketch Judging that

Action?

Desiring to move one's hand

it is desirable to move one's hand

Deciding to move one's hand

Trying to move one's hand

Moving one's hand

No

No

Yes, as 2nd-order agency

Yes, as ist-order agency

Yes, as ist-order agency

No

Possibly,

Possibly,

Yes

but

but

Other relevant features Purposive?

No

typically not typically not Means-end justifiable?

No

No

Yes

Yes

Yes

Specialpurpose?

n/a

n/a

Yes

No

No

Subject to the will?

No

No

No

Yes

Yes

rationality which decisions share with the actions which they determine. What unites all that we deliberately do - the practice both of our decisions to act and the actions which they determine - is being governed by practical reason. As the book has argued, decisions to act and the actions which they explain are all alike means-end justifiable occurrences. They are all alike justified by reference to desirable ends which they themselves are likely to further. And that means that decisions and the actions which they explain are all governed by reason in its distinctivelyjfrrac/zW - practice- or agencygoverning - form. For means-end justifiability is indeed reason in its distinctively practical form. And that is because means-end justifiability is reason in the form that governs any exercise of control. Any rational exercise of control involves producing outcomes that are desirable and preventing outcomes that are undesirable. And means-end justifiable doings are made rational precisely by the fact that their doing would produce some outcomes that are desirable, and prevent others that are not. The book has defended the agency of the will by arguing for the truth of an Action model of decision rationality. Decisions to act are justified by the ends which they themselves further, and not simply

272

The psychology of freedom

by the ends furthered by the actions decided upon. But the book has also shown that this model is fully consistent, not only with the will's agency, but also with its executive function. It is central to our conception of the will, I have argued, that decision-making serves to apply reason as it concerns subsequent action. The point of taking decisions about which actions to perform, is precisely to ensure that we subsequently perform the right actions. Any plausible theory of decision-making has to be accommodate the essentially executive function of the will, as well as its agency. One way of accommodating the executive function of the will, as we have seen, is to adopt a Pro Attitude model of decision rationality. If decision-making serves to apply reason as it governs subsequent action, then why not claim that decision justifications consist, purely and simply, in justifications for performing the action decided upon? But the Pro Attitude model, I have argued, is false. And it is the executive function of the will which proves the Pro Attitude model's falsehood. It is the use of decisions to co-ordinate subsequent action which refutes the Pro Attitude model, and establishes the Action model. The use of decisions to co-ordinate action shows that decision justifications are not identical with our justifications for the actions decided upon. There can be action co-ordinatory justifications for deciding to do A that, prior to the decision being taken, are not also justifications for doing A. And, quite generally, the likely co-ordinatory effect of a present decision or intention to do A can increase our justification for eventually doing A - thanks to our present intention to do A, there is an increased risk of misco-ordinated action if we do not do A as intended — but without thereby adding to our justification for now intending to do A. To do justice to the executive function of the will we need, not the Pro Attitude model of decision rationality, but REASON-APPLY, REASON-APPLY places a restriction on the ends which can justify deciding to perform an action A. Ends can only justify deciding to perform an action A if they provide at least as much justification for doing A thereafter. The book shows how REASON-APPLY explains important intuitions about decision-making - intuitions that sharply distinguish decisions from the agency of the actions or attempts at action which those decisions explain, REASON-APPLY explains why

