The Structure of
Trust An Evolutionary Game of Mind and Society By
Translation of Shinraino Kozo: Ko...
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
The Structure of
Trust An Evolutionary Game of Mind and Society By
Translation of Shinraino Kozo: Kokoroto Shakaino Shinka Gemu. 信頼の構造：心と社会の進化ゲーム Tokyo: Tokyo University Press. 1998. Translated by
Dedicated to my father, Shunji Yamagishi, and the late mother, Kasumi Yamagishi, who gave me strength to trust myself and others
Acknowledgments It is well understood that a society cannot function if there is no trust among its people. Trust is a lubricant of social relations, making relations possible among people and organizations. An absence of trust will reduce greatly the efficiency of interpersonal relations, including social and economic relations. In this sense, trust is privately possessed social capital that enriches a person's life, and at the same time, publicly possessed social capital making our society an easy place to live. This book is a summary of my research on trust over the past ten years, which has tried to connect the understanding of trust as social capital, the aspect of trust emphasized by social scientists such as James Coleman (1990), Robert Putnam (1993a), and Francis Fukuyama (1995), with the understanding of trust as a form of cognition and behavior, the view shared by psychologists and social psychologists. I owe a debt of gratitude to many people for conducting the research presented in this book. I take this opportunity to express my thanks to all of them. An Abe Fellowship granted to me in 1993 for trust research provided an initial impetus. Up until then, I had been aware of the importance of trust while working on the issue of social dilemmas. Through results of U.S.-Japan comparative experiments on social dilemmas, I had discovered a difference in the level of general trust among Americans and Japanese in a direction opposite to the common sense view of the two societies. However, the time and research funding granted by the Abe Fellowship made me think more carefully about implications of that finding. Being a first-term fellow, I am afraid that I too often caused extra work for the office staff. I am deeply grateful to the staff of the Abe Fellowship office for the sincere care and warm assistance they always extended to me. I was very lucky to have the trust research initiated by the Abe Fellowship adopted as a research project of the Institute for Social Systems Research, a division of the Institute for the Nuclear Safety System, headed by Juji Misumi. As a result, my research on trust became a fully-fledged research project. Without support of the Institute, most of the studies presented in this book, especially the U.S.-Japan cross-societal questionnaire survey would not have been possible. Here I express my deep appreciation to the director, Juji Misumi, and the associate director, Akira Yamada, of the Institute, as well as to the associate researchers at the Institute including Yasuhiro Haruna and his colleagues. In three international workshops on trust and additional domestic workshops, all sponsored by the Institute, I received invaluable suggestions on various aspects of my research from the participants, which aided the development of the emancipation theory of trust introduced in this book. Once again, I must express my thanks to the director, Juji Misumi, and the associate director, Akira Yamada, for making the workshops possible, as well as for their invaluable suggestions and encouragement at the workshops. My thanks also go to those who provided valuable comments at the workshops: Karen Cook of
4 Stanford University, Diego Gambetta of Oxford University, Robert Frank and Mary Brinton of Cornell University, Peter Kollock of the University of California, Kazuo Yamaguchi of the University of Chicago, Frank Miyamoto of the University of Washington, Midori Yamagishi of Osaka International Unversity, Eiji Takagi of Saitama University, Tatsuya Kameda of Hokkaido Unversity, Kazuhisa Takemura of Tsukuba University, Yoshinori Saito and Ichiro Numazaki of Tohoku University, Taro Kamioka of Hitotsubashi University, and Kumiko Mori of Aichi Shukutoku University. The international workshops sponsored by the Institute for Social Systems Research spurred a rapid growth of interest in trust research. In particular, my long-standing collaborator, Karen Cook of Stanford University who attended all three of the trust workshops, in collaboration with Russell Hardin of New York University and Margaret Levi of the University of Washington, held a new series of international workshops on trust with support of the Russell Sage Foundation. I presented early versions of the “emancipation theory of trust” and the “cognitive resource investment model of trust development” at some of those workshops, which provided excellent opportunities for receiving useful comments and suggestions. I thank the three organizers for kindly inviting me to the workshops. Grants from the Japanese Ministry of Education, Science, Sports and Culture supported some of the studies presented in this book. I thank the Ministry for its support. I conducted many of the experiments presented in this book jointly with my students, rather than alone. My collaborators in those studies include Motoki Watabe (currently at Kyoto University), Nahoko Hayashi (currently at Kansai University), Nobuhito Jin (currently at Shukutoku University), Nobuyuki Takahashi (now at the University of Arizona), Yoriko Watanabe (currently at Hokkaido Musashi Women’s College), Motoko Kosugi (currently at CRIEP), Tohko Kiyonari, Masako Kikuchi, and Riki Kakiuchi (currently at Institute for Social Research). Shigeru Terai collaborated in a study that is not cited in this book but constitutes part of the trust research project . I thank all the above former and current graduate students who worked as my collaborators. Special thanks go to Yoriko Watanabe who as coordinator of subjects scheduling was extremely dedicated. Thanks also go to two other graduate students, Yohsuke Outsubo and Masanori Takezawa, for their help in various points of the research. Furthermore, many undergraduate students helped us conduct experiments. While it is impossible to list all of their names here, I am thankful to all the students who helped with the experiments. Moreover, I express gratitude to all participants in the experiments as well as to professors at Hokkaido University who donated their precious class time for our effort to recruit participants, in particular Yasuma Fukuchi and Kiyoshi Moriya. Tatsuya Kameda, my colleague at the Department of Behavioral Science, Faculty of Letters, Hokkaido University, has been my companion in earnest
5 discussions of my ideas and hunches, and kindly has provided me with valuable ideas. Another colleague of mine, the late Hiromi Shinotsuka, not only endlessly provided valuable comments on our research during indefinitely extended graduate seminars, but also took very good care of us in our daily academic life and provided indirect support to our trust research project. I am deeply grateful to these two invaluable colleagues. Concerning publication of this book, Kazue Ito of Tokyo University Press was of enormous assistance, providing me with valuable advice from the perspective of a specialist when I was undecided as to the target of the book. She has helped me on other occasions as well. I thank her here. The series of trust studies presented in this book began in a cooperative paper by myself and Midori Yamagishi of Osaka International University, which was presented by Midori Yamagishi at a conference held in Hong Kong in 1989. My association with Midori had a long history before that time, but the paper was our first joint work. After that, we began to work together on several more papers on trust. During that time, she supported my research life in various ways both as a colleague and a life partner. I thank her for her support.
Table of Contents Acknowledgment
Paradoxes of Trust
Market for Lemons Social Uncertainty and Trust The First Paradox The Second Paradox The Third Paradox Experiments on Sensitivity to Information Experiments on Prediction of Other's Behavior Chapter 2
Multi-faceted Nature of Trust Expectation of Natural and Moral Orders Expectation of Competence and Expectations of Intentions Trust and Assurance Social Uncertainty and Caution General Trust and Information-based Trust Character-based Trust and Relational Trust Summary of the Relations among Concepts Related to Trust Trust and Trustworthiness The Paradoxes of Trust Revisited Chapter 3
The Emancipation Theory of Trust
Trust as Encapsulated Self-interest Trust Produced in Stable Relations Limitations of Reductionist Approach Social Uncertainty and Formation of Commitment Relations Commitment Relations Formation of Yakuza-type Commitment Relations Pruitt and Kimmel's Goal/Expectation Theory
The Tit-for-Tat Strategy Axelrod's Computer Simulation Reduction of Social Uncertainty by Formation of Stable Relations Kollock's Analysis of Rice and Rubber Trades Experiment of Rice and Rubber Trades Network Prisoner Dilemma Experiment Transaction Cost and Opportunity Cost Trust as Emancipator from Relational Confinement Trustworthiness as a Trait of the Selected and Trust as a Trait of the Selecting Incentive and Intention Chapter 4
Security in Japan, Trust in the U.S.
The US-Japanese Comparative Questionnaire Survey Comparison of General Trust between Japanese and Americans Importance of Commitment Relations Importance of Reputation Honesty and Justice Summary of US-Japan Comparative Study Chapter 5
Trust and Commitment Formation
Experiment 1 Cross-cultural Experiment and Cross-Societal Experiment Purpose of the Experiment Procedure Findings Decline in Trust in Strangers Experiment 2 Procedure Findings Relationship with the Theory Experiment 3 Behavioral Measure of Trust The Benevolent Dictator Game
Findings The Role of Experimentation Chapter 6
Trust As Social Intelligence
Selection by the Consequence Social Environment The Game Approach The Evolutionary Game Approach In Search of a Missing Link Advantage of Being Credulous? High trusters Are More Sensitive to Information Prisoner's Dilemma with Variable Dependence The Results of the Prisoner's Dilemma Experiment with Variable Dependence The First Detection Experiment The Second “Detection” Experiment Social Intelligence and General Trust The Cognitive Investment Model of Trust Development The Missing Link Revisited Is Trust Needed? Chapter 7
In Search of a Foundation of an Open Society
Evolutionary Game and Co-evolution Trust and Social Intelligence Trust, Trustworthiness, and Social Intelligence Equilibrium within Individuals and between Individuals Assurance of Security and Network Extension Breakdown of Trust and Breakdown of Assurance of security Subjective Transformation of Assurance of Security to Trust Modern Society, Assurance of Security, and Trust References
Introduction This book is written around the central message that collectivist society produces security but destroys trust. Although various definitions are possible, I define a collectivist society as a society in which group members cooperate at a much higher level with each other than with out-group members, people outside group boundaries. Of course, people in any society cooperate at a higher level with those in their group than with people from other groups. However, I call a society particularly strong in such “in-group favoritism” a collectivist society. The prototypical collectivist society in this sense is a pre-modern, traditional village, or, to put it more generally, a communal society—Gemeinschaft, as it is called in sociology. In the modern world, the family could be considered a collectivist group in this sense. In a collectivist society, mutual cooperation inside the group occurs easily. Insofar as they deal only with their own people, group members feel no need to be very cautious about being cheated or exploited. As an instance of such a collectivist society, think of an isolated village in the middle of mountains in a time when there were no cars, buses, trains, or paved roads giving access. Wary behavior that we are used to today, such as locking the door when we go out, hardly would be necessary there. Houses often would not have locks. People would feel safe even if they did not take precautions by, for example, locking the door. We do not need to travel back in time to a remote mountain village to see examples of a collectivist society in which people do not need to be on guard against each other; such examples can be easily seen in modern societies as well. For instance, in comparing Japanese business practices with those of the West, it often is said that first building a trust relationship with the business partner is of critical importance in Japan, and that doing so takes much time. However, once such a “trust relationship” is built, a simple telephone call can complete a business deal without troublesome exchanges of contracts. The lesson is that it is important to form a “trust relationship” in order to successfully conduct business in Japan. This example indicates the collectivist nature of Japanese business relations that provides assurance of security to people inside the relationships. That is, once a “trust relationship” providing sources for mutual cooperation is constructed, people feel safe and assured in dealing with each other even when there are no contracts providing protection. This is the essence of the first half of the central message of this book—collectivist society produces security. So far, no one will disagree. However, I did not decide to write a book in order to write about such a cliché. Rather, I decided to write a book because of the importance of the second half of the central
10 message. More specifically, I did so because I thought the second half of the message—collectivist society destroys trust—was important in the discussion of how to transform Japanese society for the future. Unlike the first half of the message, the second half is not so obvious. I have published a number of papers about this issue in both domestic and foreign journals, and in edited books (Kakiuchi, & Yamagishi, 1997; Kikuti, Watanabe & Yamagishi, 1997; Kyounari, Yamagishi; 1996, Yamagishi, 1997; Yamagishi, in press; Yamagishi, Cook & Watabe, 1998; Yamagishi, Jin & Miller; 1999, Yamagishi & Komiyama, 1995; Yamagishi et al., 1995, 1996; Yamagishi & Yamagishi, 1994, 1998). However, although such academic publications are suitable for reporting specific survey or experimental results, they cannot convey adequately the complete message suggested by the findings. I always have felt that something was lacking in those papers. I decided to write this book in order to introduce specific research findings to readers, but at the same time to focus new attention on the overall message that emerges from the findings. Although the entire book is needed to give a full account of the message, I begin by explaining its essence in an extremely simplified manner. Please recall the aforementioned example of a traditional village community deep in mountains in the days when there were no trains, buses, automobiles, paved roads, or telephones, radios, television sets, or even mail. Villagers living there typically spent their whole life in the small community. It was hardly necessary for them to be on guard against being taken advantage of, being cheated, being exploited, or being robbed by their neighbors.* In this sense, people could feel safe being in the community. The issue addressed in the second half of the message concerns whether trust can be fostered among such villagers. Intuitively, we might think that such communities are the most favorable environment for fostering trust. Challenging this intuition, the central message of this book asserts that such a community does not constitute an environment favorable for fostering trust. Rather, it inhibits trust from developing. Why is that so? We understand intuitively that in a collectivist society represented by such communities, people tend to feel secure with their own kind in the group, but not to trust unknown outsiders. That implies a difference between assurance of security among compatriots on the one hand, and trust in other people’s human nature in general, trust that goes beyond one’s own group, on the other. That difference is the starting point of this book. The second half of the central message asserts that people who are immersed in secure relationships with compatriots (such as in the community deep in the mountain) have difficulty forming trust toward people in general. *
Translator’s note. The author had in mind traditional Japanese rural communities in the pre-modernization era. Not all traditional communities are collectivist in this sense.
11 I believe that understanding that point, which appears obvious and is often overlooked, has significant implications for the debate over what Japanese society should be in the future. This is because the familiar, collectivist way of consolidating in-group cooperation will not work well in the Japanese society of the future. In other words, in future Japanese society, the traditional practice of closing off relations to the outside and building internal cooperation will damage rather than enhance economic, political and social efficiency. That is a problem reflected in the call for “deregulation,” and a problem especially well recognized in the field of business. As Iwao Nakatani states in The Root of the Japanese Problem (Nakatani, 1990), internal cooperation among those who have special privileges at various positions of the society—such as between business firms or corporations and government bureaucracy or politicians—has characterized Japanese society. The so-called cozy relationship between industry and government is a typical example of such a closed relationship. The cozy relationship protects industry from outside competition and assures it of stable profits. Some day, however, the cost of such a practice must be faced. The public now is seriously concerned over the possibility that it may be too late, as we finally suffer the cost in a depressed level of competitiveness. Hence the current call for “deregulation.” To understand that collectivism destroys trust is important since general trust plays an extremely important role in the transformation of a closed, collectivist society into a more open society. In past research on trust as well as in the popular view of trust, the relation fortification aspect of trust—trust fortifies bonding among people—has been emphasized. In contrast to that traditional emphasis, this book emphasizes another aspect of trust, its relation extension aspect. That is, trust emancipates people from closed relations and leads them to form spontaneous relations with new partners. I am not the only one who suggests the role of trust as the “emancipator from closed relationships” or the “constructor of spontaneous relations.” There are other scholars who make the same claims. Among others, Francis Fukuyama’s (1995) argument merits special attention here. He argues that in societies with a strong tradition of familism, that tradition creates difficulties for the formation of spontaneous groups or organizations because familism hinders development of trust beyond the confines of the family and thus impedes economic development that requires effective management of organizations. Fukuyama regards societies in which development of trust beyond the family is limited due to strong familism as low trust societies. Conversely, he regards societies in which people have trust towards general others beyond the family as high trust societies. Fukuyama cites China, Southern Italy, and France as examples of such low trust societies, and the U. S., Germany and Japan as examples of high trust societies. I do not completely agree with Fukuyama’s argument, particularly on the point that
12 Japanese society is a high trust society—which is opposite to my view, to be introduced later. However, my argument parallels Fukuyama’s in holding that tightly knit groups (such as the strong family in the case of Fukuyama) prevent trust toward general others from developing beyond the confines of the group. My treatment of collectivism may be regarded as a generalization of Fukuyama’s argument of familism, generalized to tightly knit groups in general instead of focusing narrowly on the family. Fukuyama argues that strong family ties prevent trust from developing towards people outside the family. I argue that there is no reason why the trust-suppressing role of strong bonds between people should be limited to the family. Despite the difference between familism and collectivism as the major obstacle to the development of trust, both Fukuyama and I come to the same conclusion that spontaneously formed groups and organizations are important for the effective social and economic functioning of a society, and that trust toward others beyond the confines of family or group is important for the spontaneous formation of groups and organizations. That conclusion is shared not only by Fukuyama and myself, but also by other scholars. For example, Robert Putnam (1993a, b) claims that civic tradition is indispensable for efficient management of democracy. He also argues that general trust beyond the family or the group is central to the civic tradition. He supports this argument by showing that Northern Italy, characterized by a higher level of general trust and a stronger civic tradition, has achieved superior economic development and a more efficient democratic system than Southern Italy, characterized by strong familism. Putnam (1993b) also points out that American society recently has been facing a decline in trust, and warns that the decline in trust not only could hinder economic performance but also could produce a crisis of democracy. My message in this book is basically the same as those of Fukuyama and Putnam. Japanese society has achieved efficiency in social and economic functioning through mutual cooperation within highly cohesive, closed, and secure relations. However, the time has come when opening relations to the outside is more profitable than closing off outsiders. A typical example of this can be found in business relations. Stable business relationships with specific partners have been a crucial aspect of so-called “Japanese management” practices. A stable relationship with a particular business partner permits various types of flexibility; for instance, urgent orders can be met, or only a telephone call suffices for a business transaction. The so-called “lifetime employment system” also can be considered in terms of a stable relationship between employer and employee. However, savings in “transaction costs” gained by dealing only with specific partners in fixed relationships can be larger or smaller than the accompanying increase in “opportunity costs.” Opportunity cost here is defined as the difference between the profit potentially available from dealing with
13 other partners and the profit expected in maintaining the current relationship. In this sense, dealing with only specific partners means paying the opportunity cost. Insofar as one cannot expect large profit from shifting transaction partners, that is, insofar as the opportunity cost of maintaining current business relations is not too large, the commitment formation strategy that saves transaction costs would be a wise business practice that enhances overall profits. On the other hand, in a situation where the opportunity cost is great, such a strategy will reduce overall profits. In Japanese society today, opportunity costs of a stable commitment strategy are increasing rapidly not only in economic or business affairs, but also in diverse areas of society. For example, currently in marital relationships, as in business relationships, opportunity costs of keeping stable relations with the same partner are growing. Of course, the savings in transaction costs conferred by a stable marital relationship are quite enormous even today, but opportunity costs accompanying it—giving up opportunities for re-marrying a better partner or the loss of freedom of living alone—are rising swiftly. Thus, in the future, success in reducing opportunity costs, namely, not missing advantageous opportunities, will have critical importance for being successful in society. Looking at the same phenomenon from the macro, society-wide perspective, the efficiency of the whole society depends on the degree to which opportunities are appropriately matched with talents, that is, the right people are assigned to the right positions. Thus, to use opportunities for individuals effectively and to match opportunities with talents effectively for the entire society, Japanese society needs to abandon the collectivist behavioral pattern centered around the security of stable relations, on the one hand, and distrust in and discrimination toward outsiders, on the other. Furthermore, successful functioning of an “open society,” entailing effective matching of opportunities and talents, requires trust toward general others that extends beyond the group boundary. In this sense, the nurturing of general trust that is not confined within the boundaries of group or relationship is the key to successful transformation of Japanese society from a security-based society characterized by closed and collectivist social relations to a more open type of society in which opportunities play a dominant role. In the face of such a fundamental transformation of society, finding out how to nurture general trust beyond the boundaries of group is one of the most urgent tasks faced by social and human scientists working in or on Japanese society. This book summarizes the results of a series of studies I have conducted ultimately to answer this question, and offers my tentative answers to the question at this stage of research, in an easy-to-read form. As to how to foster general trust, I believe that research findings introduced in this book will prove to be of great significance in that they teach us that high trusters are not naïve or gullible. I present details of that point in Chapter 6. In short, the results of the series of experiments I and my colleagues conducted consistently show
14 that, compared to low trusters, high trusters not only are more sensitive to information suggestive of the trustworthiness of a specific person but also can predict more accurately whether or not the person will perform trustworthy actions. Conversely, those who are neither sensitive to informational clues about the trustworthiness of a partner nor predict accurately whether or not the partner actually performs trustworthy actions tend to presume, without any ground, that “it’s best to regard everyone as a thief,” as stated in a widely used Japanese proverb. The general conclusion derived from these studies is that trusting other people or human nature in general is different from being convinced blindly that other people are trustworthy. Rather, to trust others is to have mental composure on the bases of improved sensitivity and skills for discerning trustworthy from untrustworthy people, and to assume, until proven otherwise, that people are generally trustworthy. Those who do not have such mental composure come to the conclusion that “it’s best to regard everyone as a thief” to protect themselves. This point of view implies that to foster general trust it is not necessary to preach the importance of the virtue of trust. What is necessary is to provide opportunities for acquiring social intelligence in a broad sense of the term. Treating trust as social intelligence, and encouraging investment in the development of social intelligence at both individual and societal levels is the key to successfully nurturing general trust. This is the conclusion derived from the research presented in this book, although it is still only tentative at this stage of trust research.
Chapter 1 Paradoxes of Trust
Everyone agrees that humans are social beings. It is unimaginable to live in complete isolation from society and other people. The importance of trust for humans as social beings, that is, as people living with relationships with other people, cannot be over-emphasized. As to the above statement, some might say, “No, I do not trust anyone else at all.” Some might say: “People are completely untrustworthy—even my relatives, spouse, children, and parents.” However, even those people actually are more trustful of others than they think. Those who completely distrust everyone will find it impossible, for example, even to go shopping. One who never trusts others at all would not pay for goods until he had completed a careful examination of each item. However, there is hardly any grocer who would sell vegetables to that Mr. Distruster, who would not pay until he had cooked the vegetables and verified that they had no problems. Mr. Distruster could not ride a taxi, because there is no guarantee that the driver would not turn out to be a robber. He could not send his child to school because there is no guarantee that the child would not be sexually harassed, or kidnapped on the way to school. Mr. Distruster always would have to be on full guard when he walked on the street because any one of the passers-by might turn to be a mugger, and if he were mugged, no one would be willing to help him. A cursory look at the life of Mr. Distruster gives us an impression of how miserable such a life would be. Take the example of renting a car. Many people, like myself, would not examine every word of the contract closely, nor ask for convincing explanations for each word or phrase not completely understood. In signing the contract without fully understanding each and every word, many people would trust the rent-a-car company and would believe they were not placing themselves in jeopardy. Many people will sign a contract without scrutinizing it, or even without skimming over it. On the other hand, Mr. Distruster in this case would examine the contract in full detail. It would take hours or even days for a lawyer to read it carefully and come to a full understanding of each of the legal terms written there. It is hard to argue that the life of Mr. Distruster who takes hours or days wrestling with the contract before renting a car for just a day is a fruitful life. Lack of trust produces enormous social waste, as well as making personal life miserable as shown in the case of Mr. Distruster. Let me present an example from my own experience, although it represents only a small fraction of the enormous
16 waste produced by distrust in Japanese society. I am grateful for research funds provided by the Japanese government for studies that made this book possible. However, I wish that I had had total control over use of the funds, since I would have used them much more effectively than I did strictly following the regulations established by the government. For instance, a purchase order for foreign books has to go through the department purchasing office to designated brokers, according to current regulations. This way of purchasing books costs two or three times more than buying them directly at a discount price at international conferences with subsequent reimbursement, for example. Moreover, books ordered following regulations may take up to a year to arrive if things go badly. It is often the case that by the time they arrive I no longer need them! All aspects of the research provide numerous examples of this kind. Thus, at least half of the research funds are wasted simply to meet regulations. This is the kind of waste distrust produces. If taxpayers trusted researchers, there would be no need to constrain researchers’ use of research funds through such detailed regulations, with the inevitable waste. Detailed regulations about use of research funds are the consequence of taxpayers’ distrust that research funds would be misused if researchers were permitted to use them freely. Thus, taxpayers pay for their distrust by forcing waste on research funds with scores of regulations. Generally, people attribute inefficiency of government offices to laziness of government officials. That could be the case, but we should not overlook the fact that inefficiency is forced upon government officials by countless rules and regulations. Rules and regulations that make government offices inefficient are the consequence of citizens’ distrust of officials. Enormous wastes are produced by that distrust. Of course, the reason that people do not trust officials is that not all officials are trustworthy. People will be exploited if they trust officials and permit officials to use tax money any way they please. In a society where all people behave in a trustworthy manner and all trust each other, such waste produced by rules and regulations will not exist. Trust thus has critical importance to people as social beings. Reflecting its importance, trust is being studied in almost all of the social and human sciences. With casual reflection, I can think of studies of trust in psychology, social psychology, sociology, economics, political science, law, philosophy, cultural anthropology, and business administration. I do not know of any other research topic studied in so many disciplines. Those disciplines represent almost the whole range of human activities, and thus the fact that trust is studied in them testifies to the important role trust plays in all aspects of human social activities. On the other hand, I must admit that trust research is not as advanced as the number and variety of scholars in so many disciplines might suggest. The meaning of the term “trust” itself differs from discipline to discipline. Furthermore, researchers in a given discipline hardly understand the fact that their use of the term
17 differs from that of researchers in other disciplines. Clarifying differences in the use of the term “trust” is an important task, and I intend to do so in Chapter 2. However, my main goal in this book is not to give an account of the use of the term, or an overview of past trust research. Instead, what I intend to do in this book is to summarize puzzling questions raised in our trust research as “three paradoxes of trust,” and then to reproduce in print the theory-making process we—my colleagues and I—have taken to understand those paradoxes. Of course, this book is not a literal reproduction of how we proceeded through the research. Actual research often starts with intuitive and rather vague ideas, which gradually are refined and finally are forged into a theory through empirical studies such as experiments, questionnaire surveys, computer simulations, and so on. It would be a waste of the reader’s precious time if that process were to be reproduced exactly in the book. On the other way, it would be dull and boring for the reader if only final results were summarized and reported. So, in writing this book I have adopted the strategy of re-creating the research process as a series of puzzles and their solutions. Thus, in the first chapter, I introduce the “three paradoxes” concerning trust. Then, in Chapter 2, preparatory to solving the paradoxes, I clarify various conceptions of trust. In remaining chapters, I present the theory and empirical research conducted to solve the paradoxes.