Conclusion

273

the influence of decisions on subsequent action is always rationality-preserving - why rational decisions to act leave us disposed to act rationally thereafter, REASON-APPLY also explains why I cannot sensibly order you just to take a particular decision why I cannot sensibly command you 'Just decide to raise your arm' - as I can order you to perform an action, or to try to do something; and why decisions are not subject to the will, as actions and attempts at action are. That REASON-APPLY is true explains, in particular, a very important fact about intentions to act - their pro attitude-like character. For even if the Pro Attitude model of decision and intention rationality is false, it is a profoundly important fact about the common-sense psychology of the will that the Pro Attitude model at least looks true. Intentions to act, as we ordinarily conceive them, do very much resemble pro attitudes to action such as desires and practical judgments. Why else should the history of philosophy be so full of theories, of which Bratman's is only the latest, which treat intentions as pro attitudes themselves? Others before Bratman have actually assimilated intentions to specific pro attitudes - identifying them with practical judgments in the case of the Stoics and Davidson, and with desires in the case of Hobbes. Like intentions, pro attitudes such as desires and practical judgments are not subject to the will. And they cannot be commanded either. Just as there is no command 'Intend to raise your arm', so too there are no commands 'Desire to raise your arm' or Judge it desirable to raise your arm'. And the book shows that all this is precisely because desires and practical judgments are no more justifiable in terms of any old desirable ends that they might further, than are intentions and the decisions which form them. Indeed, far from being justified by ends which they themselves further - as they would have to be to be subject to the will or commandable — the rationality of desires and practical judgments directed at action entirely depends on the rationality of their object. What justifies a desire or practical judgment directed at action is simply what justifies performing that action. Now, quite consistently with the means-end justifiability of intentions to act, REASON-APPLY ties their rationality to the rationality of their objects in a very similar way. That is, REASONAPPLY ties the rationality of intentions to act to the subsequent rationality of the actions intended. Any end which justifies an

274

The psychology of freedom

intention to act must, once that intention is formed, thereafter provide as much justification for the action intended, REASONAPPLY helps explain, then, how intentions can be deeply pro attitude-like, without actually being pro attitudes themselves. Get rid of REASON-APPLY, and a core feature of the common-sense psychology of the will goes with it. Far from being in tension with an Action model of decision rationality, I have argued, REASON-APPLY is fully consistent with it. As a constraint on justifications for taking decisions to act, REASONAPPLY is not an ad hoc principle plucked out of the air, but is generated by the fact that the will is motivation-perpetuating in its operation — the very feature of the will to which I appealed in my account of why freedom of action depends on freedom of will. Remember why we should suppose that the will is motivationperpetuating in its operation. Decisions to act cause their execution by influencing what actions decision makers are thereafter motivated to perform. Now it might be thought that decisions cause their execution by exploiting a concern, which all decision makers possess, with decision execution for its own sake. But I have argued that, since decision execution is not obviously desirable in itself, this is a concern which decision makers can lack. What all decision makers do of course care about, at least at the time of their decisions, are the very ends which motivate them to decide to do A — characteristically, though as Dan's case shows not always, ends which from the outset they expect doing A to further. The most plausible account of decision execution, then, is that decisions to do A simply ensure that these decision-motivating ends thereafter motivate the decision maker to do A. Decisions cause their execution by perpetuating their own motivation. It is not surprising that there is no analogue of REASON-APPLY to constrain the rationality of what is a superficially similar but really importantly different capacity for action-generating action — our capacity for conation. That is precisely because, unlike decisionmaking, conation or trying does not generate action by perpetuating its own motivation. As we have seen, trying to raise one's arm causes one's arm to rise in a motivation-independent manner. And so justifications for trying to do A are not limited to ends which justify doing A thereafter. But though constrained by REASON-APPLY, our second-order agency is agency none the less. It is governed by reason in a

Conclusion

275

form appropriate to the exercise of control - by reference to desirable ends which performing it furthers. And that means that, though like our conations, our decisions are characteristically nonpurposive, as Dan's case showed they perfectly well can occur in purposive form. The will, then, really is what our belief in our freedom requires it to be - a faculty for action-generating action, which serves to apply reason as it governs the actions generated. Through the will, we can exercise deliberate control over our future action. And we can exercise that control in the way that matters - in a way which applies our deliberations about how we should subsequently act. We have precisely the freedom of will that we need to be free.