Market for Lemons As described above, trusting other people saves us time and energy in social relations, and our lives become enriched with such savings. Put more generally, trust is a social lubricant promoting interpersonal and social relations. The understanding that trust enriches social life is a common starting point for trust researchers. An example showing the importance of trust in that sense—an example often used in economics—is the “market for lemons” discussed by George Akerlof (1970). Lemons here are not the fruit, lemon, but an American term for cars with hidden problems. The used car market is characterized by information asymmetry between buyer and seller in which the seller knows problems while the buyer cannot detect them easily. Knowing this, the buyer takes the possibility of hidden problems into consideration while bargaining with the seller over price. A price reflecting the possibility of hidden problems would yield a hefty profit for the seller if the sold car actually were a lemon but only a small or no profit if it were a sound car. A buyer who cannot tell good cars from bad ones will not pay a high price, if she does not trust the seller, even for a car of high quality for which the seller paid a high wholesale price. Insofar as the buyer does not trust the used car salesman at all while negotiating over price and has no ability to discern a lemon from a cream puff, the salesman will make no or very little profit from sales of a car with no hidden problems that comes
18 with a high wholesale price. To make reasonable profits, the seller must sell a nice-looking car with concealed problems. Such behavior on the part of sellers increases the probability of used cars on the market being lemons. In turn, buyers then pay more attention to the possibility of getting a lemon and are willing to pay even less. As this process continues, lemons come to dominate the used car market. In the resulting market for lemons, buyers cannot buy a decent used car and sellers have difficulty finding buyers. It is an undesirable situation for both sellers and buyers that could have been avoided if buyers trusted sellers. The market for lemons characterizes not only business transactions involving the selling and buying of used cars, but also various situations in daily life. In a society where no one was trustworthy, the cost merely of protecting oneself and one’s property would be huge, and time and energy that could be used for other productive activities would be compromised greatly. Furthermore, in a world where no business partner was trustworthy, business based on credit would be impossible and so would be the efficient modern system of business. A common understanding of the social function of trust in this sense underlies trust research conducted by social scientists such as economists, political scientists or sociologists.
Social Uncertainty and Trust The above section gave an example of the market for lemons and pointed out that trust is a lubricant for social relations, a common understanding shared by many social scientists who work on trust. This implies that social uncertainty is a necessary precondition for trust to play a significant role. In the market for lemons, the buyer lacks sufficient information about the seller’s intention concerning selling a lemon, that is, cheating. Seen the other way around, the problem of lemons will not exist if the buyer can judge the seller’s honesty correctly. We define a socially uncertain situation to be one in which information about a partner's intention is needed and yet such information is lacking. It should be noted here that lack of information about a partner’s intention by itself does not make a situation socially uncertain. For instance, new cars seldom have hidden problems and thus there is less opportunity for cheating in the new car market compared to the used car market. This is why social uncertainty is lower in the new car market than in the used car market even when the amount of information about the salesperson’s honesty is the same. This sounds like something not worth mentioning and often is taken for granted. In a situation where social uncertainty does not exist, that is, where a partner makes no profit by cheating or where a partner’s intentions are transparent, whether or not the partner is trustworthy does not matter. However, this obvious point has significant implications for discussions to follow. For this reason, I repeat the point again. It is in situations in which social uncertainty is large that trust is needed. Conversely, in situations in which the possibility of cheating and exploitation does
19 not exist, that is, in which social uncertainty does not exist at all, trust has no role to play.
The First Paradox The argument that trust is needed, or plays an important role, only in socially uncertain situations sounds obvious. Yet it involves a serious problem, since it conflicts with another commonly shared understanding about trust. Another widely shared understanding about trust is the “common sense” idea that trust is engendered in a long-lasting and stable community such as the family. Restated using the same terms we used describing the first common understanding about trust, this idea is that trust is engendered in an environment devoid of social uncertainty. This common sense idea—that nurturing trust needs an environment devoid of social uncertainty—is shared by many people as is the other common understanding that trust plays a role of lubricant of social relations. Suppose we accept the second commonplace that trust is produced in stable relations with little social uncertainty. Then, trust is unnecessary from the very beginning because there is little social uncertainty in such relations. That is, trust is engendered in situations in which trust is not needed; this is the conclusion of accepting both the first and the second common understandings at the same time. This is the first paradox of trust we discuss in this book. On the one hand, trust is most needed in situations of high social uncertainty, situations where trust is most difficult to produce. On the other hand, trust is not needed in stable relations where trust is most easy to produce. These two propositions form a paradox. Each of the two ideas that produce the above paradox is, by itself, a “common sense” view that is shared by many researchers. However, accepting both ideas at the same time yields conclusions that look mutually conflicting. Why accepting the two common premises causes such a seeming paradox is an interesting question.
The Second Paradox The second paradox is that the level of general trust toward other people in Japanese society is much lower than that in American society, yet Japanese society is characterized by widespread networks of stable relations that are extended to every corner of the society, and in which collectivist social relations play dominant roles. The first paradox of trust, introduced in the preceding section, is produced by the combination of two premises generally accepted as common sense (the premise that trust facilitates the formation of social relations under the situation of high social uncertainty, and the premise that trust is produced in relations with low social uncertainty). The second paradox, on the other hand, is based on the second premise alone—that is, the premise that trust is engendered in relations with low social uncertainty. In studies of Japanese society in general, and particularly of Japanese business
20 practices, Japanese society is regarded as one in which social and business practices are based strongly on mutual trust. Often it is claimed that business practices in Japan are based on trust to a much greater degree than in Western societies where business practices are mostly based on contracts. It is also claimed that social relations in Japanese society depend more strongly on trust, or trust plays a more significant role in Japan than in the U. S. This view is so widely shared by so many that it is almost accepted as “common sense.” However, results of US-Japan comparative questionnaire surveys do not support this common sense view of American and Japanese societies. Instead, results consistently show that Americans have a stronger tendency to trust others than do Japanese. For instance, in my own study (Yamagishi, 1988a), in which the respondent’s level of general trust or trust toward general others was measured by an eight item General Trust Scale, the average level of general trust of 852 American students was conspicuously higher than that of 212 Japanese students. Research presented by M. Yamagishi, a collaborator of mine (Yamagishi, M., & Yamagishi, 1989), that reported the questionnaire results for 167 University of Washington students and 165 Hokkaido University students suggested that Americans believe more strongly than Japanese that people are generally honest. A more systematic survey (Yamagishi & Yamagishi, 1994) also reached the same conclusion. This survey used both student samples and general population samples randomly chosen from residents of Seattle in the U. S. and residents of Sapporo in Japan, and showed that Americans had a higher level of general trust than Japanese. The difference was found consistently among students, general citizens, and within each sex. Results of this survey will be presented in detail in Chapter 4. In those studies, a trust scale my colleagues and I developed was used to examine the degree of trust that respondents felt toward other people in general.* Of course, it is legitimate to ask how valid responses to the questionnaire responses were, that is, whether or not respondents’ general trust was measured correctly through questionnaire items. That issue is inherent in any survey using questionnaires, but fortunately, the General Trust Scale used in this research (or earlier versions of it) has been fairly successful in predicting the actual behaviors of participants in a series of experimental studies (e.g., Kakiuchi & Yamagishi, 1997; Yamagishi, 1986, 1988a, 1988b, 1992; Yamagishi & Cook, 1993, Yamagishi & Sato, 1986; Yamagishi et al., 1996). The predictive validity of the scale thus has been demonstrated. That is, what the General Trust Scale measures is not an abstract belief that has no influence on people’s actual behaviors. Rather, results of those studies indicate that what the scale measures (i.e., the respondent’s level of general *
For example, the average of the following six items was used to measure level of respondents’ general trust in Yamagishi and Yamagishi’s survey (1994). They are: “Most people are basically honest.” “Most people are trustworthy.” “Most people are basically good and kind.” “Most people are trustful of others.” “I am trustful.” “Most people will respond in kind when they are trusted by others.”
21 trust) has very strong influence on people’s actual behaviors that should reflect their trust. However, while comparing American and Japanese responses, the first two studies used only student samples from Hokkaido University and the University of Washington, and samples in the third study were limited to residents of Seattle and Sapporo. Considering those limitations in the nature of the samples, “the second paradox of trust”—that Americans have a stronger tendency to trust others in general than do Japanese—may sound like an overstatement. In response, I can point to a similar finding from another study using national representative samples in both Japan and the U.S., conducted by the Japanese Institute of Statistical Mathematics. This questionnaire survey was conducted with representative national samples including 1,571 American respondents and 2,032 Japanese respondents. Among numerous questionnaire items used in this study, the following three items are directly related to trust. First, 47% of the Americans chose the response, “people can be trusted,” to the question: “Do you think you can put your trust in most people or do you think it’s always best to be on your guard?” In contrast, only 26% of the Japanese gave the same answer (in 1978, and 31% in 1983). Similarly, 62% of the Americans answered “not true” to the question: “Do you think that other people are always out to make use of you if ever they see an opportunity, or do you think that’s not true?” Only 53% of the Japanese sample answered the same way (in 1978, 59% in 1983). Moreover, 47% of the Americans responded that “people try to be helpful” to the question: “Would you say that most of the time people try to be helpful or that they are mostly just looking out for themselves?” Only 19% of the Japanese gave the same answer (in 1978, 24% in 1983). Japanese society is considered to have stable and long-lasting social relations that are made up of strong mutual trust. And yet, those studies found Japanese to be less trustful of others in general than Americans. This is the second paradox. The research project on trust that underlies this book was prompted by that consistent finding, which challenges the “common sense” belief that Japanese society is a high trust society.
The Third Paradox The third paradox is that people who have a stronger tendency to trust others in general are not gullible or naïve as they are usually assumed. Rather, they are more sensitive to information potentially revealing trustworthiness of others and are also more accurate in predicting whether or not others will take trustworthy actions. This paradox emerged from findings of the series of experiments that had been prompted by the second paradox. We generally regard trusting others as desirable, but, at the same time, we are apprehensive of the possibility of being cheated or exploited by others. We are concerned about people who easily trust others’ words. When we see someone
22 around us trust others too quickly, we may hint at the possibility of being cheated. We may look down on such a person as too naïve, especially when he or she is not very close to us. From our experiences, we acquire the “common sense” belief that clever people do not trust others easily, or, conversely, that people who trust others too easily are naïve people who know little of the world. The third paradox challenges this “common sense” belief. It is necessary to present some of our experimental results in order to demonstrate the existence of this paradox. Details of those experiments will be presented in Chapter 6. Here, I am going to present the minimal essence of the experimental findings in order to show the third paradox to readers.
Experiments on Sensitivity to Information The third paradox was noticed first due to a series of experiments I conducted with one of my graduate students, Motoko Kosugi (Kosugi & Yamagishi, 1998). The experiments used vignettes, rather than being studies of actual behavior in the laboratory. Each respondent was asked to imagine the situation described in a handout and to indicate what he or she would think in such a situation. With that kind of scenario experiment, there is no guarantee that when actually confronted with a real situation respondents would think and behave in the way they describe in their responses to the imaginary situation, when they say, “If it were I, I would think such and such” or “I would do such and such.” Rather, it is safest not to consider this to be the case. Thus, we must be very careful in conducting such experiments. The biggest mistake would be to interpret results of scenario experiments as a substitute for real behavior. However, this does not mean that scenario experiments are totally meaningless. Such experiments can be a useful tool when the responses themselves yield useful information, instead of being used as a substitute for real behavior. The experiments I present below used scenarios, yet what is examined in those experiments is not an approximation of participants’ natural behavior. Rather, the experiments are used to learn how participants deal with the specific information provided to them. The booklet distributed to the respondents contained 15 scenarios. Each scenario described a situation in which a target person might betray other people’s trust and behave selfishly. Respondents were asked to predict whether or not the target person would act in a trustworthy manner. What was to be examined is whether participants’ estimation of the target person’s trustworthiness would be different between high trusters, who have a high tendency to trust others in general, and low trusters, who have a low tendency to trust others. Respondents were divided into high trusters and low trusters according to their scores on the trust scale score included in the post-experimental questionnaire. In making the estimation of the target’s trustworthiness, information about the target person as well as information about the situation was provided. Examples of
23 the information about target person A are: “A is a person who does not say thanks to someone who has been kind to him or her”; “A cuts into a line at a supermarket.” The purpose of this experiment was to see how estimation of the target’s trustworthiness would change as information about the target person was provided. Two types of information about the target person were provided, namely, positive information potentially revealing the trustworthiness of the target person and negative information potentially revealing the lack of trustworthiness. Figure 1.1 shows two graphs of how estimation of the target person’s trustworthiness changed as positive information was provided (left panel) and negative information was provided (right panel). First, look at the left margin of each graph. It indicates the mean probability estimation by high- or low trusters that the target person would act in a trustworthy manner when no information about the target person was provided.
Probability that target acts in a trustworthy amnner
Probability that target acts in a trustworthy manner
1 Pieces of Information
Pieces of Inforamtion
Figure 1.1 The effect of positive or negative pieces of information on high- and low trusters’ estimation of the target person’s trustworthiness, as shown in Kosugi & Yamagishi (1998), first experiment
24 When no information was provided about the target person, high trusters thought that the target person would act in a trustworthy manner to a greater extent than did low trusters. This means that the general trust scale used to classify the respondents reflected fairly well the degree of their general trust, that is, the extent to which they think that an unspecified partner would act in a trustworthy manner. Having thus demonstrated the validity of the general trust scale, we will investigate how different high trusters and low trusters are in their sensitivity to (positive or negative) information potentially revealing trustworthiness or lack of it. Generally it is considered that low trusters, who tend not to trust other’s benevolence, are more sensitive to negative information than high trusters, who tend to trust others, whereas high trusters are more sensitive to positive information than low trusters. Low trusters, who are always suspicious of others, would regard the target person as not trustworthy given the slightest reason to doubt. Conversely, high trusters would believe that the target person is trustworthy with only little positive evidence. Do experimental results support this expectation based on our common sense? First, the left graph of Figure 1.1 shows that both high trusters and low trusters raised their estimated likelihood of the target person’s acting in a trustworthy manner when positive information was provided. Especially when two pieces of positive information were provided, the difference between high and low trusters in their estimation of the target person’s trustworthiness seems to have increased. As expected, high trusters rapidly increased their estimation of the target person's trustworthiness when information potentially revealing the target person’s nice quality was provided, while low trusters’ estimation did not rise so quickly. However, this differential sensitivity to positive information between high and low trusters did not reach statistical significance. What was the case when negative information was provided? The right graph of the same figure shows that both high and low trusters reduced their estimation of the target person’s trustworthiness when negative information was provided. It further shows a difference between high- and low trusters in how rapidly the estimation decreased. Contrary to the common sense expectation described above, high trusters responded more quickly to negative information than did low trusters, and the differential sensitivity to negative information was statistically significant. These results were consistent with the common sense view that high trusters are more sensitive than low trusters to positive information, although the differential sensitivity was not statistically significant. On the other hand, responses of high and low trusters to negative information were contrary to the common sense view; high trusters responded more quickly than did low trusters. In sum, compared to low trusters, high trusters responded more sensitively to either positive or negative information. More generally, high trusters were more attentive or cautious to
25 potential social risks and were more sensitive to information potentially revealing trustworthiness or the lack of it in other people than low trusters. Though this conclusion seems quite challenging, it is not radically different from previous findings by Julian Rotter and his colleagues that showed that general trust and gullibility was unrelated. Rotter defined trust as “an expectancy held by an individual or a group that the word, promise, verbal or written statement of another individual or group can be relied upon” (Rotter, 1967: 651). He is famous for the Interpersonal Trust Scale (ITS) he developed to measure trust toward general others (Rotter, 1967, 1971). The trust scale I used in my research had been developed partly based on his ITS. After examining studies that had used ITS, Rotter (1980a, b) concluded that one’s level of general trust is independent of one’s gullibility (or naïvite or stupidity). If we accept the research findings Rotter examined, we should not be surprised to see that high trusters, those who have a high level of general trust or trust toward others in general, are not particularly gullible compared with low trusters. However, how can we interpret the above experimental findings that high trusters paid more attention than did low trusters to whether or not others are trustworthy? This is the third paradox of trust. The reader might ask, though, “Is this really true?” because the findings are so radically different from our common sense beliefs. The reader might say: “I don’t believe the findings. They may have happened that way for some accidental reasons.” It would not be surprising if such questions occurred to the reader since I myself had such questions when I first saw the result. I said to myself that I would only pay serious attention to this paradox if I conducted the same experiment again and obtained the same result. Accordingly, Motoko Kosugi and I conducted a second experiment (Kosugi & Yamagishi, 1996). The second experiment used almost the same procedure as in the first experiment. There were only a few differences. For example, we increased maximum number of pieces of information from two to three, and the number of scenarios from 15 to 16. The number of participants was smaller (75 instead of 257 in the first experiment). The second experiment largely replicated the findings from the first experiment, especially the most important finding that high trusters were more sensitive than low trusters to negative information and lowered their estimation of the target person’s trustworthiness more rapidly. That is, when negative information about the target person was provided, high trusters reduced estimation of the target person’s trustworthiness more rapidly than did low trusters. As a result, high trusters’ estimation of trustworthiness fell below low trusters’ estimation as negative pieces of information were provided. Results from these two experiments were consistent on three key points. (1) High trusters tended to estimate the target person’s trustworthiness more than did low trusters when no information about the target person was provided. (2) When positive information about the target person was provided, both high and low
26 trusters increased their estimation of the target person’s trustworthiness, and little difference was seen between high- and low trusters. (3) However, when negative information was provided, high trusters responded more quickly than low trusters by reducing their estimation rapidly. In sum, the experimental results suggest that high trusters follow the policy, “no witness, no punishment;” they do not suspect others without ground. On the other hand, given suspicious cues they quickly become cautious. In contrast, low trusters follow the policy of regarding “everyone as a thief;” they are suspicious of others without any ground. On the other hand, they do not pay much attention to information that in fact provides reason to be suspicious.
Experiments on Prediction of Other’s Behavior Results of the two experiments introduced above conflict with our intuitive view of high trusters and low trusters, and yet, I would not have used such an extreme expression as the third paradox of trust solely based on those findings. Findings from two additional experiments justify use of the expression. I conducted the following experiment in cooperation with my students, Masako Kikuchi and Yoriko Watanabe (Kikuchi, Watanabe & Yamagishi, 1997). Participants in this experiment were asked to decide whether to give or not to give his or her money to a partner in a situation called “Prisoner’s Dilemma.” The experimenter provided participants with an 800 yen (about $8) endowment before the experiment started. Each participant then played a Prisoner’s Dilemma game with a series of partners, who were other participants in the study. In each game, each participant was asked to decide whether to give 100 yen to the partner or to take 100 yen from the partner. If a participant gave 100 yen, the experimenter doubled its value then awarded it to the partner. Thus, the partner received 200 yen if a participant gave 100 yen. If both gave 100 yen, both received 200 yen. This yielded a net profit of 100 yen for each, since each gave 100 yen and received 200 yen. On the other hand, when a participant took 100 yen from the partner, the money, 100 yen, became the participant’s. The experimenter also doubled the loss to the partner, so that the partner suffered a loss of 200 rather than 100 yen. Thus, if both took 100 yen, each earned 100 yen and at the same time suffered a loss of 200 yen, resulting in a net loss of 100 yen. This is the nature of the Prisoner’s Dilemma used in this experiment. Prisoner’s Dilemma refers to a relationship in which both people prefer the outcome of mutual cooperation to that of mutual defection, and yet for each person defection yields a more desirable outcome than cooperation. In such a relationship, those who pursue only their own self-interest will choose to defect. If both choose to defect, they have to live with the outcome of mutual defection, which is less desirable than the outcome of mutual cooperation. The purpose of this experiment was to examine participants’ accuracy in
27 judging whether other participants gave their money to the partner (Cooperation) or took money from the partner (Defection) in the Prisoner’s Dilemma. For this purpose, before they participated in the Prisoner’s Dilemma experiment, participants took part in a group discussion on garbage collection issues with their future Prisoner’s Dilemma partners. They were told that the discussion session was a part of another study, and was totally independent of the experiment that followed. In fact, the real purpose of this “discussion experiment” was to give participants an opportunity to get to know each other and learn about each other’s character traits. In the group discussion, which lasted about 30 minutes, would participants perceive cues sufficient to predict others’ choices in the Prisoner’s Dilemma? This is one of the questions we wanted to learn the answer of in this experiment. Another question we wanted to pursue—and this question was more important than the first—was whether a participant’s level of general trust would have any relationship with how accurately he or she predicted other participants’ choices. The common sense view that high trusters are naïve and credulous people who can be deceived easily implies that the less naïve and credulous low trusters will be able to judge others’ choices more accurately. On the other hand, from the previous experimental finding showing that high trusters are more sensitive than low trusters to information about others, we would predict that high trusters will do better at judging others’ choices. The most important purpose of this experiment was to see which of the two predictions would hold. The experiment was conducted in the following way. While the experiment included two conditions, I present only one of them here, the “partner unknown condition.” The other condition will be presented and discussed in Chapter 6. In the “partner unknown condition,” each participant was informed that he or she was paired with one of several other participants, but was not told who the specific partner was. The participant then was asked to decide whether to give 100 yen to or take 100 yen from the unknown partner. Thus, in this condition it was impossible to base the choice between giving and taking on personal liking or disliking of the partner, in such a manner as: “I will give 100 yen to B since B is pretty,” or, “I will take from C since C is arrogant.” Thus, in this condition, the decision of giving or taking 100 yen was expected to reflect the participant’s general character rather than his or her attitudes toward specific partners. Those who give 100 yen to the partner in this condition are considered to be those who have a strong tendency for forming cooperative relationships with others, or those with a tendency to be altruistic. As will be explained in Chapter 6, anonymity in the decision of giving or taking 100 yen was guaranteed almost completely not only to other participants but also to the experimenter. After having made the decision to give or take 100 yen, participants in the partner unknown condition were informed that they had been matched with two partners. They were told with which two of the other five participants they had been
28 matched. Then, for each of the two partners, they were asked to predict whether the partner had decided to give or take 100 yen in the Prisoner’s Dilemma experiment. To induce the participant to take the prediction seriously, an additional bonus of 100 yen was provided for each correct prediction. Now let’s examine the result of this experiment. The result shown in Figure 1.2 indicates that high trusters were more accurate in predicting other participants’ behaviors (giving or taking 100 yen) than medium trusters or low trusters. Prediction accuracy toward the two partners was calculated by adding 0.5 points when a partner acted as predicted (the partner gave 100 yen as predicted or took 100 yen as predicted) but 0 points when the prediction failed. With this method, the participant received 1 point when his or her predictions for both of the two partners were correct, 0.5 point when one prediction was correct and the other was wrong, and 0 when neither prediction was correct.
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Low Trusters
Figure 1.2 Accuracy of prediction of the other participants’ choices in the partner unknown condition in the experiment by Kikuchi, Watanabe & Yamagishi
29 Figure 1.2 shows the existence of differences in the accuracy of prediction as a function of general trust of the judge as measured by the general trust scale. That is, compared with medium trusters or low trusters, high trusters were more accurate in predicting other participants’ behaviors with whom they had had a short discussion. The difference was statistically significant. Against the common sense view that people who trust others are naïve and gullible, this result indicates that high trusters are in fact prudent people who are more sensitive than low trusters to information potentially revealing other people’s trustworthiness. This result is consistent with those of the previous experiments. The result of this experiment was confirmed further by the result of the second experiment to be presented in Chapter 6. It is thus a fairly robust result. The third paradox is not based on the result of a single experiment; rather, it is grounded on findings from several different experiments. Now, we have seen the three paradoxes this book addresses. Let us move on to the challenge of solving the puzzle raised by the three paradoxes.
Chapter 2 Conceptual Clarifications
In Chapter 1, I presented three paradoxes of trust. In so doing, however, I did not define the central concept, trust. There was a reason not to do so. It is because the three paradoxes would no longer be paradoxes once a clear definition of trust was provided. The purpose of Chapter 2 is to explain this point by sorting out various concepts related to trust. This will show us that the paradoxes are generated by our failure to distinguish various conceptually independent aspects of the term, trust. Multi-faceted Nature of Trust First, let us examine everyday use of Japanese words related to trust such as shinrai, shinyoh, shinrai suru (the verb form of shinrai), and shinyoh suru (the verb form of shinyoh). 1. I had a tooth ache last night. I had spent a lot of money to get my teeth treated. I was very angry over this and I will not trust the dentist any longer. 2. I decided to buy a personal computer to do some work at home. I saw an advertisement by a manufacturer whose name I had never heard of. Judging from the advertisement, they are selling high quality computers at low prices by mail order. Reputable manufacturers that can be trusted are selling similar computers at much higher prices. I cannot decide whether to buy a computer from a manufacturer with a reputation of trustworthiness or from the mail-order company. 3. My daughter brought her fiancé home and told me that she wanted to marry him. After talking to him, I found him to be honest and trustworthy. I feel safe in entrusting my daughter to him. I could keep mentioning additional examples, but let us stop here and consider the above examples. It is obvious that the same word, “trust,” is used with different meanings in the three examples. In the case of the dentist, what mattered was not some intention of the dentist to hurt me, but the dentist’s inability to do the job right. In contrast, in the case of “my” daughter’s fiancé, the issue is not his ability but whether or not he is a person of malicious character. The difference between the two, trust in ability and trust in intention, is easy to understood. A chef who is skilled in cooking delicious and yet highly poisonous fugu (globefish) may or may not have character traits desirable in the fiancé of my daughter.
31 Present in trust in the reputable manufacturer is an additional element that is not included in trust in the fiancé or the dentist. The consumer trusts products of a reputable manufacturer, but not of an unknown manufacturer, because reputation is an asset that took the reputable manufacturer many years to acquire. Thus the manufacturer will not dare to ruin its reputation with defective products. Products of a reputable manufacturer can be trusted, not necessarily because its managers have noble qualities but because they would suffer loses if they did anything to ruin their company’s reputation. Thus, we see there are differences among trust in the dentist, the fiancé and the manufacturer. However, this does not mean that they have nothing in common at all. In all of these cases, one feels safe in entrusting somebody with something (such as one’s tooth, daughter, or computer). If we keep mentioning examples of trust in action, we will be able to reach the “core” of trust, in the sense of elements common to all examples. Then, we might define trust in terms of that “core.” However, I do not consider it productive to define trust in that way. The reason will become clear as you continue reading the rest of this chapter. My approach to defining trust is thus fundamentally different from that of Luhmann (1979) and Barber (1983), who attempted to define trust through what was common to different uses of the term.