Bibliography

Anscombe, E. Intention, Blackwell (1957) Aquinas, T. Summa contra Gentiles, Book 2, tr. A. Anderson, Notre Dame 0975) Summa Theologiae, volume 17, tr. T. Gilbey, Blackfriars (1970) Bennett, J. 'Why is belief involuntary?', Analysis 50 (1990) 87-107 Bradley, F. H. Ethical Studies, Oxford (1876) Bramhall,J. 'A defence of true liberty', Works, volume 4, Oxford (1844) Bratman, M. Intention, Plans and Practical Reason, Harvard (1987) Broome, J. Weighing Goods, Blackwell (1991) Davidson, D. 'Intending', Essays on Actions and Events, Clarendon (1980), pp. 83-103 'Replies', Essays on Davidson, eds. B. Vermazen and M. Hintikka, Clarendon (1985), pp. 195-254 Dennett, D. Elbow Room, Oxford (1984) Elster,J. Ulysses and the Sirens, Cambridge (1979) Fischer, J. M. (ed.) Moral Responsibility, Cornell (1986) Frankfurt, H. 'Alternate possibilities and moral responsibility', The Importance of What We Care About, Cambridge (1988),' pp. 1-11 'Freedom of the will and the concept of a person', The Importance of What We Care About, Cambridge (1988), pp. 11-26 'Identification and wholeheartedness', The Importance of What We Care About, Cambridge (1988), pp. 159-76 Gauthier, D. 'Deterrence, maximisation and rationality', Moral Dealing, Cornell (1990), pp. 298-321 Morals by Agreement, Oxford (1986)

'In the neighbourhood of the Newcomb-predictor', Proceedings of the Aristotelian Society 89 (1989) 179-94

Ginet, C. On Action, Cambridge (1990) Harman, G. 'Practical reasoning', Review ofMetaphysics 29 (1976) 431—63 Hobbes, T. 'Of liberty and necessity', Works, volume 4, ed. W. Molesworth, London (1840), pp. 229-78 Leviathan, ed. R. Tuck, Cambridge (1991) Hornsby,J. Atfiowj, Routledge (1980) Hurley, S. Natural Reasons, Oxford (1989) 276

Bibliography

277

Inwagen, P. van An Essay on Free Will, Oxford (1983) Inwood, B. Ethics and Human Action in Early Stoicism, Oxford (1985) Kahn, C. 'Discovering the will: from Aristotle to Augustine', The Question of'Eclecticism', eds. J. Dillon and A. Long, California (1988) Kane, R. Tree will: the elusive ideal' Philosophical Studies, 75 (1994) 25-60 Kant, I. Groundwork oftheMetaphysic of Morals, ed. H. Paton, Harper and Row (1964) Religion within the Limits of Reason Alone, ed. J. Silber, Harper and Row (i960) Kavka, G. 'The toxin puzzle', Analysis 43 (1983) 33-6 'Some paradoxes of deterrence', Moral Paradoxes of Nuclear Deterrence, Cambridge (1987) Lewis, D. 'Are we free to break the laws?' Philosophical Papers, volume 2 (1986), pp. 291-8 'Devil's bargains and the real world', The Security Gamble, ed. D. MacLean, Ottowa (1984), pp. 141-54 Locke, J.An Essay Concerning Human Understanding, ed. P. Nidditch, Oxford (i975) Mele, A. 'Intention, belief, and intentional action', American Philosophical Quarterly 26 (1989) 19-30 'He wants to try', Analysis 50 (1990) 175-8 'Intending for reasons', Mind 101 (1992) 327-33 'Intentions, reasons and beliefs: morals of the toxin puzzle', Philosophical Studies 68 (1992) 171-94 Springs of Action, Oxford (1992) Nussbaum, M. The Therapy ofDesire, Princeton (1994) O'Shaughnessy, B. The Will, volumes 1 and 2, Cambridge (1980) Parflt, D. Reasons and Persons, Oxford (1984) Pink, T. 'Purposive intending', Mind 100 (1991) 343—59 'Justification and the will', Mind 102 (1993) 329-434 Ramsey, F. 'Truth and probability', Foundations of Mathematics and Other Logical Essays, ed. R. B. Braithwaite, Routledge (1931), pp. 156-98 Rawls,J.^4 Theory of Justice, Oxford (1972) Schelling, T. Choice and Consequence, Harvard (1984) Sidgwick, H. The Methods ofEthics, Hackett (1981) Slote, M. 'Selective necessity and free will'', Journal of Philosophy 79 (1982) 5-24 Velleman, D. Practical Reflection, Princeton (1989) 'What happens when someone acts?', Mind 101 (1992) 461-81 Wiggins, D. 'Towards a reasonable libertarianism', Needs, Values, Truth, Blackwell (1987), pp. 269-302 Williams, B. 'Deciding to believe', Problems of the Self, Cambridge (1979), pp. 136-51 Shame and Necessity, California (1993) 'Nietzsche's minimalist moral psychology', Making Sense of Humanity, Cambridge (1995), pp. 65-78

Index

Abelard, P., 177 acrasia and the executive nature of intention, 30-1 and rationality-constitutive practical requirements, 134-5 ACTION, see Action model of decision/intention rationality action co-ordination and executive function of the will, 66-71 how action-coordinatory justifications establish the Action model of decision rationality, 228-46 Action model of decision/intention rationality, 8-11, 144-57 implies ACTION as a principle of decision/intention rationality, 145 how in tension with common-sense psychology of will, 9-10, 149-57 how the tension might be resolved through combining the Action model with REASON-APPLY, I56, 163-5 proved, 228-46 Teleological versions of, 166-79 proof of a non-Teleological version implying REASON-APPLY, 247-68 actions (see also first-order agency), 14-15 bodily, 52-9 compared to decisions, 7-8, 34-52 intentionality of, 219-22 practical primacy of, 184-5, 2 2 3 purposiveness of, 14-15 rationality of, 7-8, 47-52, 138-41, 181-3 as subject to the will, 192-5 Additional Motive theory of decision execution, 121, 125 arguments against, 126-35 agency (see also models of agency) characterised in terms of a distinctively practical reason, 7-8, 20-1, 47-52, first-order, 3-6, 56

second-order, 3-7, 15-18, 21-5, 33-46, 56 not found in third-order form, 205-6 and ethical theory, 45-6 intentionality of, 207, 215-25 sub-intentional, 217, 219 as means-end justifiable, 49—52, 159-62 phenomenology of, 47-8 psychology of, 1-32, 269-75 and purposiveness, 7-8, 14-15, 41-4 non-purposive agency, 49-59, 242-4 rationality of, 7-8, 47-52, 138-41, 159-63 special-purpose agency, 156, 164, 247-68 when subject to the will, 195-200 why agency alone is subject to the will, 199 and Teleology, 166-86 agent-continuants, see identity of agents altruism, rationality of, 174-6 Anscombe, G. E. M., 9, 141 Aquinas, T., 24, 29-30, 104, 177, 215-16 autonomy (see also freedom) and Incompatibilism, 89-93 Autonomy-within-Reason model of agency, see models of agency belief and intention, 19, 70-1, 222-5, 259 and intentional agency, 219-25 practical judgment as a form of, 37-40 belief-dependent effect of decisions, 72-3, 115, 117, 172-3 and deliberation, 73-4 Bennett, J., 160 Bentham,J., 177 blameworthiness, 11—13 bodily actions, 52-9 bounded rationality, 71 Bradley, F. H., 69 Bramhall, J., 25, 79, 215-16 Bratman, M., 7, 9, 19, 68, 70, 125, 133, 135, 141, 216-17, 244, 273