Expectation of Natural and Moral Orders From the above examples, we can see that the word “trust” involves various meanings. The broadest definition of the term would be that of Barber (1983) who defined trust, based on the definition of Luhmann (1979), as “expectation of the persistence and fulfillment of the natural and the moral orders.” That is, trust is defined as the conviction that the orders or rules existing in this world will not be broken easily. If we define trust using what is common to all uses of the term, almost anyone will end up with a similar definition. According to that definition, I believe that the sun will rise again tomorrow because I trust the sun. However, that definition of trust does not appeal to those who live in the world of the Japanese language, since we do not use such words as shinrai or shinyoh to refer simply to conviction. Did Luhmann include expectations of both natural order and moral order in the definition of trust because he was used to a culture in which trust is used interchangeably with conviction and thus he failed to see the fundamental difference between the two types of expectations? Probably not. He included the two types of expectations in the definition of trust because there is a commonality between the two, and he thought that the commonality is the core of trust. The commonality is simplification of complex reality. Humans’ ability to process information is limited such that it is impossible to process all information in the real world. People usually process information by simplifying it in certain ways. Luhmann attributes the significance of trust to its
32 ability to help people find and maintain simplifying rules out of a complicated reality. For Luhmann, trust is a mechanism, which cognitive psychologists call “cognitive parsimony,” for simplifying information processing. From that perspective, I can certainly use the same term “trust” to describe my belief that “the sun will rise again tomorrow” and that “my wife is not having an affair.” However, the position I advance in this book is fundamentally different from that approach. I do not consider that the essence of trust lies in “cognitive parsimony.” I will discuss that issue in later chapters. Instead, I take the position that trust is produced by deliberate information processing rather than by simplifying information processing. A major purpose of this book is to explain that point. In this book, I propose excluding expectations of natural order from the definition of trust. Expectations of natural order and expectations of moral order may share one element in common—i.e., they provide a mechanism for cognitive parsimony—but they do not share anything else, neither their causes nor the roles they play in social relations. If we take the point of view that “cognitive parsimony” is not critical for the understanding of trust, there is no reason to treat the two expectations with the same concept, trust.
Expectations of Competence and Expectations of Intentions Based on the reasoning above, this book abandons the broadest definition of trust proposed by Luhmann and Barber. Barber himself did not deal with the broadest definition of trust in his actual analysis, but only with expectations of moral and social order. However, it should be noted that even a definition of trust in terms of expectations of moral and social order still involves qualitatively different elements. The first distinction to discuss here concerns the difference between trust as expectations of competence and trust as expectations of intention. This difference corresponds to the two sub-categories of trust as expectations of moral and social order proposed by Barber. He argued that there are two kinds of trust even when trust is defined as expectations of moral and social order, and that failure to distinguish between the two has been a major source of confusion in past discussion of trust. Barber’s two subcategories of trust are: (1) “Expectation of technically competent role performance from those involved with us in social relationships and systems.” (2) “Expectation that partners in interaction will carry out their fiduciary obligations and responsibilities, that is, their duties in certain situations to place others' interests before their own.” My colleagues and I have proposed a distinction similar to Barber’s between two types of trust: (1) trust as expectations of a partner’s competence and (2) trust as expectations of a partner’s intentions (Yamagishi & Komiyama, 1995; Yamagishi & Yamagishi, 1994; Yamagishi et al., 1995). In short, it is the distinction between
33 expectations of one’s ability to perform properly what one has said one would do, and expectations of one’s intention to do what one has said one would do. The two sources of persuasion discussed in the persuasive communication literature, the competence of the persuader’s specialty and the trustworthiness of the persuader’s intention, also correspond to the two types of trust Barber and I discuss. When we board a plane, we believe that the pilot has competence sufficient to control the plane. This is trust as expectations of competence. On the other hand, when a wife believes that her husband is not having an affair, she is not thinking that her husband does not have enough competence to have an affair. There may be such cases, but mostly the wife’s expectations are about the husband’s intention—“I believe that my husband is not having an affair since he is not such a person,” or “…because he loves me”—not about his competence. This corresponds to trust as expectations of intention. In daily life we speak of these two types of trust—i.e., expectations of competence and expectations of intention—interchangeably without paying much attention to the distinction. It is often the case that failure to distinguish between them does not cause much of a problem. And yet, there are occasions in which a failure to make that distinction causes serious misunderstandings. An example is residents’ concerns about nuclear power plants (Watanabe, Haruna & Kitada, 1994). A plan to build a nuclear plant near their homes naturally would raise their concern. The power company or government would announce the safety of the plant through various media and would try to persuade people that the multi-level safety system would eliminate the risk of radioactive hazards to the residents. Here, the power company or government considers that the issue is the technical competence of the nuclear plant, and that the residents have doubts about the technical competence of the nuclear plant. It is often the case, however, that efforts by the power company or government to disseminate information about the safety of nuclear power plants turns out to be of no use in addressing concerns of residents. Such efforts often fail because residents have doubts about the intentions of people who disseminate the safety information. They suspect that the power company has an incentive not to reveal the truth about the safety of nuclear plants. While the power company views residents’ trust in the power plant as an issue of technical competence, residents view it as an issue of intentions. Given this difference, it is natural that the power company and residents talk past each other. Trust as expectations of competence and trust as expectations of intentions do not share either causes or results in common. The lack of commonality between the two types of trust is evident to anyone who looks for commonalities between trust in a fugu chef and trust in a clean politician. The only overlap between the two types of trust is that people are assured of safety. While it is possible to call whatever produces assurance of security in interpersonal relations “trust,” what produces the
34 assurance is almost completely different for the two types of trust. Thus, I take the position of distinguishing the two, and I restrict the term “trust” to the second type—i.e., expectations of intentions. I would like to make clear, before presenting my arguments, that discussions of trust presented in this book cannot be applied to trust as expectations of competence. This does not mean, though, that trust as expectations of competence is not important. What we need to do is to place each of the two in the appropriate context so that we can avoid unnecessary confusion.
Trust and Assurance The above discussion has largely demarcated the range of trust to be discussed in this book. Trust as discussed in this book concerns a partner’s intention to take advantage of you for his or her own benefit, such as deceiving you or exploiting you. Having circumscribed the range of trust in this way, we should notice that there is another important distinction within this range—i.e., the distinction between assurance of security and trust. This is an important distinction within trust as expectations of intentions that has not been noticed clearly in past studies of trust. When our action is based on trust in someone, we incur a risk of being taken advantage of by that person, risk that could be avoided by not trusting. Here, risk refers not only to risks to one’s body or life, but also risks to wealth, reputation, self-esteem, etc.—risk to self-interest in the broadest sense. If there is no risk involved in acting in a trustful manner, the action involves no trust. I explained this point in Chapter 1 with the example of social uncertainty in the market for lemons. In the market for lemons laden with high social uncertainty, if you take the word of a salesman at face value you probably will feel regret a few days or weeks later. Trusting someone is expecting him or her not to act in a selfish manner despite the existence of social uncertainty, that is, in a situation in which his or her selfish behavior would cause you loss. What makes people trust someone in this sense, then? People would trust a person who does not appear to be vicious or selfish. People would trust a person when they are convinced that the person has a strong conscience. The decision to trust a person or not thus depends to a large degree on evaluation of that person’s character—whether that person has the kind of character that makes him or her behave in a trustworthy manner, even sacrificing his or her own self-interest. More generally, the expectation that a person does not intend to act in an exploitative manner depends on information about his or her behavioral disposition. However, evaluation of other people’s behavioral dispositions is not the only source of expectations for benign behavior in socially uncertain situations. There are cases in which you can expect a person not to exploit you even if you know that he or she is mean and has no warm feelings toward you. These are cases in which the mean person knows that acting in an exploitative manner toward you produces an
35 outcome that is against his or her own self-interest. To understand this point, let’s consider a fictitious example of “a thousand needle machine.” This machine is surgically implanted in a person’s throat, and sends a thousand needles to the throat automatically whenever the person breaks his promise.* Imagine that “a thousand needle machine” has been implanted in someone’s throat. One thousand needles are forced into his throat as soon as he breaks a promise. Even a person who has broken promises millions of times would rather keep a promise than have a thousand needles forced in his throat, once the machine has been implanted as punishment. In this case, your expectation that the person implanted with a thousand needle machine does not intend to tell a lie is not based on inference of his noble character or his warm feelings toward you, but on his self-interest. Trust in this sense, that is, trust based on evaluation of a person’s gains and losses derived from his or her own action, differs greatly from trust as expectations of the person’s intentions based on his or her character or behavioral disposition. In a popular classic Chinese novel, a Buddhist priest, Sanzang, is sure that Supermonkey Sun will not betray him. Sanzang’s trust in Sun is not based on nobility in Sun’s human (or monkey?) nature. Rather, Sanzang’s trust is based on knowing the magical phrase controlling Sun’s head ring. Sanzang has total control over Sun’s behavior since as soon as Sun shows unwillingness to follow an order Sanzang can make Sun suffer by chanting the phrase, causing the ring to shrink. In this example, it is clear that knowing Monkey Sun’s character is hardly the same as knowing how to shrink the magic head ring. Knowing how to shrink the magic head ring will not make Sanzang think of Sun as a noble being. If knowing the magical phrase made Priest Sanzang think highly of Sun’s character, he would have made a big mistake in understanding monkey Sun. Similarly, if we keep using the same word “trust” to describe both expectations of benign intentions based on evaluations of someone’s character or behavioral disposition on the one hand, and expectations of his benign behavior based on his self-interest on the other, our discussion will be hopelessly confused. In fact, past research on trust has been filled with such hopeless confusion. I am proposing, in this book, to adopt another term—assurance or assurance of security—to distinguish the two aspects of trust. Assurance or assurance of security in this sense is the portion in the expectation of someone’s benign behavior that is rooted in the evaluation of his self-interest. For instance, the expectation that a person who has been implanted with a thousand needle machine will not break promises is not trust but assurance according to this definition. In contrast, trust is the portion in the expectation based on evaluations of his character or his feelings toward you. *
Translator’s note: Japanese children guarantee promises by proclaiming; “I will swallow a thousand needles if I tell a lie.”
36 I have repeatedly emphasized that trust requires the existence of social uncertainty to be meaningful. In contrast, assurance as defined above is the subjective reflection of a situation in which social uncertainty has been eliminated. For instance, in the Mafia world with “Iron Rules,” the boss need not know if his soldier is a person of noble quality or heartily loyal to him; he only needs to teach the soldier that anyone who cheats him will be executed soon. As long as the boss is powerful, nobody will dare to betray him unless secrecy is assured for the betrayal. In this example, the boss eliminates social uncertainty in his organization by enforcing “Iron Rules.” Those examples should be sufficient to make clear the fundamental difference between assurance and trust. Trust as defined here is expectations of benign behavior from someone in a socially uncertain situation because of beliefs about the person’s dispositions (including his feelings toward you). Assurance, on the other hand, is the belief that such social uncertainty does not exist.
Social Uncertainty and Caution In situations in which social uncertainty does not exist or is thought not to exist, people will feel it unnecessary to be cautious in interaction with others. In situations in which social uncertainty exists, it is necessary to be careful in dealing with others. This means that situations in which trust is meaningful are also situations that require us to be cautious. Conversely, trust is not needed in situations in which assurance of security is provided, or in which we do not need to be cautions in dealing with others. This suggests that trusting others and feeling a need to be cautious in dealing with others do not necessarily constitute two ends of the same dimension. In fact, it has been repeatedly shown in factor analysis studies that trust toward general others and belief in the need to be cautious in dealing with others do not constitute two extremes of the same factor. For instance, a colleague and I performed factor analysis on our U.S.-Japan comparative questionnaire survey data (Yamagishi & Yamagishi, 1994) which revealed a factor consisting of items measuring degree of trust toward general others, such as: “Most people are basically good and kind.” “Most people are trustworthy.” “Most people are basically honest.” “I am trustful.” “Most people will respond in kind when they trusted by others.” That factor was independent of another factor representing belief in the need to be cautious to defend oneself in social relations, consisting of items such as: “In this society, one has to be alert or someone is likely to take advantage of you.” “People usually do not trust others as much as they say they do.” “One can avoid falling into trouble by assuming that all people have a vicious streak.” “In this society, one does not need to be constantly afraid of being cheated” (reversed item). “There are many hypocrites in this society.” “People are always interested in their own welfare.” “No matter what they say most people inwardly dislike putting themselves out to help others.”
37 Two factors similar to those seen in the above research have emerged repeatedly in other studies. For example, Kaplan (1973) conducted a factor analysis of Rotter’s Interpersonal Trust Scale and obtained three factors. Of the three factors, two are the “sincerity” factor, which corresponds to the general trust factor in the above study, and the “caution” factor, corresponding to the caution factor in the above study. Items with high loadings on the sincerity factor in Kaplan’s study include such items as: “Most idealists are sincere and practice what they preach.” “Most people answer public opinion polls honestly.” Items with high loadings on the caution factor include such items as: “In dealing with strangers one is better off to be cautious until they have provided evidence that they are trustworthy.” “In these competitive times one has to be alert or someone is likely to take advantage of you.” A similar pair of factors—a general trust factor and a fear of exploitation factor—have been obtained in other studies (Yamagishi, 1988a; M. Yamagishi & Yamagishi, 1989). These factor analyses suggest that general trust that people are generally trustworthy, and the belief that one can be taken advantage of unless one is alert (belief in a need to be cautious) are not at opposite ends of a single dimension. Those who think that people in general are trustworthy are not necessarily careless in social relations. Similarly, those who feel a need to be cautious in dealing with others are not necessarily distrusters who presume that everyone is a thief. This is the conclusion of the factor analysis studies.
General Trust and Information-based Trust We often assume that high trusters with a strong tendency for trusting others are naïve and gullible people who are easily taken advantage of by others. The third paradox introduced in Chapter 1 challenges this “common sense” belief. Many studies, including the ones presented in Chapter 1, have demonstrated that high trusters are not necessarily gullible. The key to understanding this seemingly counter-intuitive finding is in the distinction between general trust as default expectations of other people's trustworthiness (that is, how trustworthy a character people generally have) and information-based trust, or trust based on information about specific others. In Chapter 1, I introduced the Interpersonal Trust Scale that Rotter developed and the General Trust Scale that my colleagues and I developed. What those scales measure is not the respondent’s trust in specific others, but his or her tendency to trust other people in general. Thus, trust toward general others instead of specific others is called “general trust.” General trust can be defined as trust toward someone about whom no concrete information is present, or trust toward someone about whom you know nothing except the fact that he or she is a human being. When one estimates the trustworthiness of someone about whom one knows nothing, one has to use the trustworthiness of people in general as a “default.”
38 In cases in which specific information about a certain person exists, no doubt the information will be used in judging the trustworthiness of that person. As introduced in Chapter 1, Rotter demonstrated that high trusters are not necessarily gullible. He defines gullibility in terms of one’s inability to make good use of information suggesting untrustworthiness in others. According to his definition, gullible people are those who are insensitive to information suggesting lack of trustworthiness in others. People who are not gullible are those who are sensitive to such information. It can then be concluded that whether one trusts or distrusts others when no information about them is available (a reflection of the level of one’s general trust) and one’s ability to use information suggesting the level of trustworthiness in others (one’s gullibility depends on this ability) are at least logically independent from each other. Let us call the latter, that is, evaluation of others’ trustworthiness through use of specific information about specific others, “information-based trust.” The above discussion defining general trust as default expectations of others’ trustworthiness, however, does not exclude the possibility of understanding general trust as information-based trust in which the information on which to base trust concerns people in general rather than specific persons. According to this view, general trust differs from character-based trust (i.e., a form of information-based trust, to be discussed next) only in the range of information underscoring trust, not in the nature of trust itself. With the view that general trust differs from character-based trust only in the range of relevant information, it is possible to think of types of trust intermediate between the two. For instance, trust in a certain category of people based on information about them would be situated between character-based trust based on information about specific individuals and general trust based on information about people in general.
Character-based Trust and Relational Trust Three types of information can be used as the basis of your trust in a specific person—information (1) about the person’s general human nature, (2) about the person’s feelings and attitudes toward you, and (3) about the person’s incentives. As we have already seen, the third one—the person’s incentives—corresponds to the thousand needle machine we discussed in distinguishing between trust and assurance. We need not discuss again implications of the third type of information. The following section discusses the other two types of information. First, let us consider information-based trust when the information on which to base trust is information about a particular person’s general character. One may trust someone because one believes that the person is someone of integrity or that the person is no way a bad fellow. This kind of information might have come directly from one’s own experience interacting with the person, or may have been obtained indirectly. Most people would trust a person whom they have known for a long time
39 and whom they have never seen or heard act in a base manner. People often trust such a person when they are convinced that they have enough knowledge about trustworthiness as part of the person’s human character. Information revealing someone’s trustworthiness as a character trait is not limited to information obtained from direct contact with that person. Other types of information also may be used to infer someone’s general trustworthiness, including information other people provide about the person or information inferred from the person’s social status, social roles or qualifications. Among many sources of such indirect information about someone’s general trustworthiness, the most important is the person’s reputation. People would not trust someone completely based only on his reputation. But reputation would be important as a screening device in cases when one wonders whether to socialize with someone one does not know. Furthermore, some category markers that are easily observable from outside also are used in assessing someone’s trustworthiness, often in combination with the person’s reputation. The most frequently used marker is social category. In particular, categories that are considered to reflect a person’s character traits to some degree are likely to be used as information in assessing his or her trustworthiness. Typical examples of such social categories are social statuses and social roles. Furthermore, prejudices or stereotypes based on a social category also can be used in estimating a person’s trustworthiness. Inference of a person’s trustworthiness based on his or her social status can be a kind of stereotype. In the case of social status, however, some foundations for the inference (relevant to some extent) often are involved. For example, a medical doctor may be inferred to be capable of self-control since it would be hard for anyone who has no self-control to pass the strict examinations for certification as a doctor. Sometimes, however, people attribute trustworthiness to a social category without any reasonable ground. Let us call such a case stereotype-based trust. Examples of this would be beliefs that people from such and such a town are honest and trustworthy and people from such and such a town are shrewd. Now, we have dealt with the first type of information that produces information-based trust, information used in assessing trustworthiness as a person’s character trait. We call trust based on that type of information “character-based trust.” Let us be reminded, though, that the person in whom we trust is not necessarily a person of noble character. Sometimes we trust someone whom we know to have done awful things to other people; we trust him not to betray “me” although he may be mean to other people. For instance, the James brothers in the Wild West knew that they were cruel to other people, but they trusted that they never would betray each other. In such cases, your expectation that someone will not betray you, that is, your expectation that the person will act in a trustworthy way, is not based on that person’s inferred character, but on information about the person’s feelings or love toward you. Here, you trust a specific person not because you know
40 the person has a trustworthy character, but because you know the person has warm feelings and positive attitudes toward you. Let us call trust based on that kind of information—information about someone’s feelings and attitudes toward you—“relational trust.”
Summary of Relations among Concepts Related to Trust Figure 2.1 presents concepts related to trust in a summary form. Only terms in bold are concepts included within the range of “trust” in this book. Let us review those concepts by referring to Figure 2.1.
Expectation s of natural order
Expectatio ns of competence
Expectation s of moral order
Assuran Expectatio ns of intentions
Relational trust Trust Character-based trust
Information-bas ed trust
Personal trust Category-b ased trust
Relations among concepts related to trust
41 First, this book excluded expectations of natural order, such as “the sun will rise again tomorrow,” from the analysis of trust. No one will object to this decision except those who see the essence of trust in “cognitive parsimony” or simplification of complex reality as Luhmann did. I do not think that the essence of trust is in cognitive parsimony. I take this position because empirical studies presented later indicate that trust involves roles and functions that far exceed that of summarizing complex information as entailed by cognitive parsimony. The next distinction we have seen is between expectations of competence and expectations of intentions. In both English and Japanese, daily usage of the word, trust, involves both meanings. It is common to both meanings that one feels secure dealing with a trusted person, but the reasons or conditions for that feeling are totally different. In this book, expectations of competence are excluded from analysis of trust. Then, we distinguished between trust and assurance or assurance of security. This distinction is not as self-evident as the previous ones. I believe that some of the confusion in past discussion of trust came from difficulty in making this distinction. According to the distinction between trust and assurance I propose, trust is expectations of other people’s (particular individuals or general others) intentions based on one’s judgment of their character traits or their feelings toward one. On the other hand, assurance comes from the judgment that there is no incentive for a partner to take advantage of you.* In other words, assurance is defined in this book as the judgment that the relationship between you and an interaction partner involves no social uncertainty. The significance of the distinction between trust and assurance will be made clear latter. Now we have finished defining trust, distinguishing it from other concepts such as expectations of natural order, expectations of competence, and assurance. The next step is to analyze sub-categories of trust. First, I divide trust into character-based trust and relational-trust. Character-based trust is based on judgment of trustworthiness as a general character trait. That is, character-based trust is the expectation that acting in a trustworthy manner belongs to a person’s general disposition. In contrast, relational trust is based on judgment of a person’s attitudes and feeling toward ego. That is, relational trust is the expectation that a person is *
Bacharach & Gambetta (1997) argue that information about an interaction partner’s character is necessary in order to know the effect of incentives for the partner. For instance, they quote an anecdote that Mafia bosses who emigrated to the U. S. could not trust blacks (could not feel assured of security, as it is called in this book, with blacks) as their henchmen. The reason is that no threat is effective for the Blacks, a Mafia boss explained. The boss trusts his men (or, he is assured of his men’s compliance) only because his threat works. Thus, he cannot trust Blacks for whom threat of execution for betrayal has no disciplinary power. In this sense, information concerning whether or not his man is a person who fears the threat of execution plays a critical role in determining if the man is trustworthy. If we accept Bacharach and Gambetta’s argument, what distinguishes assurance from trust would be the type of character trait—trustworthiness vs. sensitivity to threat—to base one’s prediction on, while both are sub-types of information-based trust.
42 disposed to act in a trustworthy manner toward “me,” no matter what the person does to others. According to the type of information, character-based trust may be divided further into personal trust, based on information about a specific individual, category-based trust, based on information (or, prejudice or stereotype) about people belonging to a specific category, and general trust, based on information, knowledge or beliefs about other people or human beings in general.
Trust and Trustworthiness The relations between trust-related concepts shown in Figure 2.1 include another distinction, the distinction between trust in a broad sense and trustworthiness. Past treatments of trust have given this distinction little attention, but I think it is important. According to the definition of trust proposed in this chapter, trust is the assessment of other people’s trustworthiness. Trustworthiness concerns whether or not someone actually will act in a trustworthy manner, or whether or not the person can be relied upon. Trustworthiness is a character trait of a trustee, that is, his or her disposition to act in an altruistic or ethical manner even when the action is not backed up by self-interest. While trustworthiness is a character trait of a trustee, trust is a character trait of a truster. It is usually the case that trustworthy people—people with noble character or those strongly disposed to altruistic actions—are trusted by people around them. If a trustee’s character trait (i.e., her trustworthiness) is reflected correctly in surrounding people’s judgments of it, trust is reduced simply to reflection of her trustworthiness. If this were generally the case, trust research would become the study of how people acquire trustworthy character traits, and we would not need to consider characteristics of those who trust trustworthy people. I would rather adopt the position that trust is not simply a reflection of someone’s trustworthiness. The reason is that I see the essence of trust in expectations of someone’s intentions that are not based on objective evaluations of that person’s trustworthiness. In other words, it is my position that differences between individuals in the degree to which they trust the same person should not be treated simply as reflections of errors in their judgments. I argue that individual differences in trust in the same person (and in the same situation) is not simply a reflection of random errors, nor of inability to discern that person’s trustworthiness precisely, but a reflection of individual differences in the way people evaluate others’ trustworthiness based on incomplete information. Details of this argument will be discussed in the next chapter. What I emphasize here is just that trust and trustworthiness are separate concepts, such that the former is a characteristic of the truster while the latter is a characteristic of the trustee. Discussions of trust, especially by social scientists, including economists and sociologists, seldom have distinguished between trust and trustworthiness. In many cases, trust is treated simply as a reflection of trustworthiness. For instance, what is
43 important in the market for lemons discussed earlier is not whether the buyer trusts the seller or not, but whether or not the seller will act in a trustworthy manner. As can be understood from this example, to social scientists who treat trust as a lubricant of social relations or as social capital, trust means trustworthiness as a characteristic of the trustee rather than a characteristic of the truster. Failure to note this difference would cause senseless confusion between psychologists and social scientists.
The Paradoxes of Trust Revisited Finally, let me demonstrate here that the three paradoxes introduced in Chapter 1 in fact stem from confusion among concepts related to trust, especially confusion between assurance and trust. Once shown the demonstration, the reader might feel like a child who has been shown the secret to a magic trick. While the secret behind the seeming paradoxes is simple—distinguishing between assurance and trust—failure to make this simple distinction caused much of the confusion in past discussions of trust. Let us start with the first paradox: trust is not needed in stable relations where trust is easiest to produce, and trust is needed most in situations of high social uncertainty, situations in which trust is most difficult to produce. As we have seen by now, what is produced in stable relations with low social uncertainty is in fact assurance of security but not trust. Details of this point will be discussed in Chapter 3, and thus here it is necessary only to understand that stable relations are relations of low social uncertainty, in which assurance of security is provided. Once we understand that assurance but not trust is provided in relations of low social uncertainty, the first paradox may be paraphrased in the following way. Assurance is provided most easily in stable relations in which trust is not needed, while assurance is hard to provide in situations of high social uncertainty in which trust is needed. As previously discussed, assurance refers to a situation in which social uncertainty does not exist. Thus, the first paradox could be paraphrased further in the following way. In stable relations of low social uncertainty, assurance can be provided easily while trust is difficult to generate. On the other hand, trust is needed under high social uncertainty because assurance is not provided there. The above statements are in fact simple restatements of the definitions of assurance and trust. There is nothing surprising there. The first paradox looks like a surprising paradox only for those who fail to distinguish assurance from trust. Let us move on to the second paradox: the level of general trust is lower in Japan than in the U.S., despite the fact that in Japan networks of stable relations extend to every corner of society and collectivist, stable social relations play dominant roles. As with the first paradox, the second paradox turns out to be a non-paradox if we understand that it is assurance of security, not trust, that is
44 provided by stable relations with low social uncertainty. Whereas a common belief is that people trust each other at a higher level in Japanese society than in American society, they are actually being assured in Japan that their partners are constrained not to betray them. Survey results showing a U.S.-Japan difference in the level of trust—the level of trust is lower for the Japanese than for Americans—are not counter-intuitive once we understand that the common sense belief actually refers to high constraints on behavior in Japanese society whereas the survey results are about trust in human nature rather than confidence in the constraining power of social relations. This, however, does not explain why Americans are more trusting than Japanese. Explanation of this will be given in Chapter 3. Finally, let us consider the third paradox: high trusters who generally tend to trust others are not naïve and gullible; rather, they are sensitive to whether or not other people are trustworthy, and are accurate in judging whether or not other people will act in a trustworthy manner. The key to solving this paradox is not the distinction between trust and assurance but the distinction between general trust and information-based trust. On the one hand, high trusters are those who have high default expectations of other people’s trustworthiness. On the other hand, whether one is gullible or not depends on whether one appropriately processes information relevant to the trustworthiness of the person concerned. Thus, the third paradox indicates that those who have a high default value in assessing other people’s trustworthiness without information are accurate processors of information about other people’s trustworthiness. The distinction between general trust and gullibility is not yet sufficient to solve the third paradox. A question still remains. Why do those who have high default expectations about others’ trustworthiness have the ability to process properly information related to trustworthiness of a specific other? The key to answering the question again lies in the distinction between assurance and trust. Nevertheless, this key is not simply to rephrase a particular type of trust as assurance, as in the case of both the first and the second paradoxes. I sketch here only a rough outline of the full, complex argument I present later. The key to understanding the third paradox lies in the point that assurance of security does not require prudence and caution, while prudence is needed for trust. Whether or not assurance is provided in a certain relation can be judged by looking at the nature of the relation. Those who live in an environment where security is assured by the nature of incentives (that is, an environment with low social uncertainty) are not pressed by the necessity of judging whether or not surrounding people are trustworthy, and thus they are not likely to be trained fully to make such judgments properly. Suppose people used to a low uncertainty environment are placed in a situation of high social uncertainty and forced to judge if people they are meet are trustworthy. Since they have not developed skills to deal with potentially untrustworthy others, they will not be able to discern effectively trustworthy others
45 from untrustworthy ones. Under such a situation, their best strategy would be to assume that all strangers are untrustworthy. In contrast, those used to a high uncertainty environment who have had sufficient experience at judging other people’s trustworthiness will have developed skills for processing information related to other people’s trustworthiness. The need to assume that “everyone is a thief” would be much smaller for them than for those who have not developed proper skills for detecting signs of untrustworthiness in people. This means that those who can process information about a specific person’s trustworthiness properly will be less inclined to assume a person is a “thief” without ground. This is the outline of how to solve the third paradox. More detailed explanations will be presented in Chapter 6.
Chapter 3 The Emancipation Theory of Trust
I am going to present in this chapter the outlines of major theoretical approaches for analyzing trust, and then present my original approach I have developed in this book—the Emancipation Theory of Trust, which differs greatly from the former approaches. My approach differs greatly from the previous ones in that it emphasizes the relation-extension role of trust in addition to the relation-consolidation role (the role the previous approaches have traditionally focused on) of trust. The relation-extension role has been completely overlooked in the previous studies of trust. I will argue that full understanding of trust would require both of the two roles—relation-extension as well relation-consolidation roles.
Trust as Encapsulated Self-interest Most trust researchers agreed that trust plays the role of lubricant for social relations as discussed earlier. Looking at this preposition from the vintage point of the conceptual distinction introduced in Chapter 2, we are now given two possible interpretations of this proposition. One interpretation is that trusting others plays the role of lubricant for social relations, and the other is that acting in a trustworthy manner plays the role of lubricant. With regard to this, most social scientists take the position that trust is the reflection of trustworthiness. From this point of view, the role of lubricant for social relations is basically played by people’s acting in a trustworthy manner. In most discussions by social scientists of trust as a reflection of trustworthiness, what they are talking about is actually assurance. That is, it is often assumed in their discussions that people expect others to act in a trustworthy manner because of the existence of social relations in which acting in a trustworthy manner benefits themselves. Of course, the issue of trust as a reflection of trustworthiness does not logically correspond to the issue of assurance. The expectation that someone will act in a trustworthy manner might come from the noble qualities inferred in that person. However, in actual discussions of trustworthiness as a lubricant of social relations, discussions proceed using the same logic as in the discussions of assurance. Namely, expectations that someone will act in a trustworthy manner are assumed to be backed up by the knowledge about that person’s self-interest. I called the idea that trust is ultimately grounded in the trustee’s self-interest the groundedness approach (Yamagishi, in press). This approach is mainly adopted
47 by economists and other social scientists (sociologists, political scientists, and so on) who stand in the tradition of rational choice theory. This approach could be said as the central approach of trust research in social science. Hardin’s (1991, 1992) idea that trust is encapsulated interest is representative of this approach. The essence of this approach lies in the case that one trusts a person A to do X because A’s self-interest is in accordance with doing X. According to my definition, however, the above case is an example of assurance, not of trust. Hardin’s trust as encapsulated interest corresponds to the conviction that any person who have been implanted with “a thousand needle machine” will never tell a lie. Those who wish to be trusted (or to be considered assuring), that is, those for whom being trusted (or being considered assuring) brings better outcomes than not being trusted (or being considered assuring), can produce a state of being trusted by having “a thousand needle machine” implanted in their throat, or acting in a similar manner. This is the essence of the groundedness approach. Of course, “a thousand needle machine” does not actually exist, but the same effect can be achieved by putting oneself in a position in which betrayal of trust produces negative outcomes to oneself. For that purpose, a “hostage” can be posted to provide an assurance of “honest” conduct. Hostage in this case refers to, for instance, warranty for a used car (repair to a car with warranty is a loss to the dealer, not to the consumer), or trustee’s reputation (maintaining good reputation is more important than immediate gain of cheating for the people who need to rely on reputation). Various hostages are actually used to produce trust (or assurance). Exchange of human hostages among war loads in the Japanese civil war era represents the mechanism of producing trust in this sense (i.e., assurance in my definition). According to the groundedness approach, trust (or assurance) exists in this world (that is, people can feel secure in social relations) because there are situations in our world in which being trusted (or being felt safe) yields better outcomes than not being trusted (or not being felt safe). In such situations, “those who pursue self-interest” produce sufficient evidence of trustworthiness (or provide assurance of security) by voluntarily providing hostages.
Trust Produced in Stable Relations Many readers may feel at odd with the account of trust from the groundedness approach. They might say: “We trust because we are not certain of the consequence.” Their view is also shared by some students of trust, especially psychologists, who consider trust as something that is beyond simple calculations of interest. Lewis & Weigert (1985) expressed this idea succinctly when they stated that “trust begins where simple prediction ends” (p.976). If we call this approach a psychological approach, we can easily see that the same word “trust” is used in quite differently in the above-mentioned groundedness approach and in the psychological approach. What the groundedness approach deals with is basically trustworthiness of the trustee—whether or not someone will act in a trustworthy manner. What the
48 psychological approach deals with is willingness “to trust.” Researchers who have adopted the groundedness approach (mainly economists and other social scientists, such as sociologists and political scientists, who take the position of rational choice theory) pursue the question of “why people act in a trustworthy manner.” Those who have adopted the psychological approach poses another question: “Why do people trust others even when available information is insufficient to predict others’ behavior?” The two camps have been pursuing different questions. It is no wonder why there has hardly been any communication between the two camps. The groundedness approach and the psychological approach differ not only in questions they pursue but also in their explanatory principles. The ultimate principle to explain trustworthiness in the groundedness approach is the incentives that induce people to act in a trustworthy manner. It is the logic that trust (assurance of security, according to my definition) exists in this world because people act in a trustworthy manner so as to enhance their self-interest. In contrast, the psychological approach lacks the ultimate principle to base explanations on. For instance, Rotter, who is a representative of psychological approach in trust research, defines trust as “an expectancy held by an individual or a group that the word, promise, verbal or written statement of another individual or group can be relied upon (Rotter, 1967),” but does not provide explanation of why such trust exists at all. Erikson’s (1963) research in clinical psychology specifies in what environment (especially what kinds of relationships with parents) children develop trust in others (specialized others or generalized others). However, no explanation is given concerning why such environment produces trust. In other words, the “explanatory principle” held by psychologists is no more than the principle of generalized reflection of trustworthiness. That is, a child raised in a stable and safe environment where no one around exploits others comes to learn to trust others (at least those in the immediate environment), and this trust is generalized so that the child comes to trust others in other environments. From this psychological “explanatory principle,” what causes people to trust others is their experience in the stable and safe environment where people around act in trustworthy manners. In this way, willingness “to trust” is ultimately reduced to (the surrounding people’s) trustworthiness even in the psychological approach. That is, people are assumed to come to acquire high general trust (i.e., strong tendency to trust others) in an environment where surrounding people act in a trustworthy manner. Hardin, the adovocate of the encapsulated self-interest view of trust, called the idea that trust is cultivated in a stable environment as thick relation theory of trust. He points out that trust is eventually based on trustworthiness of surrounding people even in the thick relation theory of trust. Thus, the thick relation theory or the psychological approach is reduced to the groundedness approach, he claims.
Limitations of Reductionist Approach As shown above, no account of trust that is not ultimately reduced to specific others’ or to surrounding people’s trustworthiness—trust that exists independently of specific others’ or surrounding people’ trustworthiness—has been proposed. That is, in the past trust research, willingness “to trust” has been either directly (the current exchange partner is trustworthy) or indirectly (one has been interacting mostly with trustworthy people in the past) attributed to other people’s trustworthiness. This approach to trust has been accepted by many researchers. It involves no problems on the surface; it does not seem to be problematic that understanding of trust means specifying the conditions in which people act in a trustworthy manner. Now, let us examine if the “reductionist” approach to trust provides any help to understanding the three paradoxes of trust presented in Chapter 1. Let us examine this starting with the first paradox—the paradox that trust is cultivated in a stable and intimate relations in which people behave to each other in trustworthy manners, while trust is needed in relations with high social uncertainty. The issue here is how trust nurtured in safe relations can be brought into relations in which trust is difficult to produce. The “reductionist” approach states that both trustworthiness and willingness to trust are produced in safe relations, but it cannot not explain how trustworthiness and willingness to trust can be extended beyond such relations. We have seen in Chapter 2 that the paradox only looks like a paradox due to the failure to distinguish assurance and trust. With this distinction, trust is needed in relations with high social uncertainty, whereas assurance is produced in stable relations with low social uncertainty. The “reductionist” approach can explain assurance but not trust. Thus, this approach is of no help for understanding the first paradox, of which solution requires the distinction between assurance and trust. How about the second paradox? The paradox is that Americans are generally more trusting than Japanese. This paradox, too, is the result of the confusion between trust and assurance as shown in Chapter 2. Given that what characterizes Japanese society and Japanese business practices are in fact assurance but not trust, the “reductionist” approach can easily explain the high level of assurance in Japanese society but not why the level of general trust is low in Japan. As for the third paradox, the “reductionist” approach is completely helpless. The “reductionist approach” can never explain the reason why high trusters are sensitive to information suggesting other people’ trustworthiness. Why high trusters who generally trust others are sensitive to information about other peoples’ trustworthiness, or why those who are sensitive to information about other peoples’ trustworthiness generally trust other people, can not be explained either by the groundedness approach that asserts that one trusts someone because that someone is trustworthy, or by the psychological approach that claims one trusts others because one is raised among trustworthy people. The emancipation theory of trust introduced in this chapter explains the three
50 paradoxes based on the distinction between trust and assurance introduced in Chapter 2. The theory consists of the six propositions to be introduced below.
Social Uncertainty and Formation of Commitment Relations In Chapter 1, I used the example of market for lemons to illustrate that trust plays the role of lubricant in social relations. This refers to a proposition that trust is most needed in situations of high social uncertainty. The emancipation theory of trust starts with the first proposition that addresses the importance of social uncertainty for trust. Proposition 1: Trust is meaningful only in situations where social uncertainty exists. Trust is not needed in situations where there is no possibility of being deceived or exploited by others. Now, imagine you were in a situation where social uncertainty is high and to trust or not to trust someone is very important. You gain large profit if you trust the person and the person is in fact trustworthy. Please note that the words “profit,” “benefit,” or “interest” is used in a broad sense involving not only money or materials; they also refer to psychological gains such as satisfaction, love, friendship, and so on. For example, in cases when a girl is reciprocated with love for her trust in her boyfriend, we will talk about the love using the term “profit”; the girl has earned profits in return for her trust in her boyfriend. The same act of trust may also incur losses instead of profits, when the person you trust turns out to be untrustworthy. The word “losses” here include not only monetary or material losses but also psychological losses such as miserable feelings. Let us examine an example here. Suppose you have discovered your partner has been unfaithful to you. After many quarreling, your partner deeply regretted his past behavior and begged you: “please give me one more chance to make up for it.” Whether to trust or not to trust your partner’s word is critical. If your partner is really trustworthy, there is a greater chance to have a happy relation by trusting your partner and giving him one more chance than trying to find a new partner from scratch. However, if your partner have not been sincerely penitent of his past deeds or was penitent at the time but soon has lost his penitence, the chances would be that you have a happy relation with a new partner than trusting his words and giving him another chance. How would people do when they face such a situation of high social uncertainty? There might be several ways to deal with such situations. The first is to trust the partner. In the above example, you do not know whether or not the partner is really penitent, but you decide to trust the partner and try your effort to build a warm relation one more time. By trusting, social uncertainty can be reduced at least subjectively. Trust can make it is possible to build a social relation that is impossible
51 without trust. But, at the same time, the possibility of failure exists. The partner in the above example might repeatedly betray your trust. Those who trust others can be exploited by untrustworthy others. In this sense, it should be concluded that those who trust someone without paying enough attention to discern his or her trustworthiness are naïve and stupid. Is it really so? Are those who do not easily trust others under any circumstances cleverer—that is, pursue self-interest more efficiently—than those who easily trust others. Or can we say they are the people who pursue self-interest more efficiently? The emancipation theory of trust that I will present below starts with a negative answer to this question. The idea that trusting others may have its own merit constitutes the intuitive foundation of my argument. In other word, the starting point of the emancipation theory of trust is the hunch that there may be situations in which being “naïve” and “credulous” may bring in more beneficial outcomes than being overly prudent. The other way to deal with social uncertainty is not to reduce social uncertainty subjectively but to reduce it objectively. In short, it is to build a device in a specific relationship according to which selfish behavior brings disadvantage to the people themselves. Given this device, no one will try to exploit others, and there will be no risk of being exploited in that specific relationship. One of such devices is what economists call “hostage exchange” or “hostage posting.” Hostage here refers to the losses that would be incurred to oneself if one conducts undesirable behavior. It was argued in Chapter 2, using the example of warranty to used cars, that the hostage posting in this sense is to provide assurance to the interaction partner, or to reduce social uncertainty for the partner. The seller of lemons will have to pay for the repair if he provides warranty for the car he sells. Then, the seller would not want to sell lemons. Once buyers understand this, and feel safe in buying used cars, sellers will eventually make better profits. The reduction of social uncertainty through hostage posting may be difficult to achieve between an unfaithful lover and his partner. This is because posting a hostage is effective only when the value of the hostage is greater than the benefit of acting in an unfaithful manner. For instance, a man may promise that he will buy a diamond to the partner if he ever acts unfaithfully, but there is no guarantee that he will keep the promise. Furthermore, for some people the “benefits” coming from unfaithful action are greater than the cost of buying a diamond worth tens of thousands of dollars. Thus, reduction of social uncertainty by hostage posting is frequently used in impersonal relations such as business relations, but not so often in personal relations. An alternative method often used when hostage posting is hard to perform is the formation of a stable and intimate commitment relation with specific partners. Dealing only with specific partners of a long-standing relation is a means to reduce social uncertainty. This is a method of uncertainty reduction that can be used when it
52 is difficult to post a hostage. This method is perhaps the most commonly used means in human history to deal with social uncertainty.
Commitment Relations Before proceeding further, let me first define “commitment relations.” The word “commitment” is used with various meanings though not in as many as the case of trust. The most common usage of this word is to refer to a state in which people are psychologically committed or “attracted” to each other. Social psychologists usually talk about commitment relations in this sense; that is, as relations in which people like or love each other and want to be together. In contrast to this psychological use, social exchange theorists represented by Emerson (1976) define commitment not in terms of emotional bonding but simply as the continuation of a relation. According to this approach, a commitment relation is defined to be one in which the same partners interact with each other even though they could have made better profits by defecting the relation. Once such a commitment relation is formed, it would be often the case that emotional bonding such as mutual liking and loyalty develops. But, according to this approach, such emotional bonding itself is not defined to constitute a commitment. I have adopted the second definition of “commitment relations” for this book. That is, commitment relation is defined to exist when two partners refuse temptation of more profitable offers from outside and stay in the relation with the same partner no matter what the reason for this might be. The reason for adopting this definition is that emotional bonding is not the only reason for people to stay in the same relation despite temptations of better offers from outside. In the next section, I divide commitment relations into two types, “lovers-type” commitment relations which are formed by emotional bonding and “yakuza-type” commitment relations which are formed as a means of protection against hostile (or hostile looking) outside society. What plays the important role in the following analysis of trust is “yakuza-type” commitment relations, not “loves-type.”
Formation of Yakuza-type Commitment Relations I have stated above that formation of commitment relations is an effective means of reducing social uncertainty. Let us ponder on this for a little more. First, it should be noted that that this argument does not apply in the case of “lovers-type” commitment relations. This is because the partners of a lovers-type commitment relation treasure each other, and thus there is almost no social uncertainty in such relations. Conversely, in relations with high social uncertainty, it would not be easy to engender mutual liking and mutual attraction in order to reduce social uncertainty. For example, a comsumer’s effort to come to like a used car salesman would be in vain as a means to eliminate social uncertainty from the purchasing of a used car. It
53 is clear that the formation of lovers-type commitment relation is of not much help for the purpose of reducing social uncertainty from a social situation. On the other hand, yakuza-type commitment relation is the relation that is formed exactly for this purpose. It is the relation that is formed among people who face a highly uncertain social situation in order to reduce uncertainty inside the relation. People who face a risk of exploitation by others would want to reduce the risk by dealing only with specific partners. Then, they are forming a yakuza-type commitment relation with the specific partners. The two who are in deep love with each other and does not pay any attention to anyone else constitute a lovers-type commitment relation. The two who are afraid of getting STD in new relations and for whom safety in sex is the only reason for not seeking alternative relations constitute a yakuza-type commitment relation. There are a few reasons why yakuza-type commitment relations reduce social uncertainty. First of all, as demonstrated in the study of iterated Prisoner’s Dilemma by Axelrod (1984) on which I will discuss in the next section, the incentive for unilateral betrayal is reduced when a relation is expected to continue. Axelrod argues that mutual cooperation emerges in a Prisoner’s Dilemma situation—an example of the situation characterized by social uncertainty—when the situation is expected to last over a long period. He explains the formation of mutual cooperation in such a Prisoner’s Dilemma situation in terms of the effectiveness of the tit-for-tat strategy. The study of Prisoner’s Dilemma by Axelrod is a very well known study, and one that is unavoidably mentioned in the treatment of problems of cooperation, competition or trust in social relations. Here, I would like to introduce this research and related studies. Pruitt and Kimmel’s Goal/Expectation Theory While I briefly introduced Prisoner’s Dilemma in Chapter 1, I would like to provide a more extended discussion of it here. The term “Prisoner’s Dilemma” originally came from an illustration of two suspects (or prisoners) who have been arrested for a serious crime. This example was used to explain the incentive structure existing in the Prisoner’s Dilemma situation. Two persons A and B were arrested on a trivial crime they are known to have committed, while they are suspected of having committed a more serious crime. Their confession is needed to prove the serious crime at the court because the district attorney does not have sufficient evidence to prove them guilty for the crime. Without confession, the district attorney has to give up prosecuting them for the serious crime, and instead prosecute them for the trivial crime. Thus, if the two keep silent on the serious crime they will be sentenced only for the trivial crime. Facing this situation, the district attorney in charge of this case talks to the two suspects separately and makes the following proposal. “We already know what you have done. It is wise for you to confess now. If
54 you confess while your accomplice continues to pretend to be innocent, you will be freed without being prosecuted. If you try to pretend to be innocent while your accomplice confesses, then, you will only receive life imprisonment since there is no room for extenuation. If both of you confess, each of you will be sentenced to ten years in prison on extenuation. If you two do not confess, we will have no other way but to prosecute you on another charge. In this case, both of you would be sentenced for one year in prison.” In this case, a relationship exists between the two suspects. 1. Each suspect has to make a choice between confessing or not confessing. For the two suspects, not confessing is “cooperation” while confessing is “defection.” 2. For each suspect, confession (the choice of “defection”) brings in a better outcome (that is, receiving lesser punishment) than no confession (choosing “cooperation”) no matter if the accomplice confesses or not. For instance, when the accomplice chooses not to confess (choosing “cooperation”), the suspect who confesses is acquitted whereas the suspect who does not confess receives a year of imprisonment; the consequence of confession (being free) is more desirable for the suspect than the consequence of not confessing (one year in prison). Similarly, when the accomplice chooses to confess (choosing “defection”), the suspect who confesses gets ten years of imprisonment whereas the suspect who does not confess gets his whole life in imprison; again, confession bring in a better outcome (ten years in prison) than no confession (the rest of life in prison). 3. However, both of the two suspects get ten years of imprisonment when both of them confess. If neither of the two confesses, both of them get only one year of imprisonment. The outcome of mutual cooperation (not confessing) is thus more desirable than the outcome of mutual defection (confessing). As demonstrated in this example, Prisoner’s Dilemma refers to a relationship in which “defection” produces more desirable outcome than “cooperation” for each regardless of the choice of the other, whereas the two suffer from undesirable outcome if both seek their own interest and choose “defection.” The nature of the relationship between the two who face a Prisoner’s Dilemma is usually represented by a “payoff table” or “payoff matrix” as shown in Fig. 3.1. This payoff table shows how much benefit (or loss) each receives as a consequence of the two persons’ (A and B’s) choices of “cooperation” or “defection.” For example, the upper right cell represents the benefits that A and B obtain, respectively, when A chooses defection (confession) while B chooses cooperation (no confession). A’s benefit is shown above the diagonal line and B’s benefit below the diagonal line in each cell. In the actual Prisoner’s Dilemma research, the benefits in the payoff table are often represented by points or money. In such research, participants are shown the payoff table and are asked which alternative they want to choose, “cooperation” or “defection.” More than a thousand experimental studies have been conducted on
55 Prisoner’s Dilemma. One of the major findings from those experimental studies is that the principle that guides choice behavior of experimental participants varies, while using the same payoff matrix, depending on whether the choice is made once between two players or it is made repeatedly between the same players.* In one-shot Prisoner’s Dilemma experiments, participants usually do not choose “cooperation” at a very high level. In iterated (i.e., repeated between the same partners) Prisoner’s Dilemma experiments the choice of “cooperation” often decreases at early stages, but later it begins to increase and consequently cooperation rate tends to be higher than that in the one-shot Prisoner’s Dilemma. Why does such a difference emerge?
A’s Choice Cooperation (not to confess)
Defection (To confess)
1 year Cooperation (Not to confess)
Defection (To confess)
10 years 10 years
Payoff matrix of the Prisoner’s Dilemma between two suspects
First, let us consider the case of the one-shot Prisoner’s Dilemma. An egoist who cares only his self-interest will of course choose defection, because benefit from the choice of defection is greater than that from the choice of cooperation *
Readers who are interested in experimental studies of Prisoner’s Dilemma are advised to read
56 regardless of the partner’s choice. Participants who chooses “cooperation” in such experiments are considered to do so because of their motives beyond self-interest, such as altruistic motivation to care for the partner’s interest or motivation to care for the social norms or conscience that dictates the right behavior in such a situation. In the iterated Prisoner’s Dilemma, that is, in the Prisoner’s Dilemma that is played repeatedly between the same two people using the same payoff matrix, there emerges a chance that the principle of “altruistic egoism” as I called it (Yamagishi, 1990a, 1990b) come to play a role. When one of the two players keeps choosing “defection,” the situation of this one-sided “exploitation” will not last for long even though the partner may initially choose “cooperation.” Even those who understand the need for mutual cooperation and choose “cooperation” will eventually come to choose “defection” due to anger or to self-protection if the partner keeps choosing “defection,” resulting sooner or later in mutual defection. Then, those who attempt to exploit the partner’s initial willingness to cooperate will eventually come to understand that it is not easy to keep taking the lion’s share and that their own defection only brings a state of mutual defection. People who play the iterated Prisoner’s Dilemma thus come to understand from their experience that there is no way to get out of the quick sand of mutual defection unless they give up the honey of one-sided defection and aim for mutual cooperation. Pruitt and Kimmel (1997) extracted the following conditions for the production of mutual cooperation in iterated Prisoner’s Dilemma through examination of over one thousand studies conducted during the 20 years prior to the mid 1970s. (1) First, people have to understand that mutual cooperation is required to get out of the state of mutual defection. That is, they have to come to understand that reckless pursuit of self-interest works against their own self-interest and that their self-interest is served only when they secure long-term interest through achievement of mutual cooperation. However, not all who have come to understand this necessarily choose “cooperation.” This is because even those who hope to achieve a state of mutual cooperation are afraid that the partner may take advantage of their willingness to cooperate; they would not cooperate unless they are convinced that the partner does not capitalizes on their cooperation. In order for those who desire mutual cooperation to actually take the cooperative choice, then, their distrust in the partner needs to be somehow eliminated, so that they can expect that the partner like themselves are hoping to achieve mutual cooperation and would cooperate if they themselves do. In other word, (2) those who have adopted attainment of mutual cooperation as their goal do actually choose to cooperate only if they trust that the partner would not take advantage of their willingness to cooperate. Pruitt and Kimmel called the above two propositions “goal/expectation theory” of cooperation. In short, the theory specifies the two conditions for cooperation to take place in Prisoner’s Dilemma situations. First, people have to Yamagishi (1990a, 1990b) or Yamagishi (1995).
57 understand that one-sided exploitation cannot be sustained and thus they have to achieve mutual cooperation to gain desirable outcomes. At the same time, they need to expect that the partner also wants to achieve mutual cooperation and thus would not take advantage of their cooperative behavior.
The Tit-for-Tat Strategy The above mentioned Pruitt and Kimmel’s theory states that people become willing to cooperate in order to secure long-term interest through their experience of repeatedly playing the Prisoner’s Dilemma with the same partner. They further state that even those who are willing to cooperate to secure long-term gain would be hesitant to cooperate unless they trust the partner. Indeed, it has been shown in many experimental studies on Prisoner’s Dilemma that cooperation rate differs greatly depending on whether or not participants trust the partner. Results of those studies are mostly consistent with Pruitt and Kimmel’s claim that mutual cooperation is difficult to achieve unless participants trust the partner. It is also well known, though, that mutual cooperation can be achieved among egoists even when they do not trust the partner. The most well known case in which mutual cooperation is achieved among egoists is when at least one party adopts the “tit-for-tat” strategy. The tit-for-tat strategy refers to the principle often adopted by the player of iterated Prisoner’s Dilemma, in which the player chooses to defect following the partner’s choice of defection in the previous trial, and chooses to cooperate following the partner’s choice of cooperation in the previous trial. It is the principle for the choice (or the strategy) in which the player repeats what the partner has chosen in the previous trial. Let us use the example of the iterated Prisoner’s Dilemma represented by the payoff matrix shown in Figure 3.2 to see what will happen when one player adopts the tit-for-tat strategy. Participants in Prisoner’s Dilemma experiments are usually shown such a payoff table and then are asked to make a choice between “cooperation” and “defection.”* With this payoff table, each of the two players receive 200 yen for the trial in which the two choose to cooperate. When one defects and the other cooperates, the one who defects gets 300 yen and the other gets nothing. Finally, each gets 100 yen when both choose to defect. Two players, A and B, play the Prisoner’s Dilemma over many trials. Suppose A is a self-centered person who does not feel the need for mutual cooperation while B desires to achieve mutual cooperation and has adopts the tit-for-tat strategy. Having adopted the tit-for-tat strategy, B would choose to defect only when A has defected in the previous trial; otherwise, B would choose to cooperate. When A chooses to defect, B will also defect, resulting in the defection-defection cell. Then, A gets 100 yen. In contrast to this, A gets 200 yen when A chooses cooperation and *
More neutral words are often used to describe the two alternatives in actual experiments.