278

Index Broome, J., 139 Butler, J., 177 Campbell,J., 128 Close Analogy assumption (about freedom of will in relation to freedom of action) stated, 189 and Will Subjection theory of freedom, 189-90 refuted, 208-12 commands as presupposing the means-end justifiability of what is commanded, !54-5> l 6 ° to take particular decisions, 154-5, 265-6 to form particular pro attitudes, 160, 273 to perform actions, 155 to try, 155, 265-6 common-sense psychology as a normative theory, 28-9 and Enlightenment psychology, 28-9 of agency, 14-18, 28-9 of the will, 1-2, 18-21, 29-32, 117-21 tension in its conception of the will, 1-2, 9-11, 137-65 Compatibilism, 76-8 and the Psychologising conception of freedom, 88-93 conation experiment, 58, 261-3 conationist theory of agency, 3 conations, 52-9 as first-order agency, 56 compared to decisions, 55-9, 244, 261-7 as possible objects of command, 155, 265-6 as subject to the will, 58-9 as motivation-independent in their effects, 56-7, 264-5 motivation of, 52-9, 143-4, 244, 263 rationality of, 52-9, 261-7 as not governed by any analogue of REASON-APPLY,265-7

Consequence argument, 91-2 Consequentialism, see Teleology control, see freedom Dan's decision as establishing the Action model of decision/intention rationality, 234-44 Davidson, D., 9, 23, 27, 29, 43-4, 47, 51-2, 54, 141,217-18,245,273 decision-commands, 154-5, 265-6

279

decision execution Additional Motive theory of, 125-35 Motivation-Perpetuation theory of, 121-5, 259-60 decision drug taking as first-order agency analogue of decision taking, 95-9, 148-57, 266 as motivation-influencing, 96-7, 266 as neither freedom- nor rationalitypreserving, 97-9,148—57 as not special-purpose agency, 147—8, 257-8 as governed by no analogue of REASONAPPLY, 148—57, 266

decision prizes, 147-59 decision theory, see probabilistic decision theory decisions (see also intentions) as intention formations, 1, 20-1 as second-order agency, 3-4, 15-18 as special-purpose agency, 156, 164, 247-68 as end-specific agency, 252-60 as essentially end-specific agency, 256-60 as essentially intentional agency, 217-19, 222-5 as executive in function, 1-2, 29-32, 66-71, 112-21, 135-6, 146, 149-59, 225-7 as reason-applying in effects, 112-21, 135-6, 146, 149-59, 225-7 as belief-dependent in effects, 72-3, 115, 117, 172-3 as motivation-perpetuating in effects, 6-7, 112-15, 116-25, I5°>lb?>~^ 243, 258-60 as rationality-preserving in effects, 6-7, 93-5, 116-21, 149-54, 196-7 as governed by REASON-APPLY, I O - I I ,

152-9, 165, 185-6, 196-7, 247-68 as determinants of action, 72-6 as not subject to the will, 8, 40-1, 44-5, 195-206 as raising the stake on performance of the action decided upon, 74-5 as adding to our justification for acting as decided, 74-5, 129—33, 238-41 the impossibility of deterrent decisions, 171-6 motivation of, 7-8, 35-7, 41-4, 144-5 phenomenology of, 47-8 rationality of, 8-11, 47-63, 137-86, 228-75 without prior deliberation, 204-5 their effect on deliberation, 73-4

280

Index

decisions (cont.) tensions in common-sense psychological conception of, 1-2, 9-11, 137-65 compared to actions, 7-8, 34-52 compared to conations, 55-9, 244, 261-7 decisions to decide are decisions to act in a decisionaffecting way, 192-206 as decisions to take particular decisions, 192-200 as decisions to make up one's mind, 200-6 as decisions to deliberate, 200—6 deliberate, see intentional deliberation as first-order agency, 203 as wondering what to do, 73 as reviewing a decision, 73-4 deliberative conception of the will, 22-3, 29-32 and the Psychologising conception of freedom, 103-5 refuted, 245 Dennett, D., 26, 47-8, 218 Dependence (of freedom of action on freedom of will) introduced, 2—7, 64—6 whether explained by Incompatibilism, 76-88 as explained by the Psychologising conception of freedom, 101-21 Depsychologising conception of freedom, 80 refuted, 101-21 desirability and Teleology, 139, 166-8 of decision execution, 129-33 desires and action-motivation, 14-15 and the motivation of end-specific action, 248-53 second-order, 34-41 and intentions, 25-6, 34-41, 245, 273-4 and practical judgments, 37-40 as pro attitudes, 59-61 as not subject to the will, 40-1, 195-206 Elster, J., 5, 109 end-derived practical requirements, 134-5 END-SPEC, see end-specific agency end-specific agency defined, 248