58 B responds to this with cooperation, resulting in the cooperation-cooperation cell. When B has adopted the tit-for-tat strategy, A is practically left with the choice between these two alternatives since B responds to A’s choice with the same choice. Given this, A the egoist who cares only for his self-interest will of course choose the alternative that gives him 200 yen rather than 100 yen; that is, he will choose to cooperate. Mutual cooperation will then be achieved. It is important to note that B does not need to trust A for adopting the tit-for-tat strategy. There is no one-sided exploitation by A, because B stops cooperating when A defects. The effectiveness of the tit-for-tat strategy in iterated Prisoner’s Dilemma has been demonstrated in many experiments.
A’s Choice Cooperation
An example of the payoff matrix (in yen) used in Prisoner’s Dilemma experiments
Axelrod’s Computer Simulation To demonstrate that cooperation is possible when the Prisoner’s Dilemma is repeated between the same partners, Axelrod (1984) conducted a computer simulation. He incited this research by writing to famous game theorists in the world inviting them to submit a strategy of their choice for a computer tournament. Game theorists are those who mathematically study what will happen when two or more people follow a certain behavioral principle in a certain interdependent situation. In response to Axelrod’s invitation, fourteen game theorists agreed to enter
59 the tournament. Those fourteen people are well-known game theorists who should know much better than ordinary people the implications of various strategies in the Prisoner’s Dilemma game. The fourteen game theorists who entered the tournament submitted, in the form of computer programs, the strategies that they regarded as the most successful one. These programs included various strategies, ranging from the simplest strategy such as the tit-for-tat, which simply repeat the partner’s choice in the previous trial, to complicated strategies. For example, one of such complicated strategy keeps defecting insofar as the partner does not respond, and switches to cooperation as soon as it finds out that the partner is responding. Axelrod conducted a computer simulation with the fourteen submitted strategies plus the random strategy that randomly chooses cooperation or defection. Each strategy played a Prisoner’s Dilemma game 200 times with each of the 15 strategies including a copy of itself. The average of the 15 simulations was taken as the overall performance score of the strategy. The winner of the tournament was determined by comparing those performance scores of the 15 strategies. The winner of the tournament was the simplest strategy, tit-for-tat. Besides tit-for-tat, “nice” strategies that do not defect unprovoked generally performed well. In contrast, “nasty” strategies that aimed at taking advantage of the partner failed to perform well. The same tournament was held again, and sixty-two people from six countries entered the second tournament. Invitation was published in several magazines for personal computer users together with the result of the first tournament. The winner of this second tournament was again the simplest, the tit-for-tat strategy. The result of the Axelrod’s computer tournaments demonstrated the effectiveness of the tit-for-tat strategy as a means to promote the player’s own long-term self-interest. Then, the question arises whether or not people actually adopt the strategy only because it is effective to promote their own long-term self-interest. In order to answer this question, Axelrod performed another computer simulation, this time on evolution of strategies. Suppose there is a group which consists of people who have various strategies. Some may use the tit-for-tat strategy, some may adopt a strategy of unconditional defection, and some may adopt unconditional cooperation strategy. All these people are connected to others with a Prisoner’s Dilemma network. They play an iterated Prisoner’s Dilemma with each of the other people in the group. As people play the iterated Prisoner’s Dilemma with various others in the group, differences in the profits they make will emerge. Those who adopt a certain strategy will earn more or less than that of those who adopt other strategies. In Axelrod’s new simulation, strategies that accumulate low scores through repeated interactions with other strategies are eliminated, replaced by copies of strategies that earn high scores. As can be expected from the result of the first simulation, the tit-for-fat strategy earned relatively good scores, and gradually replaced less well performing strategies. While various strategies existed in early stages, as time went on all strategies were
60 converted to the tit-for-tat strategy. Of course, even the tit-for-tat strategy failed to spread in the group in which all members had adopted the unconditional defection strategy. But even in such groups initially consisting of unconditional defectors, everyone eventually became to adopt the tit-for-tat strategy when several tit-for-tat players “invaded” the group simultaneously. The tit-for-tat strategy could “invade” most groups and spread their copies only if they were not alone in the group. In contrast, other strategies were eventually eliminated from the group of tit-for-tat players even when they “invaded” the group in a cluster. This simulation shows that the tit-for-tat strategy not only performs well compared to other strategies, but also can successfully “invade” groups of other strategies and spread out. Furthermore, a group of tit-for-tat players can successfully repel any “invasion” by other strategies. In short, this simulation by Axelrod suggests that altruistic-egoists come to adopt the tit-for-tat strategy as they experience interaction with various others through Prisoner’s Dilemma relations, and as a consequence, a stable state of mutual cooperation emerges among them.
Reduction of Social Uncertainty by Formation of Stable Relations Axelrod’s research is important for us in showing that not betraying the partner is in one’s self-interest in mutual cooperation supported by the tit-for-tat strategy. Using our terminology, social uncertainty is reduced and mutual security is assured when the tit-for-tat strategy is used in iterated Prisoner’s Dilemma. Such assurance of security via tit-for-tat is not possible, however, when the relation between the two is not guaranteed to last. In this sense, formation of a stable and long-lasting relation between two people, namely, formation of a commitment relation, plays the role of reducing social uncertainty and providing mutual assurance inside the relation. Furthermore, formation of commitment relations help reducing social uncertainty through enhanced predictability of the partner’s behavior due to accumulation of information about the partner. That is, a better predictability of each other’s behavior exists in such relations than in relations with a stranger with no guarantee of future interactions. This means that the formation of stable commitment relations with specific partners reduces social uncertainty, and thus, reduces demands for trust as a subjective means to deal with social uncertainty. In other words, there is no need to torment over whether others are trustworthy or not in stable commitment relations. Thus, people who face socially uncertain situations often try to voluntarily form commitment relations in which they can feel safe. These discussions lead up to the second proposition: Proposition 2: People tend to form commitment relations to deal with the problems caused by social uncertainty.
In the situation in which one may suffer exploitation by others, one will try to reduce the possibility of being exploited by associating with only specific partners (i.e. forming commitment relations). Of course, commitment in this context refers to the yakuza-type commitment, not the lovers-type commitment.
Kollock’s Analysis of Rice and Rubber Trades The second proposition would be easy to understand using Kollock’s (1994) example of rice and rubber trades in Southeast Asia. Based on the research on economic activities in Southeast Asia by anthropologists, Kollock points out an interesting difference in the form of trades of rice and of raw rubber. Raw rubber is usually traded in commitment relations that sometimes extends over several generations between particular plantation owners and particular brokers, whereas rice is usually traded in open markets between strangers. Kollock argues that the difference in the form of the two trades is determined by the level of social uncertainty involved in the trades. According to Kollock, the quality of raw rubber cannot be known before it is processed and manufactured into producs. In contrast, the quality of rice is easy to determine. Even a non-expert like me can tell good from poor quality rice with a mouthful of tasting. This difference in the difficulty to determine the quality of traded goods produces different levels of social uncertainty for buyers of rice and of raw rubber. Even buyers with a trained eye can not determine the quality of raw rubber they are buying. They thus face the possibility of getting inferior quality materials. In contrast, there is very little danger for buyers of rice to get poor quality rice because the quality of rice can be immediately determined with simple inspection. This means that the level of social uncertainty involved in the trade for buyers of rice is much smaller than that for the buyers of raw rubber. Kollock argues that raw rubber is traded through commitment relations between the same plantation owner’s family and the broker’s family, sometimes for over generations, in order to reduce social uncertainty involved with the trade. In case of rice trade, where such social uncertainty hardly exists, trade is made with the seller and the buyer who make the best offers in an auction market where commitment relations do not exist.
Experiment of Rice and Rubber Trades After illustrating the difference of rice and raw rubber trades in Southeast Asia, Kollock conducted an experiment in order to verify the proposition that social uncertainty promotes formation of commitment relations. His experiment transported the rice and rubber trades into a laboratory in an abstract form, where participants who were assigned the roles of sellers or buyers traded some goods that sellers produced. The seller first paid for the production cost of goods. The quality of the product was determined by how much the seller had paid for its production.
62 Then, the seller posted the selling price of the product, including production costs and profit margin, together with the quality of the product. The buyer considered the prices and qualities offered by sellers and then decided whose product to buy (or to buy nothing from any sellers). Finally, the buyer resold the product to the experimenter. The resale price was determined by the true quality of the product. In this experiment, the level of social uncertainty was manipulated by the possibility of whether or not the seller lied about the quality of product. In either condition, the seller presented price and quality of his or her product to the buyer. In the low uncertainty condition, the seller could not lie about his or her products’ quality. In the high uncertainty condition, the seller could do so. The high and low uncertainty condition of this experiment was thus designed as a laboratory version of the rice and rubber trades mentioned above. The purpose of this experiment was to see if commitment relations between particular sellers and buyers would emerge more often in the high uncertainty condition in the laboratory market in which the quality of the traded goods was not guaranteed. The result of this experiment supported the proposition that social uncertainty promotes commitment formation. That is, the result replicated the pattern of commitment formation in the actual rice and rubber trades, such that stable relations between particular sellers and buyers emerged more often in the high uncertainty condition where the quality of the traded goods was not guaranteed than in the low uncertainty condition where the quality was guaranteed. The same conclusion was obtained repeatedly in a series of experiments conducted by our trust research group. Here, I would like to present briefly one of these experiments, and leave detailed discussions of the rest for the next chapter.
Network Prisoner Dilemma Experiment The experiment that I present below was conducted by a colleague of mine, the late Hiromi Shinotuka, and my former graduate students, Nobuhito Jin and Nahoko Hayashi (Jin, Sekiya & Shinotuka, 1993). It examined formation of commitment relations between particular partners in the situation in which participants was allowed to freely choose the partner to play a Prisoner’s Dilemma with. Four participants participated in the experiment at the same time. The four were identified by others with assigned colors and could not see each other in person. The participants repeated the following trial 65 times. In each trial, each participant decided whom to select as a partner. When two of the four participants (for example, the red and the purple) have selected each other as PD (Prisoner’s Dilemma) partners, the other two were automatically assigned as PD partners. Participants then played a Prisoner’s Dilemma game with their partner thus decided. Specifically, each participant decided how much of the 10 yen he or she had been
63 given by the experimenter to provide for the partner. The money that one provided for the partner was doubled and given to the partner. Thus, each received 20 yen when both provided 10 yen for each other. If each cared for only his or her self-interest and provided nothing for the partner, each ended up with keeping the original 10 yen. As trials went on, participants came to select a particular person as their PD partner and formed a commitment relation with that person. Once such a commitment relation was formed, the two achieved total cooperation such that the full 10 yen rather than 9 or 8 yen was mutually provided. This commitment relation broke off immediately as soon as one deviated even slightly from the complete cooperation (for example, reduce the provision from 10 to 9 yen). These results indicate that participants try to reduce social uncertainty by forming commitment relations with specific partners, and trust in the partner’s intention definitely played a decisive role in maintaining the commitment relations.
Transaction Cost and Opportunity Cost The experiments mentioned above show that those who face social uncertainty tend to form commitment relations with specific partners. The possibility of exploitation by others is reduced when one forms commitment relations with specific partners and associating only with them. Considering these merits, for people facing high social uncertainty the principle of “ingroup association,” according to which one associates only with specific partners, seems to be the wisest behavioral principle. In the same token, it is a clever behavior to avoid “outsiders” in highly uncertain social situations. However, this conclusion is not necessarily correct because formation of commitment relations with specific partners does not always generate desirable outcomes. The reason is that formation of commitment relations reduces social uncertainty inside the relation and thus produce a secure environment, on the one hand, but it produces an undesirable side effect, i.e., opportunity cost. This issue has not been discussed before. Opportunity cost is a term used by economists, referring to the benefit that could have been obtained if the money and time that have been invested in a certain activity were invested in other activities. I use the term, opportunity cost, in this book to refer to the extra benefit that could have been earned from alternative activities in comparison to the benefit actually earned from the current activity. With this definition, we can say that a commitment relation is a relation in which one is paying an opportunity cost. That is, when one maintains a commitment relation, one forgoes opportunities for getting a better outcome offered by alternative partners. That better outcome forgone is the opportunity cost. Thus, formation and maintenance of a commitment relation reduces social uncertainty inside the relation, but one gives up the chance to deal with other partners who might give one better deals.
64 The following proposition is derived from this discussion. Proposition 3: Commitment relations incur opportunity cost. This proposition is directly derived from the definitions of commitment relations and opportunity costs, and thus no empirical test is needed to prove this proposition. Commitment relations are the relations that are maintained by the parties’ paying opportunity costs. The third proposition is only a paraphrase of this behavioral definition of commitment. An economic term, “opportunity cost,” is used in this proposition. The following issue will be clearer if we use another economic term, “transaction cost.” Transaction cost is the time, effort, money, etc., that is consumed to conduct transactions, including such examples as cost for credit inquiries and consultation with lawyers in preparing contracts. Losses from being cheated in transactions are also included in transaction cost in this book. Given this definition of transaction cost, it takes a large transaction cost, of course, to conduct a transaction under high social uncertainty than under low social uncertainty. When a commitment relation is formed with specific partners and social uncertainty has been reduced inside the relations, then transaction cost should be small inside the commitment relations. That is, the formation of commitment relations, especially yakuza-type commitment relations, can be regarded as a means to save in transaction cost. If we adopt the two terms, transaction cost and opportunity cost, both borrowed from economics, we can state that formation of commitment relations reduces transaction cost, on the one hand, but produces opportunity cost, on the other. Whether or not formation of commitment relations with specific partners is a clever behavior under a certain level of social uncertainty is determined by the relative sizes of the savings in the transaction cost and the incurred opportunity cost. Given a certain level of social uncertainty and thus a certain amount of transaction cost that can be saved from forming commitment relations, whether or not forming commitment relations with specific partners is an advantageous action to take is determined by the amount of opportunity cost incurred by the commitment formation. This discussion leads to the following proposition. Proposition 4: In a social situation in which the level of opportunity cost is high, it is more advantageous to leave commitment relations rather than staying in it. This proposition is in fact a tautology (simply repeating the definition of opportunity cost) and is empty in itself. Nonetheless, I am presenting this proposition here because I wanted to make clear the point that it is advantageous not to stay in commitment relations when opportunity cost is high. I wanted make this
65 point clear as a stepping stone for the fifth and the sixth propositions.
Trust as Emancipator from Relational Confinement The prior discussion has revealed that the formation and maintenance of commitment relations are not necessarily advantageous even under high social uncertainty. Some people may always buy electric gagets such as a TV set, a video player or a washing machine from the same neighborhood electricity shop, but many people would rather visit several electricity shops or discount stores to find the store that offers most attractive price and service. For them, savings in the opportunity cost are important considerations. Nevertheless, many people try to maintain commitment relations even when they face a situation in which it is more advantageous to give up commitment relations and seek savings in opportunity cost. It is not so easy to get out of commitment relations once they are firmly established. There are at least three reasons for this difficulty. (1) Commitment relation is, by definition, the relation in which people would not change partners despite better alternatives. They may eventually leave the commitment relation as outside opportunities keep increasing, but there must be a substantial time lag. According to this definition of commitment relations, changing partners immediately in the face of better alternatives means that a commitment relations has not been formed there. (2) Commitment partners are likely to developed mutual attraction and loyalty as a commitment relation continues. This makes them unwilling to change partners facing better alternative opportunities. While this tendency would be especially strong in the cases of friendship and courtship, it is not always the case even in business field to leave a long-lasting relation in a “business-like” manner. (3) Continuation of a commitment relation with specific partners reduces the level of trust toward outsiders, and the perceived level of social uncertainty existing outside the commitment relation comes to loom big. When one can expect greater profit by leaving a commitment relation and dealing with other people, whether one actually leaves the relation or not mostly depends on one’s level of trust in the relatively “unknown” potential partners who are not yet included in the commitment relation. The result of an experimental study to be introduced in Chapter 5 shows that the level of trust in people outside a commitment relation is reduced as a consequence of the commitment formation. This result indicates that the more strongly one is committed in a relation, the more difficult it becomes for one to leave it. In the situation in which both social uncertainty and opportunity cost of maintaining commitment relations are large, general trust, or trust in people who are not in the commitment relations, plays the role of a “booster rocket” providing necessary “thrust” for the “takeoff” from commitment relations. This is where trust
66 or willingness to trust plays its own role that can not be reduced to trustworthiness. The fifth proposition concerns the role of trust as a booster to promote exit from commitment relations. Proposition 5: Compared to high trusters (whose level of general trust, or trust in other people in general), low trusters have a stronger tendency to form and maintain commitment relations with specific partners when they face high social uncertainty.* The fifth proposition states that, facing high social uncertainty and high opportunity cost, high trusters—those who have a high level of general trust—will have a better chance to make use of advantageous relations than low trusters. That is, Proposition 6: Under high social uncertainty and high opportunity cost, high trusters will have a better chance to make more profits than low trusters. Trusting other people are often considered “irrational” in the sense that those who trust others jeopardize their own self interests. However, trusting other people can be gainful in the environment characterized by high social uncertainty and high opportunity cost. This implies that incentives to trust other people exist in situations of high social uncertainty and high opportunity cost. Here is the key to understand trust, or willingness to trust, beyond simple reflection of the trustworthiness of people around.
Trustworthiness as a Trait of the Selected and Trust as a Trait of the Selecting Now, I have explained the six propositions that constitute the emancipation theory of trust. Finally, I will discuss the incentives for trust by contrasting trustworthiness as a trait of those who are “selected” as interaction partners and trustfulness as a trait of those who “select” partners. When social uncertainty is high, being trusted by others is likely to lead to enhanced self-interest. When social uncertainty is high, it is natural that those who are trusted by others are selected as transaction or interaction partners. In this case, winning the reputation that “the person is trustworthy” leads to success. Although sneaky and untrustworthy people may also attain the reputation that they are trustworthy, the most certain way to win such a reputation is by acting in a trustworthy manner. That is to say, the incentives to exploit others are large in *
The term “commitment” is used here to refers to the yakuza-type commitment that is formed for the purpose of reducing social uncertainty.
67 socially uncertain situation, on the one hand, but there are also incentives to refrain from such exploitative behavior and act in a trustworthy manner to win a reputation of trustworthiness, on the other. This is the incentive for acquiring trustworthiness as a character trait for those who wishe to be trusted, and the incentive would be especially strong when opportunity cost of commitment relations is also high. At the same time, the same social environment provides incentives for acquiring willingness to trust as a character trait for those who trust others. The nature of the incentivse is different in the two cases. In the following discussion, let me call the opportunity cost incurred by commitment relations simply as opportunity cost. Thus, when I refer to opportunity cost, it means opportunity cost incurred by commitment relations. In a high opportunity cost environment, trustworthy people are likely to be selected as interaction partners. This is the incentive for people to develop trustworthy character. In the same environment, trustful people are more likely to leave the confines of commitment relations in search of better opportunities. Here is the incentive to develop trustful character. The emancipation theory of trust emphasizes particularly the second aspect, that is, the “selecting” aspect, because the aspect of “being selected” has been well noticed previously. The aspect of “selecting” has been almost completely missing in past studies of trust. However, with a little thinking, we can easily understand that either aspect is of no use without the other. For one, being “selected” as a partner for interaction by however many people is of no help if one is not willing to interact with them. For the other, it is meaningless no matter how many people one hopes to have relations with if one is not selected by them as a partner. Social environments of high social uncertainty and high opportunity cost provide the hotbed of both trustworthiness and trustfulness. It is necessary to cultivate both of the two at the same time to nurture either of them. The emancipation theory of trust points out the disposition to trust others in general brings in outcomes that meet one’s self-interest, and the benefit of having such a disposition exists only under special conditions (that is, when social uncertainty and opportunity cost are both high). The benefits high trusters under such an environment gain are the consequence of their willingness to leave the commitment relation with specific partners; that is, high trusters can fully exploit opportunities that lie outside the commitment relaitons. As I have repeatedly stressed, prior studies of trust were concerned only with the role of trust in consolidating relations with specific partners. It is one of the most important theoretical contributions of the emancipation theory of trust that it has demonstrated “the role of trust as an emancipator of people from the confines of stable relations.” Considering the present condition of Japanese society in which limitations of the traditional organizational principle based on stable relations are being recognized and an alternative organizational principle based on a more open relations is being sought after, more attention should be paid to the role of general trust as an “emancipator”
68 who might play a critical role in this transformation of the organizational principles. This point will be discussed again in the last chapter.
Incentive and Intention Now at the end of this chapter, I would like to clarify a misunderstanding that I may have given to the reader. I want to call for your attention in order to clarify the misunderstanding. I stated that trust is nurtured in the social environment in which people gain benefits by trusting others (namely, incentives for trusting others exist)—in the social environment in which both social uncertainty and opportunity cost are high. This assertion may be misunderstood as that in such a social environment people intentionally decide to trust others as a means of pursuing self-interest. The same can be said about the cultivation of trustworthy character traits. I do not mean to say that people in such a social environment believe that cultivation of trustworthy character traits results in one’s own benefit, or that people intentionally cultivate trustworthy character traits to seek their benefits. I am not meaning either of the above two interpretations. What I mean is quite contrary. High trusters would most likely to be believing, at least in their subjective perception of their behavior, to be completely disregarding self-interest. Those who have cultivated trustworthy character in most cases would not have intentionally worked to cultivate such a character as a means to enhance their self-interest. (Of course, there are people who think it necessary to acquire honesty and altruism to be successful, especially economically successful, as epitomized by the proverb, “honesty is the best strategy.” For instance, in the best seller book in late 18th Century Japan as well as in the U. S., the author of Self Help, Smiles, encouraged readers to adopt trustworthiness intentionally as a “strategy.”) As clarified above, I do not mean that people consciously respond to incentives. Of course, such possibilities may exist. This possibility is especially strong when the incentives are clearly visible to people. However, the incentives discussed here—i.e., the benefit that can be gained by cultivating a trustworthy character, or, the benefit that can be obtained by trusting other people in general—are not so transparent to people who face them. Especially, the incentives for trusting exist in its indirect consequence of promoting exit secure and stable relations. Such an indirect consequence of trust has been hardly discussed in prior studies of trust. It is unlikely that ordinary people acknowledge the indirect consequence that has not been noticed even by expert researchers and behave accordingly. What I mean by Proposition 6 that there exist incentives for trusting other people is that, in such as situation, high trusters (even those who are subjectively disregarding their own self-interest) are endowed with some benefits as a consequence of their behaving on their trust of which they may or may not be aware of.
69 The principle that consciously non-self-interested actions turn out to engender outcomes that contribute to self-interest is summarized in a Japanese proverb: “You will be helped in the shallow only when you throw yourself into the water.” Proposition 6 may be paraphrased in terms of this proverb. It is only by throwing yourself into the water (that is, by not being preoccupied with self-interest, and by trusting others) that you will be helped (that is, you will receive benefits), and that help comes in a particular environment, the shallow (in the environment of high social uncertainty and high opportunity cost).” Incentives for acquiring a character trait such as trustworthiness or trustfulness thus means the existence of the social environment in which people who have such a character trait and behave accordingly disregarding immediate self-interest end up with more rewards than those who are preoccupied with their own self-interest. Those who are attracted by immediate benefits of exploiting others may eventually get a poor reputation, so that no one will want to deal with them. In the social environment in which reputation plays the critical role, those who respect ethical values more than self-interest are more likely to be successful than non-ethical egoists. In the socially uncertain situation in which opportunity costs are high, those who only wish to avoid exploitation by others are likely to miss better chances before their eyes. When such chances exist, those who trust other people are more likely to be successful. As discussed above, I am not saying that people always consciously pursue self-interest. The point I want to make is that there exists situations in which those who do not consciously pursue self-interest—i.e., those who have a trustworthy character or those who trust others in general—do better than those who consciously pursue self-interest.
Chapter 4 Security in Japan, Trust in the U. S.
In Chapter 3, I presented the argument that trust toward general others emancipates people from the confines of fixed relations and, thus, having a high level of general trust may serve one’s self-interest when the level of opportunity costs is high. The argument was summarized as the emancipation theory of trust. In this chapter, I present results of a US-Japan comparative questionnaire survey my colleagues and I conducted, and examine whether or not the US-Japan differences derived from the theory actually exists. Thus, I examine some of the propositions that constitute the emancipation theory of trust.
The US-Japanese Comparative Questionnaire Survey In Chapter 1, I presented one result of a US-Japan comparative questionnaire study conducted by our research group that indicated that Americans have a higher level of general trust than do Japanese. In this chapter I present the survey results in more detail. Furthermore, I discuss other US-Japan differences that were predicted using the emancipation theory of trust.* This survey was conducted in both the U.S. and Japan. In each country, two samples—a student sample and a general population sample—were used. Student sample The student sample in this study is a convenience sample and is not representative of any well-defined population. The reason for including the student sample is to evaluate potential bias in previous results obtained from the student sample, by comparing results from the general population sample with those from the student sample. The questionnaire was administered in college classes in each country. In America, a total of 246 responses were collected of which 152 were from University of Washington students and 94 from UCLA (the University of California at Los Angeles) students. In Japan, a total of 928 responses were collected of which 225 were from Hokkaido University, 14 from Saitama university, 90 from Toyo University, 161 from Osaka International University, and 438 from Bukkyo University. General Sample Random samples were drawn from the city telephone books in both Sapporo, Japan and Seattle, Washington, in the U. S. A total of 300 names in Sapporo and 450 names in Seattle were randomly drawn from the telephone book. Responses from 208 people in Sapporo (75.1% of 277, excluding those for whom the address was incorrect) and 265 people in Seattle (64.3% of 412, excluding *
Details of the design, procedures, and results of the study are reported in Yamagishi & Yamagishi (1994).
71 incorrect addresses) were obtained. Pilot Study 1 The questionnaire used in this survey was constructed according to the following procedures. The first version of the questionnaire was constructed in Japanese, and a pilot survey using the questionnaire was conducted with 369 students in April and May of 1993. Items in the questionnaire directly related to trust were selected from Rotter’s Interpersonal Trust Scale (Rotter, 1967), Wrightsman’s Philosophies of Human Nature Scale (Wrightsman, 1974), Rempel and Holmes’s Trust Scale (Rempel & Holmes, 1986), Rosenberg’s Trust Scale (Rosenberg, 1957), and Yamagishi’s Trust Scale (Yamagishi, 1986, 1988a). In addition, a few new items were created and added to the questionnaire. The main purpose of this study was to conduct item analysis to select items to be used in the main study. Pilot Study 2 Based on results of the first pilot study, the questionnaire for the second pilot study was constructed in Japanese and then translated into English. The Japanese version of the questionnaire was used in a study with 394 Japanese students in June of 1993, and the English version was used with 300 American students in June and September of 1993. The purpose of the second pilot study was to examine whether or not items selected from the first pilot study were suitable for examining predicted US-Japan differences. Analysis of the results revealed no serious problems and we decided to move on to the main study. Back Translation Based on results of the two pilot studies, a questionnaire including 78 items with Likert-type response scales, two binary-choice items, and 6 scenario-type questions were first constructed in Japanese. These question items were then translated into English, and finally back into Japanese. The homogeneity of the two Japanese questionnaires (the original version and the back translation version) was examined with 40 Japanese students in the following two ways. First, the same 40 respondents answered both versions so we could examine item by item whether or not their responses were different between the original version and the back translation version. Second, both the original question and the back-translated question were presented, and the respondents evaluated how close the two questions were in meaning. Several bilingual researchers discussed items for which the back translation was not found to be compatible with the original in the above analyses and the items were revised accordingly.
Comparison of General Trust between Japanese and Americans As presented briefly in Chapter 1, results of the US-Japan comparative questionnaire study indicated that Americans are more trustful of other people in general than are Japanese. First, for both student and general samples in each nationality, a principal component analysis of the 32 items measuring respondents’ trust in other people (19 items concerning trust toward general others and 13 items about trust toward
72 specific others in personal relations) was conducted. In all sub-samples, this revealed a two-factor structure consisting of trust toward general others and trust toward specific others. Next, for a combined sample of both student and general population samples in both the U.S. and Japan, a principal component analysis was conducted of the 19 items concerning trust toward general others. This revealed two factors—a general trust factor and a caution factor. The two factor structure is consistent with the factors obtained in previous studies introduced in Chapter 2. The same two factors were reproduced in separate analyses with each sub-sample.
Table 4.1 Means on the general trust scale, and on individual items included in that scale.
Male Female Female Male General Students Students General Japan US Japan US Japan US Japan US n=583 n=75 n=330 n=124 n=167 n=138 N=39 n=106
General Trust Scale
Most people are basically honest
Most people are trustworthy
Most people are basically good 3.20 and kind
Most people are trustful of others
I am trustful
Most people will respond in kind when they are trusted by others
A bold number indicates that the mean is significantly higher than the mean in the other country within the sub-sample. Actual N may differ from item to item due to missing values.
Items with high loadings on the first factor, or the general trust factor, were about belief in the honesty and trustworthiness of people in general. In the following analysis, six items with high loadings on this factor were selected and their means were used as a General Trust Scale. Table 4.1 shows, for each sub-sample, the overall mean on the General Trust Scale along with individual item means. As shown in Table 4.1, the mean on the General Trust Scale for the American sample was higher than that for the Japanese sample. This Japanese-American difference was consistent with the results of previous studies introduced in Chapter 1. The fact that the Japanese-American difference exists in each sub-sample, as shown in Table 4.1, suggests the difference is general. The Japanese-American difference is
73 substantial for every sub-sample, although it is especially large for male students for whom the mean difference is larger than the pooled standard deviation. A more important point is that each of the six items composing General Trust Scale shows the same Japanese-American difference when the difference is statistically significant at all, indicating that the Japanese-American difference in general trust is not caused by only a few items. Considering that the difference is consistent with the result of the US-Japan comparison by the Japanese Institute of Statistical Mathematics, the conclusion that Americans have a higher level of general trust than do Japanese clearly is robust.
Importance of Commitment Relations The prediction that American society should have a higher level of general trust than does Japanese society is derived from the emancipation theory of trust that emphasizes uncertainty reduction by commitment relations insofar as we accept the assumption that stable commitment relations play more important social roles in Japan than in the U. S. That prediction has been confirmed by the survey results shown above as well as by the results of the survey conducted by the Japanese Institute of Statistical Mathematics. In this section, I examine whether or not Japanese really consider commitment relations more important than do Americans in order to demonstrate the validity of the assumption that stable commitment relations are more important in Japanese society than in American society. In comparing the significance of commitment relations in Japan and in the U.S., we need to consider one issue first. It is the difference between yakuza-type commitment relations and lovers-type commitment relations. As explained in Chapter 3, people facing high social uncertainty form yakuza-type commitment relations to reduce social uncertainty inside the relations. In such yakuza-type commitment relations, one need not trust one’s interaction partners. Conversely, yakuza-type commitment relations are relations in which one feels safe interacting even with untrustworthy people. A typical example of this yakuza-type commitment relation can be found in popular samurai movies in which a corrupt local magistrate offers a special privilege to greedy merchants in exchange for bribes. The corrupt local magistrate accepts the bribe not because he trusts the goodwill or integrity of the greedy merchant. While saying to the merchant, “You are a really greedy fellow,” he feels safe accepting the bribe. He feels safe because he knows there is no incentive for the merchant to disclose the bribe to anyone outside. At the same time, in giving the magistrate the bribe, the merchant does not trust the magistrate’s goodwill or integrity. He replies to the magistrate, “My lord, I am no match with you in that respect, ha, ha, ha,” and yet feels safe in the deal since he knows that the magistrate needs to give him the special privilege in order to receive future bribes from him. A yakuza-type commitment relation is a type of relation in which partners
74 benefit from acting according to the principle of “in-group favoritism,” and the maintenance of which is based on mutual understanding of the mutual benefit derived from in-group favoritism. Thus, in order to investigate how important yakuza-type relations are as a source of security for people, it is necessary to examine how strongly people expect interaction partners in specific relations to act according to the principle of “in-group favoritism.” The following three items were included in the US-Japan comparative questionnaire in order to examine how strongly people expect a particular partner to practice “in-group favoritism” in their relations with him or her: “If I were going to buy a used car, I would feel more comfortable buying it from a salesperson whom my friend had introduced me to in person rather than from a salesperson who is a total stranger.” “When negotiating over an important issue with a total stranger, it is very important to have a personal introduction by someone you know well.” “A doctor examines a patient more carefully than usual if the patient has been referred by a personal acquaintance.” All of these items measure the belief that things go better when using personal relations. In this sense, they are directly related to the belief that maintaining relations with others serves utilitarian goals. In contrast, a lovers-type commitment relation is a type of relation based on trust in the partner’s human nature. However, trust in the partner in lovers-type commitment relations may not necessarily be based on expectations concerning the partner’s general human nature or integrity. It also may be based on expectations that your partner feels positively toward you. For instance, a yakuza’s mistress who knows him to be cold and cruel toward others nevertheless would expect him to be nice to her, because she thinks he loves her. Her expectation is based on the state of the yakuza’s mind as she infers it. In this sense, her expectation of nice behavior from him should be treated as trust, not as assurance. The expectation that a commitment partner will not exploit me often represents assurance of security produced by a yakuza-type commitment relation. However, it could be interpersonal trust developed in a lovers-type commitment relation. In this study, the following four items were included to measure relational trust, which is clearly distinct from the expectation of “in-group favoritism” that accompanies yakuza-type commitment relations. “I trust a person I know well more than one whom I don’t know.” “Whatever work I have to perform, I feel more secure when I work with someone I know well than with someone I don’t know.” “Generally, a person with whom you have had a long relationship is likely to help when you need it.” “The people I trust are those with whom I have had long-lasting relationships.” A principal component analysis of the seven items about in-group favoritism and relational trust showed two mutually independent factors. As expected, the three items about the expectation of in-group favoritism and the four items about relational trust formed mutually independent factors. This result shows that the belief that one can expect real benefits from keeping stable commitment relations is
75 independent from the belief that intimately related people are trustworthy (relational trust toward intimate associates), and is consistent with the argument presented in Chapter 2 that the two are conceptually distinct. In respondents’ minds, the expectation of in-group favoritism, that better outcomes are obtainable through proper, stable relations, is separate from the belief that partners in stable relations are more trustworthy than those outside such relations. Table 4.2 compares, in each sub-sample, the Japanese and the American means of each of the three items related to in-group favoritism and the overall mean of the “Utility of Relations Scale,” composed of the three items. As expected, Japanese had a stronger belief than Americans that one receives special treatment by using personal relations and that in-group favoritism is widely practiced in the society. Considering the fact that a strong Japanese-American difference exists in each of the four sub-groups (male students, female students, male general population and female general population), we may conclude safely that a rather distinct Japanese-American difference exists with regard to the belief that maintenance of commitment relations secures special treatment from commitment partners. That is, Japanese have a stronger expectation that people generally practice in-group favoritism.
Table 4.2 Means of the utility of relations scale and their constituent items in each sub-sample
Items Utility of Relations Scale
Male Female Male Female Students Students General General Japan US Japan US Japan US Japan US n=583 n=75 n=330 n=124 n=167 n=138 n=39 n=106 3.82
If I were going to buy a used car, I would feel more comfortable buying it from a salesperson 3.98 whom a friend had introduced me to in person rather than from a salesperson who is a total stranger When negotiating over an important issue with a total stranger it is very important to 4.01 have a personal introduction by someone you know well A doctor examines a patient more carefully than usual if the patient 3.47 has been referred by a personal acquaintance
A bold number indicates that the mean is significantly higher than the mean in the other country within the sub-sample. Actual N may differ from item to item due to missing values.
76 Now, let us examine what differences exist between the U.S. and Japan in relational trust toward familiar people. Table 4.3 reports the Japanese-American comparison, in each sub-sample, of four items related to relational trust toward familiar people and the Relational Trust Scale composed of the four items. This table indicates a rather consistent Japanese-American difference in relational trust toward familiar people. Regarding the overall relational trust score—an average of the four items—Americans have a higher mean score than Japanese within each of the four sub-samples. Moreover, the same Japanese-American difference is observed for most of the items and sub-samples. All together, these results indicate that Americans have a stronger tendency than Japanese to trust familiar people, similar to what we have seen already for general trust.
Table 4.3 Means for the relational trust scale and its constituent items in each sub-sample Male Students
Items Relational Trust Scale
Japan US Japan US n=330 n=124 n=167 n=138
Female General Japan n=39
3.91 4.46 3.87 4.37 3.93 4.24 3.83 4.19
I trust a person I know well more 4.24 4.49 4.13 4.76 4.01 4.55 3.95 4.61 than one whom I don't know Whatever work I have to perform, I feel more secure when I work 4.14 4.23 3.95 3.80 4.21 3.75 3.82 3.58 with someone I know well than Generally, a person with whom you have had a longer relationship 3.80 4.62 3.88 4.53 3.70 4.43 3.64 4.35 is likely to help you when you The people I trust are those with whom I have had long-lasting 3.47 4.51 3.51 4.44 3.85 4.25 3.92 4.23 relationships A bold number indicates that the mean is significantly higher than the mean in the other country within the sub-sample. Actual N may differ from item to item due to missing values.
Those results may be summed up in the following way. On the one hand, Japanese have a stronger belief in the merit of practicing in-group favoritism, the belief that one can obtain special treatment from particular people by maintaining relations with them. This result demonstrates a Japanese-American difference in the importance of commitment relations (as a mechanism of reducing social uncertainty and thus providing assurance of security) at least at the level of people’s subjective
77 belief. This difference had been assumed in deriving the Japanese-American difference in the level of general trust. On the other hand, the results above also indicate that Americans not only have a higher level of general trust toward general others but also have a higher level of trust toward specific people in familiar relations (relational trust). This is consistent with my assertion that Japanese society is a society characterized by assurance of security but not by trust at the relational trust level as well as the general trust level.
Importance of Reputation Reputation plays two roles─one for the target of the reputation and one for the user who receives it as information. Suppose, for instance, Mr. A got a reputation as “a man who often speaks ill of his friends.” This reputation works as a “punishment” for him. On the other hand, a reputation such as “Mr. A is a really nice person. He cares about his friends when they are in trouble” would work as a “reward” for him. Thus, reputation plays the role of controlling a person’s action by working as a reward or a punishment. This aspect of reputation may be called the controlling role of reputation. At the same time, reputation works as a source of information concerning the target’s character for those who hear the reputation. This aspect of reputation may be called the informational role of reputation. The difference between the two roles of reputation has important implications for Japanese-American differences from the emancipation theoretic perspective, since the importance of the informational role of reputation is expected to be greater as social relations become more open. In an extreme case, the informational role of reputation is extremely small in a situation in which all social interactions are performed through a network of stable commitment relations. In such a case, the level of social uncertainty is low because security has been assured, and thus need for information about other people’s trustworthiness is small. To put it the opposite way, reputation as information is valuable only in situations in which social uncertainty exists. From this point of view, a situation in which reputation as information has great value could be a situation in which trust plays an important role. In situations in which most social relations are fixed and security is assured inside each relation, neither trust nor the informational role of reputation is needed. Those two are needed in social situations with high mobility, in which those who stay in stable relations incur large opportunity costs. The distinction between “weak ties” and “strong ties” in a classic study of social networks by Mark Granovetter (1974) is relevant here. He termed close relations with people with whom one frequently spends time “strong ties.” Conversely, casual relations with someone with whom one spends time infrequently are “weak ties.” Commitment relations defined in this book correspond to Granovetter’s strong ties, although some strong ties are not commitment relations.
78 Granovetter argues that people who are bound to each other with strong ties tend to form a closed circle (commitment group), and thus opportunities for them to reach outside of the mutually strongly connected group tend to be limited. As a consequence, those who rely mostly on strong ties have limited access to information from the outside world. In contrast, those who are connected to a variety of people with weak ties have access to a wide variety of information. Using our terminology, those who are bound mutually with strong ties may live a secure life, and yet, they have to pay an opportunity cost in the form of reduced access to information. Granovetter therefore can explain the finding of his study that weak ties rather than strong ties help people find new jobs in American society. This occurs because one can obtain more information through weak ties that reach out of the closed circle—and thus the possibility of finding a suitable job is greater—than through strong ties among friends. In addition to the intensity of ties suggested by Granovetter, another feature of commitment relations, especially yakuza-type commitment relations, is internal mutual control as discussed by Axelrod. People form stable commitment relations with particular partners in order to make sure they can obtain needed resources from the partners. The resources they need may be physical or economic resources such as money and services, psychological resources such as love and understanding, or social resources such as respect and status. The more important the resources are, the more important it is to secure a stable source of them. Yakuza-type commitment relations, in this sense, are relations in which resources of critical importance are exchanged. In contrast, in exchanges of information, not of needed resources, commitment relations do not reduce social uncertainty. Partners of information exchange do not control each other’s behavior even in commitment relations. What Granovetter termed a network of weak ties is a network of information exchanges in which exchanges of substantial resources do not play a major role. The key feature of such relations from our point of view is the lack of mutual control. Given the above discussion, the open social environment that we have discussed up to now is characterized by a prevalence of weak ties for information exchange rather than a network of stable commitment relations for resource exchanges. In such a social environment, the informational role of reputation will be important because, in the absence of mutual control, it is necessary to assess how trustworthy other people are. Looking at it from the other side, when dealing with a stranger in a society in which strong ties of yakuza-type commitment relations are ubiquitous, people would rather try to obtain controllability over that person than try to assess his or her trustworthiness using his or her reputation as information. They can create a secure relation with that person if they can control him or her. Such a relation would be much more secure than relying on the assessment of his or her trustworthiness by making use of reputation. A method often used to obtain controllability is to go through indirect control via a third party whom one can
79 control and who can control the target person. From this point of view, Granovetter’s conclusion that weak ties rather than strong ties are important in finding jobs may have come from the fact that yakuza-type commitment relations do not play an important role in American society. In societies in which commitment relations play a more important role, for example, Japanese society, getting connected to influential contacts via strong ties may be more important than obtaining proper information in order to find a job. The advantage of being connected to influential contacts is not only that one gets information through influential contacts, but more importantly, the influential contacts play the role of a guarantor. That is, an employer who hires an employee can feel safe if the influential contact is in a position of controlling the employee and mutual control exists between the influential contact and the employer. This is because the employer can control the employee indirectly through the third party, not because the third party guarantees the employee’s quality. That is, the influential contact plays the role of a hostage (collateral) by the employee. From this perspective the controlling role of reputation, in contrast to the informational role, is expected to play a more important role in Japanese society, in which networks of commitment relations with strong ties are prevalent. This prediction has been confirmed by a study by Watanabe (1991), which demonstrated that strong ties instead of weak ties are more effective in finding satisfactory jobs in Japan. Although the questionnaire study we conducted did not examine strong ties and weak ties directly, results of the study basically agree with the prediction. The following item was used to tap belief about the importance of the controlling role of reputation, “Most people refrain from dishonest conducts to avoid getting a bad reputation.” Another item tapped belief about the importance of the informational role of reputation, “A person’s reputation is not very useful in judging his or her true character.” (Responses to this item have been reversed in the following presentation such that a larger number means stronger belief in the importance of the informational role of reputation.) Results are shown in Table 4.4. First, as expected, American respondents in all sub-samples rated the importance of the informational role of reputation higher than did Japanese respondents. Namely, in judging a person’s trustworthiness, American respondents considered reputation to be more useful than Japanese respondents. Concerning the importance of the controlling role of reputation, the predicted difference existed, though it was not as pronounced as in the case of the informational role. There is hardly any Japanese-American difference in the student samples, but in the general population samples, especially the male sample, the difference—Japanese respondents consider more strongly that reputation controls people’s behavior than do American respondents—is clearly seen, as expected. This result supports the prediction from the emancipation theory of trust that the
80 difference in the degree to which yakuza-type commitment relations are prevalent in Japanese society and in American society, as well as the level of opportunity costs the different levels of commitment create in the two societies, determine the differential importance of the two roles of reputation—in addition to the differential importance of general trust—in the two societies .
Table 4.4 Means for the importance of the informational role and the controlling role of reputation in each sub-sample Male Students Japan n=583
Informational Role of Reputation: A person's reputation is not very useful in judging his or her true character Controlling Role of Reputation: Most people refrain from dishonest conduct to avoid getting a bad reputation
Japan US Japan US n=330 n=124 n=167 n=138
Female General Japan n=39
2.43 3.01 2.41 3.15 2.57 3.75 2.45 3.70
3.45 3.61 3.27 3.15 3.70 3.16 3.51 3.15
A bold number indicates that the mean is significantly higher than the mean in the other country within the sub-sample. Actual N may differ from item to item due to missing values. The reported mean for the controlling role of reputation is based on reversed scores, such that greater values indicate greater importance of the controlling role of reputation.
Honesty and Justice With the growth of Japanese-American business frictions and mutual criticism of each other’s business practices, Americans often voice the opinion that “the Japanese are unfair.” It is clear that a value judgment that fair behavior is good and unfair behavior is bad exists behind such an opinion. It is natural, then, for most Japanese to be unwilling to accept this opinion. However, once the value judgment part is taken away, the opinion is consistent with the predicted Japanese-American difference based on the emancipation theory. The difference between being unfair and fair, once the value judgment is taken away, is the difference between applying different standards to oneself (or one’s friends) and to other people and applying the same standard universally. In short, it is a question of practicing in-group favoritism or not. In-group favoritism often is considered to be morally undesirable, but under some circumstances it is rather desirable. For instance, there is a type of leadership style loved by Japanese called oyabun hada, in which the leader unconditionally
81 protects his subordinates from outside attacks and represents his subordinates’ interests in negotiating with others, rather than treating subordinates and outsiders according to the same universal principle. Subordinates who have such a leader can feel safe in concentrating on their roles. Let us consider the relationship between a professor and graduate students in Japan. Though it has changed somewhat recently, university professors are recruited mostly through personal connections or the old-boy network. For the sake of simplifying the argument, suppose all Japanese universities recruit professors only from their own graduates. This is not exactly the case, but let us just imagine such a situation. Given such a situation, what behavior would graduate students who dream to be scholars in the future desire and expect of their professor? Suppose there is a vacancy in the department and the professor has a decisive say in the choice of the candidate. Should the professor hire the candidate whose performance is highest regardless of whether or not the candidate is her student? Or, is it more desirable for her to recruit her own student insofar as the student has demonstrated a sufficiently high level of academic performance? Of course, the answer to this question depends on whether or not the person who answers this question is the professor’s student. Most people outside Japanese academia would think it improper to favor one’s own student in hiring a new professor, but it would be a disaster for her students if the professor acted according to the universal standard. If other colleges acted according to the same universal principle in hiring professors, her students would have a fair chance. However, what if other colleges recruit only their graduates and this professor was the only one who acted in a universalistic manner? Her students would be greatly handicapped. A professor who insisted on objective performance as the only criterion for her hiring decisions would be considered a “cold professor,” who cared little about her students. Under such circumstances, a professor who treats her own students favorably is likely to be considered a desirable professor, and such “unfair” behavior is likely to be considered morally right. What I wanted to convey with this example is the point that, in a situation in which networks of closed relations prevail, practicing in-group favoritism not only is beneficial for oneself as discussed earlier, but also is a desirable behavior that is expected by people around, especially by those who stay in the commitment relations. Stated more extremely, in a social situation in which networks of commitment relations play critical roles, “fair” behavior based on a universalistic standard without making any distinction between those inside and those outside commitment relations is likely to be regarded as undesirable or even morally wrong behavior. Conversely, the practice of in-group favoritism that treats “friends” within commitment relations with special favor is likely to be regarded as desirable and morally right. To sum up the above discussion, in-group favoritism, that is, unfair behavior
82 (from the point of view of people outside the group or the relation), is a desirable behavior within commitment relations. Conversely, it is likely to be regarded as undesirable and unfair in a situation in which commitment relations do not play important roles. Let me continue the discussion a little further. I argued that high trusters have a chance to perform better than low trusters in an open social environment in which commitment relations do not play important roles and opportunity costs are generally large. However, in order for this argument to hold true, those who have left a stable relation in search of better opportunities have to be sought by other people as a desirable partner (or, at minimum, not be refused by other people). Those who have left stable relations cannot form better relations unless they find alternative partners. In the previous section we have seen the importance of the informational role of reputation in a society where commitment relations are not prevalent. Those who behave dishonestly or unfairly toward outsiders are likely to spoil their reputations, and to find it difficult to be welcomed as new partners once they leave their current relations. Thus, it is necessary that those who have a high level of general trust at the same time should be honest (at least, they should believe that honesty is important) in order for general trust to play the role of an “emancipator from closed relations.” In contrast, in a society in which networks of commitment relations prevail and opportunity costs of staying in commitment relations are generally small, there is little incentive for people to act honestly and fairly toward unfamiliar people located outside their commitment relations. What they must not do is betray partners within commitment relations. If they act dishonestly toward the rest and thus reduce the chance of being chosen as a new partner by “outsiders,” nothing will occur to damage their self-interest. The claim that Japanese are unfair makes a certain sense from this perspective. If we accept that commitment relations play important roles in Japanese society, or at least more so than in American society, unfair treatment to unfamiliar people outside commitment relations (that is, treating them more unfavorably than in-group members) should be regarded as undesirable more in American society than in Japanese society. Furthermore, such unfair deeds would have more undesirable consequences to the actor him or herself in American society than in Japanese society. That Japanese are unfair does not mean that Japanese are morally wrong. Rather, it means that Japan is a society in which even moral people are expected to act unfairly (namely, to practice in-group favoritism). The five items shown in Table 4.5 were included in the questionnaire to examine the subjective significance of honesty and fairness toward unfamiliar people outside commitment relations. A principal component analysis of the five items revealed a unidimensional factor structure in each of the four sub-samples. Accordingly, the average of the five items was used as an Honesty/Fairness Scale, measuring belief in the importance of honesty and fairness. As Table 4.5 shows,
83 there is a large Japanese-American difference as expected in belief in the importance of honesty and fairness. First, the scale value—the average of the five items—shows a statistically significant Japanese-American difference in each of the sub-samples. American respondents consider honesty and fairness more important than do Japanese respondents. At the individual item level, significant Japanese-American differences are observed in the predicted direction. In most item/sub-sample combinations, American respondents consider honesty and fairness to be more important than do Japanese respondents
Table 4.5 Means of the honesty/fairness scale and its constituent items in each sub-sample Male Students
Japan US Japan US Japan US Japan US n=583 n=75 n=330 n=124 n=167 n=138 n=39 n=106
Honesty/Fairness Scale *I don't want to miss out on good opportunities while trying to be fair to others *Telling a lie can be justified depending on the circumstance I don't want to act dishonestly under any circumstances I am mindful not to forget the spirit of fair play under any circumstances *Being overly concerned about fairness deprives a society of its vigor
2.81 3.39 2.99 3.59 3.43 4.00 3.41 4.06 2.60 2.92 2.66 2.98 2.94 3.14 3.05 3.25 1.68 2.29 1.74 2.65 2.46 3.19 2.00 2.92 3.31 3.68 3.57 3.78 4.07 4.47 4.08 4.60 3.46 3.91 3.66 3.74 4.13 4.40 4.21 4.49 3.15 4.51 3.25 4.67 3.92 4.82 3.79 4.84
A bold number indicates that the mean is significantly higher than the mean in the other country within the sub-sample. Actual N may differ from item to item due to missing values. The reported means of items starting with an asterisk (*) are based on reversed scores, such that greater values indicate greater importance of honesty or fairness.
Summary of the US-Japan Comparative Study This chapter presented in more detail results of the US-Japan comparative questionnaire study briefly presented in Chapter 1 as evidence showing that Americans have a higher level of general trust than do Japanese. The cross-societal comparison was based on the assumption that the role that commitment relations play in society differs in importance between Japanese and American societies. Whether this assumption is correct or not cannot be judged from the questionnaire study per se, but its results confirmed the expected difference at a subjective level.
84 That is, Japanese respondents were shown to have a stronger belief that maintaining relations with specific partners yields practically useful consequences than American respondents. It was predicted further that reputation would be regarded as a more important source of information in a society in which people deal with things more in voluntarily formed relations as opposed to stable commitment relations. The Japanese-American difference predicted from this argument also was confirmed. At the same time, Japanese were shown to have another belief more strongly than Americans, the belief that reputation controls people’s behavior. It was argued further that acting honestly and fairly toward others outside commitment relations is more likely to yield personally desirable outcomes in American society than in the Japanese society, and that such behavior is regarded more as desirable in the former than the latter society. Predictions derived from this argument also were confirmed. These survey results by themselves of course do not guarantee the validity of the emancipation theory of trust. A questionnaire survey is generally not a suitable research methodology for testing a theory consisting of causal propositions. Furthermore, since the theory had not been formulated well when the research was planned, the questionnaire items included in the questionnaire may not be necessarily the most suitable for the theoretically relevant concepts. Thus, the reader is advised to take findings of the study with respect to the Japanese-American difference in general trust as well as other Japanese-American differences as proof that the emancipation theory of trust is not too greatly off the mark. The conclusion cannot be stronger than this. When it comes to testing theory, the experimental studies presented in the following chapters have more direct validity than questionnaire studies. While experimental studies provide clear answers in testing propositions, their external validity—whether or not results observed in the laboratory are applicable to the world outside it—is always suspect. I discuss this issue in the final section of the next chapter. For the present, I emphasize that it is necessary to combine and use various research methodologies such as questionnaire surveys, laboratory experimentation, and computer simulation if one wants to test a theory in such a way as to convince all kinds of people of its conclusions. Each method has its own merits and limitations or problems. Results obtained using only one method cannot persuade everyone. The ability to persuade will be enhanced greatly if various methods are used and the same conclusion is obtained consistently. Results of the questionnaire study presented in this chapter, taken by themselves, also have many limitations and problems. What is important is the point that those results are consistent with the experimental findings presented in the next chapter.