END-SPEC used to derive REASON-APPLY,

253 will agency as essentially end-specific, 256-60 Enlightenment psychology, 26 ethical theory and freedom, 11-13 and second-order agency, 45-6 executive conception of the will, 23-4, 29-32 and Psychologising conception of freedom, 106-21 proved, 245 executive function of the will, 1-2, 29-32, 66-71, 112-21, 135-6, 146, 149-59, 225-7 as fulfilled through the reason-applying effect of decisions, m-21 implies REASON-APPLY, 154

first-order agency (see also agency), 3-6, 56 as comprising actions and conations, 56 as source of future action control, 5-6, 108-11 motivation of, 14-15, 52-4 rationality of, 49-55 Frankfurt, H., 12, 27, 34-41, 106, 146, 166, 189—90, 212 freedom (see also Dependence, Survival, Compatibilism and Incompatibilism) as the freedom to do otherwise/control, 1, i5> 8 3"4 as presupposing possibility of deliberate agency, 16-17, 207, 216 as involving a capacity for rational selfdetermination, 93-5, 101-3 and Close Analogy assumption, 189 and determinism, 11-13, 76—93 and moral responsibility, n-13 and practical rationality, 23-5, 93-5, 101-3 possible multiplicity of concepts of, 86-8 Psychologising conception of, 2-7, 79—80, 93-5, 101-36 Will Subjection theory of, 189-90 future action control as essential to freedom, 106—8 as presupposing present agency, 108 through first-order agency, 108-11 through second-order agency, in—21 future actions defined, 66-7

governed by END-SPEC, 251-2

end-specific agency as special-purpose agency, 248-52 decisions as end-specific agency, 252-6

Gauthier, D., 68, 144, 165, 169-76, 179-85, 209, 234 Gauthier-plans, 68, 179-85

Index Ginet, C , 92 goodness and end-desirability, 138—40 and Teleology, 166-9 Harman, G., 68, 70 Hobbes, T., 2, 7-8, 25-9, 32, 33-45, 64-6, 79, 102, 166, 187-91, 212-17, 266, 273 Hobbesian model of agency, 33-45, 212-17 Hobbesian psychology, 25-9, 33-45 Hobbesian theory of freedom, 64-6, 166 holism of practical reason, 66-71 Hornsby, J., 52, 57 identity of agents and practical rationality, 66-71 and the will, 66-121 and self-control, 106-21 Incompatibilism, 76-93 autonomy-centred/Weak, 88-93 time-centred/Strong, 82-8 and Dependence, 76-80 and Survival, 81-93 and the Psychologising conception of freedom, 88-100 intentionality of agency, 207, 215-22 its relation to intention, 215-17 and the Single Phenomenon View, 216-17 of actions, 219-22 of decisions and intentions, 222-5 as condition of exercising freedom, 16—17, 207, 216 and rationality, 215-22 intentions {see also decisions) characterised, 19-21, 117-21, 220 inertia of, 19, 125, 259 relation to agency, 18-21 inherently active, 20-1, 228-46 not assumed to be beliefs about how one will act, 19 effect on belief, 19, 70, 222-5, 259 not desires, 245 not practical judgments, 245 as pro attitudelike, 273-4 in animals, 21—5 when formed through decisions, 20-1 as not subject to the will, 8, 40-1, 44-5, 195-206 intrinsic desirability of decision execution presupposed by the Additional Motive theory, 121 arguments against, 126-35 Inwagen, P. van, 92

281

J-IDENTITY

stated as principle of decision/intention justification, 142 and Pro Attitude model of decision/ intention rationality, 141-4 counterexamples to, 231-41 general falsehood of, 238-44 judgment, see practical judgment justifications as opposed to reasons, 47-52, 138-41 practical and non-practical, 47-52 for actions, 49-50 for conations, 52-5, 264-7 for decisions/intentions, 47-63, 137-86, 228-75 for desires and practical judgments, JUSTIFY

stated as principle of action justification, 139 denied by Gauthier, 181-3 Kane, R., 12, 78-9, 83-4, 91 Kant, I., 4-5, 82, 94, 102-5, T77> 2 7° Kavka, G., 9, 144, 147, 158, 165, 169, 176—9, 267 legislative conception of the will, 94, 102-5 Lewis, D., 144, 169, 176-9, 267 Locke, J., 55-6, 221, 261, 267 Martin, M., 262 means-end justifiability, 8, 49-55 how implies agency, 49-50 implied by purposiveness, 144 implies possibility of purposiveness, 50, *43 consistent with non-purposiveness, 52-5, of what is commanded presupposed by commands, 154-5, 160 implied by subjection to the will, 198 Means-End model of agency, see models of agency Mele, A., 20, 262 Mintoff, J., 133, 151 models of agency: motivation-based Davidsonian model, 41-4 Hobbesian model, 212-15 rationality-based Means-End model, 49—52 Autonomy-within-Reason model, 162 Plural-Justification model, 162-3

282

Index

models of agency (cont.) how motivation-based models support, and rationality-based models defuse, scepticism about second-order agency, 7-8, 41-4, 47~59> l6l~3> 228-46 moral theory, see ethical theory motivation-perpetuation as theory of decision execution, 114, 116, 121-5, 259-60 through tie-breaking, 123-4 through desire-stabilisation, 124-5 explains why REASON-APPLY governs will agency generally, 259-68, 274 helps explain why decisions are rationality-preserving, 114, 120-1, *53-4>259-68, 269 Non-Purposiveness argument (against second-order agency) stated, 41-4 refuted, 245-6 normative theory and common-sense psychology, 28-9 O'Shaughnessy, B., 154-5, 158, 160-1, 265 paradox of deterrence, 169, 177 Parfit, D., 61 phenomenology of decision making, 47-8 plans defined, 68 and action co-ordination, 66—71 plans and Gauthier-plans, 184-5 plan-executory effect of decisions, see reason-applying effect of decisions Plural-Justification model of agency, see models of agency practical judgments as beliefs, 37-40 as pro attitudes, 59-61 not second-order desires, 34-41 not intentions, 245 in animals, 21-5 whether subject to the will, 40-1, 195-206 practical primacy of action, 184-5 helps generate scepticism about secondorder agency, 209-10 used to refute Regress argument against second-order agency, 208-12 practical reason, as basis for theory of agency, 7-8, 20-1, 47-52 Pro Attitude model of decision/intention rationality, 8-11, 141-4 implies J-IDENTITY, 142

implies R-IDENTITY, 143

how in tension with common-sense psychology of will, 159-63 how rationality-based models of agency might resolve the tension with common-sense psychology, 162-3 refuted, 228—46 pro attitudes defined, 59 desires and practical judgments as, 59—61 and intentions, 59, 141-4, 245, 273-4 as not subject to the will, 40-1, 195-206 as not subject to command, 154-5, 160, 273 probabilistic decision theory (PDT) and scepticism about second-order agency, 166-9 PDT in second-order form, 168-79 Psychologising conception of freedom, see freedom psychology, see common sense psychology purposiveness (see also actions and agency) implies means-end justifiability, 144 Ramsey, F., 26, 168 rational decisions to do A as disposing one to do A rationally, see rationalitypreserving effect of decisions rational self-determination, see freedom rationality and freedom, 93-5, 101-36 and second-order agency, 21-5 rationality-constitutive practical requirements, 134-5 rationality-preserving effect of decisions, 6-7> 93~5> 116-21, 149-54* 196-7 and the executive function of the will, 112-21, 135-6, 146 as explained by motivation-perpetuating effect of decisions, 114, 120-1, 149-54, 269 Rawls, J., 68 REASON

stated as principle of action-motivation, 140 denied by Gauthier, 181—3 REASON-APPLY

stated as a principle governing decisionand intention-rationality, 153 as implied by the executive function of the will, 154 as a condition of the rationalitypreserving effect of decisions, 152-4 how in tension with the Action model of decision/intention rationality, 155-7