Chapter 5 Trust and Commitment Formation
In Chapter 4, I tested the emancipation theory of trust by analyzing the results of the Japan-US comparative questionnaire study. The results generally supported the theory. However, a questionnaire survey is not the most appropriate method for testing a theory that consists of a series of propositions. In this chapter, I present a series of experimental studies that we conducted to test the theory. Although most of the experiments presented in this chapter were conducted both in the U.S. and Japan, the purpose was not to claim that there is a so-called “cultural difference” between the U.S. and Japan. Rather, the purpose was to test the validity of the theory, on the premise that any “cultural differences” would disappear once we controlled the variables that play critical roles in the theory. More detailed discussion of this point will appear later in this chapter when it is needed.
Experiment 1 The design of the first experiment was based on Kollock’s (1994) experiment that I introduced earlier.* In the experiment that simulated trades of goods, Kollock (1994) manipulated (i.e., an experimenter changed the state as part of the experimental design) the level of social uncertainty by allowing or not allowing buyers to know the quality of the goods at the time of purchase. Results showed that social uncertainty facilitates commitment formation between particular sellers and buyers. Furthermore, results showed that commitment formation produces higher trust between exchange partners within commitment relations. However, “trust” in Kollock’s experiment is “relational trust” according to our definition. Kollock (1994) did not examine how general trust is related to commitment formation or relational trust. In other words, while results showed that relational trust in a particular partner (i.e., the expectation that the particular partner would not hurt me at least) emerges once a person forms a commitment with the partner, they showed nothing about the effect of general trust on commitment formation. Further, the effect of commitment formation on trust in people outside the commitment relations has not been examined, either. The purpose of our first experiment described below was to investigate those two questions. Experiment 1 differed from Kollock’s (1994) in several other aspects. We discuss settings in detail later, but the primary difference is that all buyers and sellers *
Results of the experiment conducted in Japan were reported in Yamagishi et al (1995). The entire set of results including both Japanese and U.S. data is reported in Yamagishi, Cook, & Watabe (1998).
86 except for one participant were computer responses in our experiment, whereas all sellers and buyers were real humans in Kollock’s (1994). Also, we manipulated (i.e., changed) social uncertainty differently. To put it briefly, the degree of social uncertainty is much greater in our experiment than in Kollock’s (1994). I would like to emphasize one other matter before we get into the description of settings of the experiment. That is, our experiment was conducted as a “cross-societal experiment.” Usually, when researchers conduct the same experiment in different countries, it is called a “cross-cultural experiment.” However, I consider this experiment to be not a cross-cultural experiment, but a cross-societal experiment. Because I am the only one who makes this distinction, I discuss it further.
Cross-Cultural Experiment and Cross-Societal Experiment The experiments presented in this chapter were conducted in both Japan and the United States. That type of experiment is usually called a cross-cultural experiment. In most cases, the purpose of a cross-cultural experiment is to show that human minds work differently in different cultures. One of the classic cross-cultural experiments is the series of experiments on “optical illusions” by Muller and Lyer. This is a well-known phenomenon that any person who has taken an introductory psychology class would know. When people compare the two lines below, which have the same length, they perceive that line (B) is longer than line (A). A large number of studies have been conducted with this illusion in different cultures, and it is now known that this illusion does not occur strongly in some non-Western countries. Thus, although this illusion once had been considered universal, repeated experiments in different cultures revealed that it is actually a phenomenon that is heavily affected by culture.*
Cross-cultural experimentation has advanced greatly both methodologically and theoretically since the time of the experiments on Muller and Lyer’s optical illusion. However, the basic purpose has stayed the same. The purpose is to find out *
The “cultural difference” has been explained as follows. The illusion occurs in cultures where there are many objects that have straight lines in everyday life (e.g., buildings, streets, etc.). In other words, it occurs among people who have grown up in an environment that has artifacts with sharp edges, but does not occur among people such as native Africans who live in an environment that does not have artificial objects that have sharp edges. However, various questions have been raised about the validity of this explanation , and no agreed-upon answer has been provided.
87 the commonalities and differences between people’s minds in different cultures. However, the purpose of the experiments that we discuss later is fundamentally different from that of cross-cultural experiments. The purpose of our experiments is not to show that people in different cultures think or behave differently. Rather, it is to show that people’s behavioral patterns are the same even in different cultures, once we control for the variables that are theoretically relevant. Our experiments start from this assumption: people behave differently in different cultures because the values of variables that affect that behavior are different in different cultures. Therefore, it follows logically that even people who share the same culture will behave differently if we change the values of relevant variables. Also, even if the cultures that they belong to are different, participants’ behavioral patterns will be the same if we can set the appropriate variables in the experimental settings to be equal across cultures. Experiments that have this purpose—that is, to eliminate so-called cultural differences in behavioral patterns by equating the values of the theoretical variables—are called cross-societal experiments here. I make this distinction clear in order to emphasize the viewpoint that what is important is not differences among cultures that are internalized in people’s minds, but variables that exist in the social environment. From this viewpoint, if we still see cultural differences in participants’ behavioral patterns after we control the variables that are considered important based on the theory, the residual cultural difference is evidence that the theory is not appropriate. In other words, the validity of the theory is proven only when we can specify enough theoretically relevant variables to eliminate the so-called cultural difference. Now, let us take a close look at this argument with the emancipation theory of trust. The reason we conducted the same experiments in both Japan and the United States is not to see cultural differences that might exist between these two countries. Rather, the purpose is to test the validity of the theory. For example, as I discussed earlier, it has been argued that the tendency to form committed relations with specific partners varies between the United States and Japan. I have no intention of rejecting this point. I believe that this difference exists because the opportunity costs for maintaining commitment relations and the level of general trust are different between the United States and Japan. Assuming this explanation is correct, then there should be no difference between American participants and Japanese participants in their tendency to form commitment relations with specific partners to reduce social uncertainty once we create experimental settings in which Japanese participants and American participants face the same levels of social uncertainty and opportunity cost and the levels of general trust are matched between American and Japanese participants. If we still see differences between Japanese and American participants in the tendency to form commitments even after setting the three theoretically important variables that affect commitment formation—social
88 uncertainty, opportunity cost, and participants’ general trust—at the same level across the two participant groups, then our theory will be found to be incomplete. In that case, we should reexamine the theory and try to develop a more appropriate theory. Through this process, I believe that the theory’s validity will increase.
Purpose of the Experiment In the broadest context, the primary purpose of the series of experiments shown in this chapter is to test the validity of the emancipation theory of trust. However, this ultimate goal cannot be achieved by one or two experiments. What we can do in each experiment is pick up one part of the theory and test hypotheses derived from that specific part of the theory. Specifically, the purpose of the first experiment was to test the following two hypotheses. The first hypothesis was derived from the second proposition introduced in Chapter 3. Hypothesis 1: Social uncertainty facilitates commitment formation. This hypothesis was originally proposed by Emerson and Cook, my mentors, in their research on commitment formation in social exchange networks (Cook & Emerson, 1978). As we discussed earlier, this was supported later by one of their students (Kollock, 1994). Once a commitment relation is formed with a specific partner, people come to behave cooperatively in order to secure a long-term benefit within that relationship. In other words, within commitment relations in which “future profit” based on the long-term relationship exists, the future benefit that one can furnish to the partner provides “assurance of security” that keeps the partner from behaving dishonestly. This assurance eliminates social uncertainty (i.e., belief that my partner might exploit me). The claim that social uncertainty is eliminated and very strong mutual cooperation is established within committed relations has been demonstrated consistently in many studies, including Jin et al.’s experiment, introduced earlier, and my experiments with Hayashi on network-PD (Jin, Sekiya, & Shinotsuka, 1993; Hayashi, 1993, 1995; Hayashi, Jin, & Yamagishi, 1993; Hayashi & Yamagishi, 1997, 1998; Yamagishi & Hayashi, 1996; Yamagishi, Hayashi, & Jin, 1994) In the network-PD situation, two actors who select each other from many possible alternatives play a Prisoner’s Dilemma. In previous research on PD, each participant’s partner was determined by the experimenter. Participants could not choose their partners. However, most situations that we face in everyday life are not of this kind. Even in relatively long-lasting relationships, such as friendships or marriages, we are not forced to form a relationship with a designated partner against our will. Therefore, in order to see what would happen in a situation in which we can choose our partners and a PD exists within the mutually chosen pairs, we started the research on network-PD. We found that, in a “selective-play situation” in which
89 actors can choose their partners and can leave relationships with their current partners, strong commitment relations form between specific pairs of actors, who mutually cooperate within the relationship. The experiments that I mentioned in Chapter 3 and related computer simulations have replicated this finding repeatedly. The first hypothesis (i.e., commitment formation is facilitated in situations with social uncertainty) thus has been supported repeatedly by Kollock (1994) and network-PD experiments. The second hypothesis is about the US-Japan difference in the tendency to form commitment relations. As we discussed repeatedly, compared to American society Japanese society is a society with strong commitment relations, especially yakuza-type commitment relations. The results of the questionnaire survey for the US-Japan comparison presented in the previous chapter also suggested that Japanese consider commitment relations more important than do Americans. The purpose of this experiment is not to show the difference in the tendency to form yakuza-type commitments between Japanese and American participants. As I discussed earlier, the goal of this experiment is not to demonstrate the so-called “cultural difference” between Japanese and Americans. Rather, the purpose of this experiment is to prove that Japanese and Americans behave similarly when theoretically important variables are set equal. According to the previous argument, two variables affect commitment formation: social uncertainty and the level of general trust. Assuming that the emancipation theory of trust that forms the basis of this line of argument is correct, there should be no difference in the tendency to form commitment relations among Japanese and American participants once those two variables—the level of social uncertainty in the experimental setting and the level of general trust among the two groups of participants—are set equal. Hence the second hypothesis: Hypothesis 2: When both the level of social uncertainty and the level of general trust are set to be equal for Japanese and Americans, there is no difference in their tendency to form commitment relations. Please note that opportunity cost, which is, according to the emancipation theory of trust, another variable that affects commitment formation, is set to be constant in this experiment.
Procedure The experimental setting represents trading practices between sellers and buyers as in Kollock’s experiment. Participants were told that there were several buyers and sellers in each group. Actually, however, there was only one real participant in each group. Although participants believed that several others were participating, all of the other “participants” were simulated actors controlled by a computer.
90 There were 200 volunteers, 100 Japanese students and 100 American students. The experiment was run in the Social Psychology Laboratory at Hokkaido University and the Social Exchange Laboratory at the University of Washington. Levels of general trust among participants were controlled so that in each country one half of the participants were high trusters and the other half were low trusters. Levels of participants’ general trust were measured by a questionnaire that potential participants filled out several weeks before the experiment. Therefore, in this experiment, there was no difference between American and Japanese participants in the level of general trust. When I designed this experiment, I wanted to test the other hypothesis that low trusters are more likely to form commitment relations than are high trusters. Unfortunately, however, data for the American participants’ individual trust scores were lost due to problems that occurred during data input. Hence, in this experiment we cannot test directly the effect of general trust on commitment formation. However, because we know that the numbers of high-trusters and low-trusters are equal among American and Japanese participants, we still can test the second hypothesis. In both the United States and Japan, each participant stayed in a small cubicle in the laboratory complex during the experiment. Each cubicle was equipped with a PC. Based on the instructions and on the information on a computer display, each participant decided what to do and used the keyboard to enter his or her decision. Therefore, there was no chance that participants would see other participants. However, because each experimental session was run with several participants, each participant heard the experimenter talking to other participants in different compartments. This procedure allowed us to eliminate unwanted and unexpected effects that might be produced by face-to-face interaction among participants and, at the same time, to reduce participants’ suspicion that they were interacting not with other participants but with a computer. At the beginning of the experimental session, each participant drew a card that determined whether he or she would play the role of a seller or a buyer in the experiment. However, there was a trick in this lot-drawing, and all participants were assigned the role of a buyer. After the drawing, each participant received 5 dollars (500 yen in Japan) as an initial endowment. The experiment consisted of many transaction periods. In each transaction period, each participant bought a commodity from one of the sellers and resold it to the experimenter. The profit that each participant made in each transaction period was the difference between the purchasing price and the resale price. If the participant purchased the commodity at a price higher than the resale price, he or she suffered a loss for that transaction. Throughout the experiment, each participant’s profits were added to the original $5 and losses were subtracted from the original $5. The amount of money that participants had at the end of the experiment actually was paid to them. In each transaction period, a buyer had to buy a commodity from one of two
91 available sellers. The “available” sellers were supposed to be determined by a lottery again. However, all participants in fact were assigned the same set of two sellers, seller A and seller C; each participant had to buy a commodity from either A or C. Please note that there was only one real participant in a group as I mentioned earlier. The participants who were assigned the role of seller A or seller C did not exist. They were actually programmed responses by a computer. As noted earlier, there were several participants in each experimental session. However, this does not mean that real participants interacted in the experiment. Each participant belonged to a different experimental group that included him or her as the only human with the other members being simulated actors. Thus, multiple groups that were mutually independent were run in each experimental session. While a buyer had two transaction partners (sellers), each seller was supposed to have two transaction partners (buyers). Therefore, if another buyer bought a commodity from a seller while a participant was deciding, the participant lost the opportunity to buy from that seller. If that happened, the participant could buy only from the remaining seller. If a participant was very slow in making up his or her mind, both sellers may have completed transactions with other buyers. In that case, the participant would not be able to complete any transaction, and the participant’s profit for that transaction period would be zero. Each simulated seller was programmed in such a way as to complete a deal with another buyer within a certain time (plus random fluctuations). Each transaction period started with sellers announcing the price of the commodity they were selling. Buyers did not know the actual quality of the commodity. If the quality of the commodity was standard, the experimenter bought the commodity for $1.40 (140 yen in Japan). Thus, if the quality of the commodity was standard and the purchasing price was less than $1.40, the participant made a profit. If the quality was below standard, the resale price to the experimenter was less than $1.40. If the quality was above standard, the resale price was more than $1.40. The buyer knew neither the quality of the commodity nor the resale price reflecting the true quality until he or she resold it to the experimenter. The seller could not control the quality of the commodity he or she sold; he or she could decide only on the price. Furthermore, the seller had only limited information about the quality of his or her commodity, and did not know exactly how good the commodity was. In other words, the seller was supposed to offer a price based on some vague idea of the quality of the commodity. Thus, the buyer could not judge whether overpricing was intentional or not. Actually, the computer program randomly determined the price so that the maximum profit for the buyer was $0.50 and the maximum loss was $0.10 in each transaction, whichever seller he or she bought from. There was no behavioral difference between seller A and seller C up to this point. However, there was a substantial difference between A and C with respect to whether or not the seller took advantage of an
92 “extortion chance” after a transaction. In this experiment, “extortion chances” were introduced to make the amount of social uncertainty clearly visible to participants. After each transaction, the computer gave the seller an “extortion chance” with a small probability. When the seller was given this opportunity, he or she decided whether or not to use it. If the seller took advantage of an “extortion chance,” he or she took a certain amount of money from the buyer to whom he or she sold the commodity. The amount of money that the seller could “extort” from the buyer increased as the experiment went on; the seller could extort $1.20 (120 yen in Japan) from a buyer during transaction periods 1-7, $1.90 during periods 8-14, and $2.50 during periods 15-20. The amount of money that the seller took from the buyer was supposed to be paid to the seller at the end of the experiment (although “the seller” was actually a computer), and the amount of money taken from the buyer actually was subtracted from the buyer’s total earnings. The difference between sellers A and C lay in their use of the “extortion chance.” The computer was programmed such that each of the two sellers would have the extortion chance twice during the first 20 transaction periods. Seller A took advantage of both opportunities and took away the amounts given above from the participant. In contrast, seller C never used the extortion chance. In other words, A was programmed to act like a selfish person who took advantage of any opportunities for self-benefit, while C was programmed to behave like a nice person who did not exploit other people even when opportunities were given. The participant as a buyer dealt with these two sellers for the first 20 transaction periods. After the 20th transaction period, replacement of transaction partners occurred, and seller F replaced seller A (the egoist). The transactions until this point, i.e., in the first 20 transaction periods, were just preparation for the experimental manipulation. The actual experiment started from the 21st transaction period. From the 21st period, a buyer had a new set of two sellers to deal with, seller C (the nice person) and seller F (a newcomer). From past experience, the buyer knew that seller C was a safe bet. In contrast, the seller had no prior experience with the new seller, F. The buyer had to decide with whom to complete a transaction. Transaction with C was safe, but F was programmed to offer a lower price than C. The independent variable, the level of social uncertainty, was manipulated by keeping or removing the “extortion chances” for transaction periods 21-30, the 10 periods during which the buyer dealt with sellers A and F. In the high uncertainty condition, not only did the “extortion chances” remain, but also the amount of money extorted increased to $4.00 (400 yen in Japan). In the low uncertainty condition, the “extortion chances” no longer existed after the 20th transaction period. The fact that the “extortion chances” disappeared after the 20th period was announced to the participant at the beginning of the 21st period when seller F replaced seller A.
Findings Results of this experiment are shown in Figure 5.1. There were 10 transaction periods after seller F replaced seller A. The graph shows how many times during the last 10 transaction periods the buyer bought from seller C, whom he or she learned to be a safe person from the experience of the first 20 periods. The new seller, F, was programmed to offer lower selling prices. Therefore, if the chance that F would behave selfishly was small, the buyer would be better off trading with F than with C. However, as this picture indicates, participants did not jump into trades with the new seller F who was offering lower prices. In the low uncertainty condition in which there was no “extortion chance,” participants traded with F in about half of the last
Figure 5.1 The number of transactions with safe seller C during the last 10 transaction periods 10
94 transaction periods. In the high uncertainty condition in which the buyer could suffer a big loss from the seller, the buyer concluded about twice as many transactions with C, who was safe but offered higher prices, as with F, whose niceness the buyer did not know. This difference between conditions was statistically significant. Furthermore, responses to one of the post-experimental questionnaire items, “how important was it to keep the same seller as a trade partner in order to develop a trusting relationship with him or her?” show that participants in the high uncertainty condition thought it more important to build a trusting relationship with the same seller than did those in the low uncertainty condition. These results support Hypothesis 1 that the greater is the level of social uncertainty, the stronger is the tendency for people to form commitment relations. Figure 5.1 also indicates that there is little difference in the degree of commitment formation (in this case, the frequency of keeping seller C as transaction partner) between Japanese participants and American participants. Although one might see a difference between American respondents and Japanese respondents in the high uncertainty condition, that difference is not statistically significant. Therefore, this result supports Hypothesis 2 that, once the levels of social uncertainty and general trust are set to be equal, there is no difference in the degree of commitment formation with C between Japanese and Americans.
Decline in Trust in Strangers Results of the first experiment suggest two things: (1) social uncertainty facilitates commitment formation and (2) when we control the experimental setting so that the level of social uncertainty, the level of opportunity cost, and the level of general trust among participants are the same, no difference emerges between Japanese and American participants in the degree of commitment formation. Besides these main findings, a few interesting findings emerged.* One is that the mere existence of social uncertainty reduces trust in “strangers” with whom one does not have a commitment relationship. Another is that the more committed one is to a specific partner, the less trust one places in “strangers.” Let us start with the first finding. This finding is based on the relationship between the level of social uncertainty and participants’ responses to a post-experimental questionnaire item that asked the degree of trust in seller F. The new seller F was not given an “extortion chance” in either condition. However, responses to a question that measured the degree of trust participants felt in F on a 5-point scale (1 is untrustworthy, 5 is trustworthy) revealed that participants in the low uncertainty condition rated F’s trustworthiness at 3.73 on average while the average response in the high uncertainty condition was 3.46. This difference is statistically significant. Because seller F was never given an “extortion chance,” *
See Kiyonari and Yamagishi (1996) for details.
95 participants would not have been able to tell whether F was a trustworthy person or not. This finding suggests that the mere existence of the possibility of being exploited in social interactions reduces the level of trust in “outsiders” with whom one has little or no familiarity. The next finding is that forming a commitment with a specific partner (in this case, seller C) reduces trust in “outsiders” with whom one does not have a commitment relation. This is based on the finding that the more committed the participant was to C during the first 20 transaction periods, the less trust he or she placed in F.* A negative correlation between the degree of commitment to C during the last 10 periods and trust in F may reflect the reverse causal relationship that participants who did not trust F formed a commitment relation with C. To avoid this anticipated criticism, however, I analyzed the relationship between the degree of commitment to C during the first 20 periods and the response to the post-experimental questionnaire item that measured trust in F. Therefore, it is safe to interpret the results as showing that commitment to C reduced trust in the “outsider,” F. Together, these results suggest that the mere existence of social uncertainty reduces trust in “outsiders,” those outside a person’s commitment relations, and that the more committed a person is to a specific partner, the more distrust he or she has in such “outsiders.”
Experiment 2 The first experiment tested one of the fundamental propositions of the emancipation theory of trust: social uncertainty facilitates commitment formation. Also, the results demonstrate that when both levels of general trust and levels of social uncertainty are set to be equal, there is no difference between Japanese and Americans in the degree of commitment formation. The primary purpose of the second experiment that I present next was to demonstrate that conclusions from the first experiment are not limited to the specific features of the particular experimental setting, such as transactions of commodities between sellers and buyers. I aimed to achieve this goal by showing that qualitatively similar results emerge in other experimental settings. Another important purpose was to test the hypothesis that we could not test in the first experiment because of problems in the logistics of the experiment. That hypothesis is that low trusters will be committed more strongly to a particular partner than will high trusters. This is directly related to the fifth proposition of the emancipation theory of trust.
A multiple regression analysis was used to derive this conclusion. I used society (Japan or US), social uncertainty (high or low), the level of general trust (high or low), and pre-commitment (the number of periods that one bought from seller C during the first 20 periods) as independent variables (the first three of which were dummy coded), and the number of periods that one bought from C during the last 10 periods as the dependent variable. The negative effect of pre-commitment was significant.
Hypothesis 3: Facing social uncertainty, the tendency to form a commitment relation with a specific partner is higher among low trusters than among high trusters. Furthermore, by expanding the previous argument, we have another hypothesis, that the relationship between general trust and the tendency to form commitment relations predicted by the third hypothesis will be the same between American participants and Japanese participants. Hypothesis 4: The effect of general trust on commitment formation predicted in Hypothesis 3 is the same for the two groups of participants: Americans and Japanese. The purposes of the second experiment are to test the two hypotheses presented above (Hypotheses 3 and 4) as well as to confirm, using a different experimental setting, the first two hypotheses (Hypotheses 1 and 2) that were supported in the first experiment.
Procedure The second experiment differed from both the first experiment and Kollock’s in two respects. The first difference is that this experiment involved an abstract setting that does not correspond to any particular real life situation, whereas Kollock’s and the first experiment used trading settings between buyers and sellers. In Kollock’s experiment on which the first experiment was based, social uncertainty was manipulated by allowing or not allowing a seller to lie about the quality of a commodity. The first experiment basically followed this transaction setting of Kollock’s experiment, but differed from Kollock’s in two respects. One was that only one participant was a real human and the other “participants” were computer responses, and the other is that the manipulation of social uncertainty was more extreme. In the second experiment, as I show below, the manipulation of social uncertainty was simpler and more abstract. Also, two, instead of one, real participants existed in each experimental group, although a computer played the other “participants” as in the first experiment. Thus, it was possible to measure commitment relations between real participants—another difference from the first experiment in which commitment relations were actually with the computer. As was the case with the first experiment, the second experiment was run in the Social Psychology Laboratory at Hokkaido University and the Social Exchange Laboratory at the University of Washington. There were 186 Japanese participants and 141 American participants. Based on level of general trust, they were divided into a high truster group and a low truster group before the experiment was run. In
97 each country, in both groups, participants were assigned to either the high uncertainty condition or the low uncertainty condition. Here, I do not describe the experimental laboratory in detail because it is basically the same as in the first experiment. Each experimental session consisted of repeated “trials” (which corresponds to “transaction periods” in the first experiment) that I describe below. Although there were 60 trials, participants were not told in advance how many trials there would be. As I explain later, each participant transacted with either a human or a computer in each trial. Social uncertainty was manipulated using “extortion chances” as in the first experiment. In each transaction with a human, each of the two participants received 10 cents as a reward for the trial. Then, one of the two was selected for an “extortion chance” to take 10 cents from the partner. The computer randomly determined who was given this chance. The participant who was given this chance decided whether or not to take 10 cents from the partner. If a participant decided to take 10 cents, he or she made 20 cents in that trial, and the partner made no profit at all. Note that this has been the description of transactions with a human in the low uncertainty condition. In the high uncertainty condition, the partner from whom 10 cents was extorted not only lost the 10 cents that the experimenter had awarded him or her at the beginning of the trial, but also suffered an additional loss of 50 cents. The participant who took 10 cents from the partner, on the other hand, only additionally gained those 10 cents—exactly the amount that he or she actually took from the partner. The extra loss of 50 cents was explained as similar to the cost of a broken car window caused by the theft of a car stereo. The consequence of the partner’s exploitative behavior (i.e., taking advantage of the extortion chance) was more serious in the high uncertainty condition than in the low uncertainty condition. That is, the loss from the partner behaving in an exploitative manner was 60 cents in the high uncertainty condition but only 10 cents in the low uncertainty condition. Thus, we can expect that the tendency to keep a relationship with a person who has not behaved in an exploitative manner—the tendency to form a commitment relation with that particular partner—will be stronger in the high uncertainty condition than in the low uncertainty condition. In the description above, I made no mention of opportunity cost. In the situation as described, participants would not lose anything by interacting solely with a nice partner. Thus, everyone surely would form a commitment relation with a nice partner even in the low uncertainty condition. In order to prevent this from happening, opportunity cost was introduced, as it was in the first experiment, such that leaving a commitment relation or not forming a commitment relation could yield extra profit. More specifically, when a commitment relation was dissolved, there was a possibility that a participant’s new partner would be a “generous” computer, as well as a possibility that the new partner would be another participant.