Index how really consistent with the Action model, 247-68 why governs will agency generally, 259-60 explains the pro attitude-like nature of intention, 273-4 reason-applying effect of decisions as fulfilling the executive function of the will, in—21 as explaining the Psychologising conception of freedom, m-21 as involving belief-dependence, 115, 117 as involving motivation-perpetuation, 114, 116, 118 as involving rationality-preservation, 116, 118 as implying REASON-APPLY, 152-4

reasons as opposed to justifications, 47-52, 138-41 reasons for action, 14-15, 140 reasons for trying, 52-9, 143-4, 244, 263 reasons for desiring, 42-3 reasons for deciding and intending, 7-8, 35-7> 41-4, 144-5. 242-4 Reduction argument (against second-order agency) stated, 33-4 refuted, 244-5 Regress argument (against second-order agency) stated, 44-5, 187-91 makes Close Analogy Assumption, 189-91 assumes Will Subjection theory of freedom, 190 builds, but inconsistently, on executive function of will, 225-7 refuted, 192-227 R-IDENTITY

stated as principle of decision/intention motivation, 142 and the Pro Attitude model of decision/ intention rationality, 141-4 counterexample to, 242-3 general truth of, 243-4 Ryle, G., 26 scepticism about second-order agency, 7-8, 20-1, 25-9, 33-63 encouraged by motivation-based models of agency and defused by rationalitybased models of agency, 7-8, 41-4, 47-59, 161-3, 228-46 and Non-Purposiveness argument, 41-4, 245-6 and Reduction argument, 33-4, 212-15, 244-5

283

and Regress argument, 44-5, 187-227 Schelling, T., 61, 109 Scholasticism, 4, 21-5, 29-30, 101-6, 215-17 second-order agency (see also agency, decisions and scepticism about second-order agency), 3-7, 15-18, 2i-5, 33-46, 56 phenomenology of, 47-8 and rationality, 21-5 Sidgwick, H., 84-5, 119-20, 130, 177 Single Phenomenon View of action intentionality, 216-17 Slote, M., 92 special-purpose agency, 156, 164, 247-68 as not subject to the will, 198 as end-specific agency, 248-52 governed by END-SPEC, 251

Stoicism, 4, 29, 101-6, 273 sub-intentional agency, 217, 219 subjection to the will, 8, 40-1, 44-5, 187-212 characterised, 194 tied to the rationality-preserving effect of decisions, 197-8 whether a condition of freedom, 206-12 why only agency is subject to the will, 199 why actions and conations are subject to the will, 192-6, 261-7 why special-purpose actions are not, 198 why decisions and pro attitudes are not, 195-206 sunkatathesis, 29, 104 Survival (given decisions to act, of one's freedom to act otherwise) stated, 81-2 and the Psychologising conception of freedom, 81-121 in tension with Incompatibilism, 81-93 proved and explained, 119-21 Teleology and agency rationality, 166-9 and decision/intention rationality, 170-85, 234 directly Teleological theories of decision/ intention rationality, 170-9 indirectly Teleological theories of decision/intention rationality, i79-85> 234 and scepticism about second-order agency, 168 toxin puzzle, 9, 147-59 trying, see conations Velleman, D., 39, 128 volitionist theory of agency, 3

284 volitions, as spurious psychological category, 55-9, 221, 261-7 Wiggins, D., 92 will, see common-sense psychology, decisions, and intentions

Index Williams, B., 27, 31-2, 45-6, 160-1, 245 Williamson, T., 117-18 Will Subjection theory of freedom, see freedom