98 Thus, if the participant kept choosing the same partner, he or she would lose the opportunity to make a transaction with a “generous” computer. When the partner was a computer, the participant’s profit on that trial was determined by “roulette.” The “roulette” was displayed on the computer screen and determined the participant’s profit randomly within the range from 12 cents to 27 cents. The computer “generously” gave 12 - 27 cents, and, furthermore, did not take money from the participant. Whether a participant was matched with the computer or another participant on the first trial was determined randomly. The probability that a participant was matched with the computer was 50%. Because there were only two real participants in each group, this random decision was common to both of them. In other words, when a participant was matched with a human, the human was always the same person, the other real participant in the group. However, participants were led to believe that the group included a larger number of people, and thus they might be matched with a different person each time. From the second trial on, whether a participant dealt with the computer or with a human participant was determined in the following manner. When the participant had dealt with the computer in the previous trial, the partner was chosen as in the first trial; the partner was determined randomly, with a 50% chance of being a human and a 50% chance of being the computer. When the participant had dealt with a human on the previous trial, each participant was asked at the start of the new trial whether he or she wanted to deal with the same partner. If both participants chose to interact with their partners from the previous trial, that is, each other, the two in fact did deal with each other in the new trial. If at least one of the two declined to interact with his or her partner from the previous trial, then both were assigned partners as in the first trial. The partner was determined randomly, with a 50% chance of being a human and a 50% chance of being the computer. Thus, a participant who had completed a transaction with a human partner in the previous trial had two choices. One was to continue dealing with the same partner. If the partner also chose the participant as a partner in the new trial, they dealt with each other again. Then, if neither took advantage of the extortion chance—that is, if a mutually cooperative commitment was formed between the two—each participant received 10 cents. The other choice was to stop interacting with the human partner. Then, it was possible that he or she would receive 12 to 27 cents from the computer. At the same time, it was also possible that the participant would have to deal with a new human partner. In that case, the participant could not know whether or not he or she would suffer severe loss, because the new human partner might or might not take advantage of the extortion chance. If a participant threw away a commitment relation with a nice partner who had not taken advantage of an extortion chance, he or she had the possibility of receiving a better profit from the computer, while he or she also faced the possibility of
99 interacting with a new, potentially exploitative human partner. In this way, the design of this experiment made assurance-pursuing behavior, paying opportunity costs in order to reduce uncertainty, a strong feature of commitment
Low Uncertainty Condition High Uncertainty Condition
0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0
Figure 5.2 The mean Commitment formation index by the uncertainty condition and the participants’ level of
First, we should see how many times participants took advantage of the extortion chance. On average, they used about half (51%) of the extortion chances given to them. Participants tended to withhold exploitative behavior more in the high uncertainty condition where exploitative behavior imposed greater damage to
100 the partner than in the low uncertainty condition. Specifically, participants took advantage of the extortion chance 67% of the time in the low uncertainty condition, while they used it only half (33%) as often in the high uncertainty condition. This might have jeopardized the experimental manipulation since the probability of being exploited was smaller in the high uncertainty condition than in the low uncertainty condition. However, responses to a post-experimental questionnaire item that asked about the effectiveness of the experimental manipulation revealed that participants in the high uncertainty condition worried about being exploited much more strongly than did those in the low uncertainty condition. Therefore, the experimental manipulation worked; in the high uncertainty condition, uncertainty was higher rather than reduced. In this experiment, a repeat transaction between people who had made a deal in the previous trial was possible only when both of them declared their intention to continue transactions with the same partner. Thus, the probability that a pair in one trial was maintained in the next trial was used as the index of commitment formation. Figure 5.2 shows this index of commitment formation, or the proportion out of the total number of trials of trials in which participants continued their transaction with the same partner, broken down by the condition. We can see that the tendency to form and maintain a commitment relation was very different between the high uncertainty condition and the low uncertainty condition. This difference is statistically significant. This result shows that the tendency to form a commitment relation with a specific partner is stronger when the degree of social uncertainty is higher. As in the first experiment, Hypothesis 1 clearly was supported. Whether the proportion of transactions with the same partner was different between Japanese and American participants was examined next. Although Figure 5.3 seems to suggest that the commitment index is slightly higher among American participants than among Japanese participants, the difference is not statistically significant. Thus, the result supports Hypothesis 2 that when the degree of social uncertainty and level of general trust that are theoretically relevant to commitment formation are controlled, the tendency to form a commitment relation with a specific partner will not differ between American and Japanese participants. The two hypotheses that had been supported in the first experiment were supported again in the second experiment. How about Hypothesis 3? Was the tendency to form and maintain a commitment relation stronger among low trusters than among high trusters? Results show that the proportion of the transactions with the same partner was 0.27 among low trusters whereas it was 0.23 among high trusters, and that the difference is statistically significant. Therefore, this hypothesis was supported as well. Figure 5.3 shows that the difference between high trusters and low trusters in the tendency to form a commitment relation does not differ much between Japanese
101 and American participants. As Hypothesis 4 predicted, the effect of general trust on commitment tendency is almost identical for Japanese and American participants, and the difference is not statistically significant.
Japanese Americans 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0
Figure 5.3 The mean Commitment formation index by the participants’ nationality and their level of general trust
These results not only replicate the results of the first experiment, but also show that the tendency to form yakuza-type commitment relations when faced with social uncertainty is higher among low trusters than among high trusters. Overall, these results support almost all the propositions included in the emancipation theory of trust discussed in Chapter 4. Relationship with the Theory This may be a good opportunity to review the experimental findings particularly in relation to the emancipation theory of trust. The first proposition of the theory is that trust is meaningful only when social
102 uncertainty exists. This proposition does not require empirical investigation. Whether one trusts others or not is basically equivalent to evaluating the probability of being exploited. Therefore, it is self-evident that trust is not meaningful when there is no social uncertainty. I adopted this logically self-evident proposition as the first proposition of the theory because I wanted to make it very clear that this is the most fundamental assumption when we think of trust. Contrary to the first proposition, the second proposition, that in general people form yakuza-type commitment relations in order to deal with problems produced by social uncertainty, requires empirical investigation. It is logically true that when one forms a commitment relation with a specific partner, social uncertainty within this relationship is reduced. However, commitment formation is not the only way to reduce social uncertainty. If people facing social uncertainty provide each other with “hostages,” social uncertainty among them will be reduced significantly. Alternatively, laws and customs, and establishment of a central authority to administer laws are other ways of reducing social uncertainty. Acquiring power is still another effective way to reduce social uncertainty. Because there are many other ways to reduce social uncertainty, logically demonstrating that commitment formation reduces social uncertainty is not sufficient; accumulation of empirical research is necessary to prove that people actually and frequently form commitment relations to reduce social uncertainty. That is why we conducted Experiments 1 and 2. Results of these two experiments in addition to results of Kollock’s experiment all consistently support this proposition. Although each experiment manipulated social uncertainty differently and measured the degree of commitment formation differently, all three of these experiments clearly supported this proposition. Furthermore, this proposition corresponds to conclusions drawn from empirical and simulation studies on network-PD. Considering this overall consistency, we can conclude that this proposition’s empirical validity is reasonably high. As in the case of the first proposition, the third proposition, that commitment formation incurs opportunity cost, does not require empirical testing. The commitment relation itself is defined to be a relation that is maintained despite opportunity cost. Therefore, it is not necessary to call this a proposition. However, I decided to call it a proposition to make explicit the fact that opportunity cost is a critical aspect of commitment relations. The same logic applies to the fourth proposition, that leaving a commitment relation is more profitable than staying in it when opportunity cost for staying is high. When the opportunity cost of commitment is high so that it exceeds the benefit gained by the reduction of social uncertainty, it is clear to anyone who can do subtraction that leaving a commitment relation is more profitable than staying in it. The fifth proposition, that low trusters are more likely to form and maintain commitment relations than are high trusters, requires empirical support as in the case of the second proposition. High trusters are defined as people whose default
103 expectation for human benevolence is high, and low trusters as people who estimate human benevolence to be low. Then, intuitively it makes sense to suppose that high trusters and low trusters differ from each other in their tendency to seek (or avoid) new relations with people in which security is not assured. Furthermore, if we adopt the assumption that the only difference between high and low trusters is in the level of the default expectation of others’ trustworthiness, this proposition is almost logically true. However, the fact that this proposition intuitively makes sense is no substitute for proof. Furthermore, there is no guarantee that high and low trusters differ only in their default expectations of others’ trustworthiness. It is possible that default expectation of others’ trustworthiness is related to other factors and that those other factors have effects on commitment formation independent of the effect of the level of general trust. Logical analysis is not sufficient to solve this type of a problem. We have to conduct empirical studies to see whether people actually behave as this proposition tells us. As far as I know, Experiment 2 presented earlier in this chapter is the first and only study to test this proposition that has been conducted so far. The arguments of Fukuyama and Putnam that were introduced earlier are basically about this proposition. They argue that voluntary formation of groups or organizations beyond already existing commitment relations (especially family bonds) is difficult in low trust societies. However, their arguments have been examined only indirectly using macro-level statistical measures. Such evidence cannot prove this proposition. Results of Experiment 2 presented earlier in this chapter support the fifth proposition. However, support by only one experiment is not sufficient to prove the empirical validity of a proposition. It is possible that the proposition would not have been supported if we had conducted another experiment using different settings. Likewise, results might have been different if the participants had not been undergraduate students. I return to this kind of problem at the end of this chapter. For now, I simply note that research is moving in a direction that empirically supports the validity of the fifth proposition, acknowledging the need for further studies using different methods in the future. The sixth and last proposition, that high trusters can be better off than low trusters when both the level of social uncertainty and the level of opportunity cost are high, does not require an empirical proof. That is because this proposition is automatically produced by the combination of the fourth and the fifth propositions. That is, in a situation of high opportunity cost in which leaving commitment relations is more profitable than staying in them (Proposition 4), high trusters who are more likely to leave them (Proposition 5) should be better off than low trusters. As I argued above, the fourth proposition’s validity is guaranteed without empirical investigation. Therefore, if we can empirically demonstrate the validity of the fifth proposition, it automatically demonstrates the validity of the sixth proposition as well.
104 In sum, among the six propositions constituting the emancipation theory of trust, only Propositions 2 and 5 require empirical tests. Two experiments presented in this chapter produced the results that supported the validity of those two propositions. In particular, the second proposition was supported not only by our two experiments but also by Kollock’s experiment, and also is consistent with the results of network-PD research. I argue that the validity of this proposition can be generalized considerably. In contrast, the fifth proposition was tested only in Experiment 2. We should continue empirical investigations on this proposition using diverse settings.
Experiment 3 The purpose of the third experiment is considerably different from that of the two experiments presented above. It is rather close to the purpose of the US-Japan comparative questionnaire study that was described in the previous chapter. The purpose of this experiment is to show that Americans tend to trust others more than do Japanese, not through paper-and-pencil responses to questionnaire items but through actual behaviors of experimental participants. Both questionnaire studies and laboratory experiments are used commonly in the behavioral and social sciences, and they have different strengths and weaknesses. The first advantage of questionnaire studies over laboratory experiments is in the generalizability of results to the population through using a sample randomly drawn from the population. This is a very important point when we examine the validity of the argument about Japanese and Americans in general. In this regard, laboratory experiments have a big handicap. When we run an experiment, we cannot draw a representative sample randomly from the population of Japan or the U. S. and bring them to the laboratory. We cannot afford the travel cost, and few people would be willing to participate. Therefore, when we run an experiment, participants are people who live near the laboratory, and most of the time are students of the university where the laboratory exists. Needless to say, students of the university where the laboratory happens to exist do not constitute a representative sample of the Japanese or the American people. Thus one might think that laboratory experiments are meaningless. However, that is not necessarily so. Questionnaire studies have their own crucial limitation; information that researchers can obtain is limited to what comes to the conscious mind of the respondents. When the purpose of the research is to investigate what people are consciously aware of, of course that is not a big problem. There still may be a technical problem of how to get people to express their true feelings and ideas. However, this is only a technical problem, insofar as respondents themselves can be conscious of the true feelings and ideas that researchers want to know. However, when the purpose of research is to gain access to what people cannot be consciously aware of or what people do not think consciously in everyday life, the difficulty
105 increases tremendously. If we want to know what respondents themselves do not know, asking people through a questionnaire is meaningless. In this regard, people are consciously aware of trust, the object of our research, and can think about it consciously. On the other hand, there is no guarantee that people are sufficiently aware of it in everyday life. Therefore, even if the results of a survey show that Americans tend to trust others more than Japanese, what this really means is not so clear. The critical problem is that ordinary people do not think consciously about the various types of trust that this book talks about. Considering that even researchers on trust have not distinguished trust from assurance of security, it is no wonder that ordinary people do not distinguish between them. If this is the case, respondents might answer with assurance in mind, for example, in responding to a question that we try to ask about trust as distinct from assurance. In order to deal with this problem, we should not ask people directly. We have to think theoretically about how trust, distinct from assurance of security, affects people’s behaviors in certain situations and actually observe expected behavioral differences. One way to do this is to observe how people behave in everyday life. Although this is a valuable research method in itself, it is not suitable for observing theoretically expected behavioral differences in their pure form because so many factors affect behaviors in everyday life. For that purpose we need an experiment whose settings eliminate various external factors. Researchers on humans and societies are divided into two camps: the survey camp and the experimentation camp. Each criticizes the other’s weaknesses and is eager to deny results of the other camp’s research categorically as meaningless. However, considering that each research method has its own weaknesses and shortcomings, this is not a happy situation. In order to draw truly meaningful conclusions about humans or societies, we cannot rely on a single method. What is needed is to use various methods to investigate the same problem and examine whether they produce consistent results or not. If results using various methods are consistent, the conclusion will be persuasive. If results are not consistent, we think about the reason for the inconsistency, and this improves our understanding of the object of our research. Understanding the danger of relying solely on one methodology, I use experimental methodology as well to investigate the Japanese-American difference in trust.
Behavioral Measure of Trust There have been many attempts to examine trust by looking at behavior in experiments. Most of them treat cooperation in a Prisoner’s Dilemma situation as an indicator of trust in a partner. Although other types of experiment have tried to measure degree of trust through participants’ behavior, many of them have conceptual problems. For example, Cash, Stack & Lune (1975) used the behavior of falling backward into the experimenter’s arms to measure participants’ trust in the
106 experimenter. As I explained in Chapter 2, what is measured by this behavior is mainly expectation of ability—whether the experimenter really can hold the participant—and not trust in the experimenter’s intentions—whether the experimenter purposefully will let a participant fall to pursue his or her own interest. This is an example of the pitfalls into which research without good conceptualization may fall. Can we think of cooperation in a Prisoner’s Dilemma as an indicator of trust in a partner? There is also a big problem here. The reason is that cooperation in a Prisoner’s Dilemma is influenced by many factors other than trust in a partner. For example, when there is a continuing relationship between two players, each player can control the other player’s behavior by changing his or her own behavior in response to what the other player has done previously. However, when it is possible for a player to control the partner’s behavior, it is not necessary for him or her to predict the partner’s behavior. Thus, when players cooperate in the course of following a tit-for-tat strategy, they are not necessarily trusting the partner. Then, how about measuring trust in a partner through cooperation in a one-shot PD, not through an iterated PD? There is also a problem in this case. The problem in this case is that if the reason for cooperation is based on altruistic motivation or influenced by a norm (in this case, the sense of obligation that one must cooperate in such a situation), whether or not one expects the other’s cooperation is irrelevant. That is, whether their partners are trustworthy or not is irrelevant for altruistic or norm-observing players. When a person cooperates due to altruistic motivation or the norm of cooperation, there is no logical reason to believe that trust in a partner affects that person’s cooperative behavior. Thus, we must conclude that cooperating or not cooperating in a Prisoner’s Dilemma is insufficient or inappropriate as a behavioral measure of trust in a partner. Therefore, we have proposed a new experimental paradigm suitable for measuring trust in its pure form behaviorally. This is the “benevolent dictator game” that I will present next.
The Benevolent Dictator Game To measure trust behaviorally, we need to observe behavior that changes only as a reflection of trust in the partner and nothing else. As I argued earlier, cooperation-defection in a Prisoner’s Dilemma is not a desirable behavioral measure of trust because it may be affected by factors other than trust in the partner. Therefore, we invented a new experimental paradigm (Kiyonari & Yamagishi, 1999).*
This experiment is similar to the “game of trust” that has been studied by Dasgupta (1988), Snijders (1996), and others. However, there is an important difference in that the recipient’s choice does not affect the allocator’s profit at all.
107 In this experiment, participants play the role of allocator or recipient. The recipient’s task is to choose between two methods of receiving money. One method of receiving money is to receive $6 (600 yen in Japan) directly from the experimenter as a reward for participation in the experiment. The other method is to let the experimenter give $15 (1500 yen in Japan) to the allocator and then let the allocator allocate this $15 between him- or herself and the recipient. The allocator is free to allocate the $15 in any way he or she wants. The allocator can allocate $15 to herself and give nothing to the recipient. Or, if she wishes, she can give herself and the recipient an equal share of $7.50 each. If a recipient believes that his partner is a fair person and expects her to divide the $15 equally between them (i.e., $7.50 each), he will entrust the allocator with control of the $15. Conversely, if a recipient believes that the allocator is an egoistic person who does not care about fairness and expects her to take most of the $15, he will choose to receive $6 directly from the experimenter rather than letting the allocator control the $15. This experiment was conducted under complete anonymity in which nobody knew who was who’s partner and what the partner did. Furthermore, we created an experimental procedure such that even the experimenter did not know each participant’s decision. Thus, a recipient was free from the worry of possibly hurting the feelings of the allocator by choosing to receive $6 directly from the experimenter. An allocator was free from the worry of being regarded as a mean person for allocating little to the recipient. Furthermore, because the experimental procedure was designed such that even the experimenter did not know each participant’s behavior, participants did not have to worry about being considered a chicken or a mean person by the experimenter. Thus, whether a recipient chooses the sure $6 or lets the allocator divide $15 depends almost completely on whether or not he believes that the allocator is a fair person. Unlike in the Prisoner’s Dilemma, motivations such as altruism or cooperation are irrelevant. This is because, whichever behavior the recipient chooses, the allocator gets the amount he or she allocates to him- or herself. In addition, because anonymity is completely guaranteed even to the experimenter, concern for what the partner or the experimenter thinks of them will not affect participants’ choice of behavior. Thus, the behavior of the recipient in this experiment is the most appropriate behavior for measuring trust. Let us explain the experimental procedure in more details. Participants were 120 American and 96 Japanese college students. There were 6 persons in each experimental group. First, the six participants had a discussion for 20 minutes. The topic was what people should do when they face a situation such as: “A ship was wrecked and many people are drowning. There is an escape boat, but it is too small for all people to be on board.” The purpose of this discussion was to give participants opportunities to know what kind of people other participants were. After the group discussion, another experimenter appeared and asked the
108 participants to participate in another experiment. When all six people agreed to participate in another experiment (actually all participants agreed), a “new experiment” started. In this “new experiment,” participants first were given the explanation of the aforementioned benevolent dictator game. Then, they were told that assignment to the role of allocator or to the role of recipient would be determined by drawing lots at the conclusion of the experiment. Then, all participants were asked to decide how much of $15 they wanted to keep for themselves and how much they wanted to give to the recipient if they were assigned to be an allocator. After this, they were asked again to imagine that they had been assigned to be a recipient, and were asked whether they chose to receive the sure $6 or to let the allocator divide. In answering this question, the only information they were told about the allocator matched with them was that he or she was one of the other five participants; they were not told with which one of the five they were matched. Finally, they were asked to imagine a situation in which a particular one of the five had been assigned the role of their allocator, and were asked to decide whether to take the sure $6 or to let the allocator divide $15. They repeated this decision five times, once with each of the other participants as the designated allocator. These decisions were made in the following manner. Each participant put his or her ID number and the decision sheet into an envelope, sealed it, and gave it to an assistant. The assistant entered the room as the experimenter left it, and collected envelopes. He randomly matched three pairs and decided who was assigned the role of the allocator and who was assigned the role of the recipient by ID numbers. Each participant’s reward was determined by the actual decisions made in the randomly matched pair of allocator and recipient. The assistant performed this procedure in front of the participants so that participants saw that no deception was involved. However, he did not know each participant’s ID number. Since the assistant knew each participant’s decision by ID number but did not know each participant’s ID number, he could not relate faces to ID numbers. Participants literally saw that anonymity was maintained while watching that the assistant randomly matched pairs. Finally, the assistant put money into the envelopes based on the behavior of the two participants in the pair, wrote down ID numbers on the envelopes, and left the room. The experimenter entered the room again and gave the envelopes to the participants. Only money was inside each envelope. Thus, the experimenter who knew the ID numbers of the participants knew neither whether a given participant was an allocator or a recipient nor how much he or she received.
Findings As had been expected from the survey results, results of this experiment showed that American participants trusted the allocator more than Japanese
109 participants. First, when paired with an unspecified allocator, only 12 % of American participants chose to receive the sure $6, while 20% of Japanese participants chose to receive the sure $6. Second, Japanese participants wanted to accept allocations from 3.7 of the other five participants on average, while American participants wanted to accept allocations from 4.2 of the other five on average. These differences are statistically significant. Thus, results indicated that the tendency to take an action that requires trust in others is higher among American participants than among Japanese participants. Does this difference reflect the fact that Americans actually allocate more fairly than Japanese? Or, are only expectations of allocation different between Americans and Japanese while actually allocation behavior itself is the same? In other words, does the fact that Americans tend more than Japanese to take an action that requires trust in others mean that Americans are more trustworthy than Japanese, or that Americans are more trustful than Japanese? The relationship between trustfulness or willingness to trust and trustworthiness was discussed in Chapters 2 and 3. Most previous studies treat trustfulness or willingness to trust as a simple reflection of trustworthiness. In contrast, I argued that one’s trustfulness is more than a simple reflection of the trustworthiness of surrounding people. The comparison of actual allocation behavior and entrusting behavior among Americans and Japanese sheds light on this issue. How much did participants actually allocate to the recipient? American participants allocated $6.14 out of $15 to the recipient on average. Sixty-six percent of American participants allocated $7.5 (half of the $15) or more to the recipient. Japanese participants allocated $6.44 to the recipient on average, and sixty-one percent of them allocated 750 yen (half of the 1,500 yen) or more to the recipient. The average amount of allocation to the recipient indicates that on average Japanese participants allocated more than American participants, and that the proportion of fair allocators is higher among Americans than among Japanese. However, neither of these differences was statistically significant. Because the directions of the differences are not consistent between the average allocation amount and the proportion of fair allocators, and because neither difference was statistically significant, we can conclude that Americans and Japanese did not differ in trustworthiness in this experiment. This means that Americans and Japanese differ in trustfulness or willingness to trust, but not in trustworthiness. In other words, the tendency to behave fairly is not different between Americans and Japanese, but the expectation that other people will behave fairly is different. The main finding of this experiment—that willingness to trust others is higher among Americans than among Japanese—is consistent with results of the questionnaire study. As I stated earlier, we cannot make a persuasive argument about the US-Japan difference based only on results of an experiment that involves only a relatively small number of college students. There is absolutely no guarantee that
110 participants represent Japanese or Americans as a whole. What is important here is that the main result of this experiment is consistent with results of many previous questionnaire studies. Results of questionnaire surveys are more representative of the general population. However, there is no guarantee that what is measured by a questionnaire is truly what we want to measure. Americans may not have a higher level of trust in general others than do Japanese but, rather, simply may believe that they think that others are trustworthy. With an experiment in which we can observe actual behavior that requires trust, what we observe is not whether people believe that they think that others are trustworthy, but whether they actually believe that others are trustworthy. The fact that results of this experiment correspond to results of the questionnaire studies increases the validity of the results of the questionnaire studies. At the same time, it also confirms the validity of this newly invented benevolent dictator game.
The Role of Experimentation I am using experiments, including the three experiments presented in this chapter as well as several other experiments, to test the theory of trust. While reading about those experiments, readers who are intrigued by the results may yet be wondering to what extent we can generalize results obtained from a small number of unrepresentative participants? I have conducted probably more than 100 experiments, although I have not counted exactly how many, and have presented the results in various occasions and places. When the audience is psychologists or social psychologists, I am not asked this question very often. However, when the audience is social scientists such as sociologists, political scientists, or economists, or undergraduate students or the general public, almost certainly I will not be able to leave the room without being asked this question. In some extreme cases, my articles have been rejected by reviewers of academic journals precisely for the reason that results of laboratory experiments cannot be generalized to the “real” world that lies outside the laboratory. What I would like to call to your attention is that behind this question is a fundamental misunderstanding about the purpose of experiments in the social sciences. The misunderstanding is that the purpose of experiments is to observe participants’ behaviors in a miniature society that experimenters create in a laboratory and then apply the observed behaviors to the real world. Actually, there are experiments with this kind of purpose. For example, engineers may build a model of a plane and try to see how air flows around the plane in a wind tunnel. Although this example is not an example of a social science experiment, it is a typical example of an experiment that serves this kind of purpose. In the social sciences, particularly in experimental economics, this type of experiment is common. However, most experiments in the social sciences are not of this kind; they are not wind tunnel experiments. They are not experiments whose purpose is to generalize
111 results observed in the laboratory to the real world. Then, what is the purpose of those experiments? The purpose is to test and develop theories. This means two things with regard to the above question. First, results of this type of experiment are not meant to be generalized directly to the real world. Researchers have no intention of doing such a thing. What is important is not the results or the participants’ behaviors, but the theory tested by the behaviors. As an example, let us consider the second experiment presented in this chapter. The situation that participants in this experiment faced does not exist anywhere outside the laboratory that was designed for this experiment. We cannot find anywhere in the world relationships in which each person alternately exploits the other person by taking 10 cents from him or her. A place where we can generalize results of the second experiment directly does not exist anywhere on earth. However, we can apply the theory that this experiment tries to test everywhere in the world. This experiment tests the proposition that the tendency to form commitment relations becomes stronger as the level of social uncertainty increases. From the practices of trading raw rubber in Southeast Asia that Kollock discussed to the relations among yakuza members, situations to which this proposition can be applied are ubiquitous. While the commitment relations are produced by the same principle, nothing concrete about the relationship between an owner of a rubber plantation and a rubber broker can be applied directly to the relationship among yakuza members. In quite the same way that concrete behaviors in the second experiment cannot be generalized to any real life situation, concrete behaviors of rubber plantation owners cannot be generalized to yakuza members. It is only when we abstract what is going on in the rubber market and conceptualize it as commitment formation to reduce social uncertainty that we can apply what is going on in the rubber market to relations among yakuza members. What is important is not what exactly is going on in the rubber market, but that commitment formation in order to reduce social uncertainty is occurring. Following exactly the same logic, what is important in the second experiment is not whether or not that kind of extortion chance actually exists in the real world (such a situation does not exist!), but that participants form a commitment relation with a specific partner when the possibility of being exploited is high. The second important point here is that representativeness of the sample is not a necessary condition for the purpose of testing theories, although it is better to have it. This point requires more elaboration. The reason we conduct an experiment is that there is a possibility that a hypothesis derived from a theory will be rejected by the result. If the possibility of rejection is zero, there is no point in conducting the experiment. In this sense, the purpose of an experiment is not to prove that a theory is correct but to prove that a theory is wrong. To prove that a theory is wrong, or to show that a theory does not hold, participants in an experiment are not required to be a representative sample of a certain specific population. No matter how biased the
112 sample is, a result that does not match the theory implies that something must be wrong with the theory. For example, let us suppose that most people behave in the way a theory predicts, but that the same result is not obtained among a certain group of people. This situation means that the theory is incomplete, and that the theory has failed to specify a certain criterion that distinguishes one type of people to whom the theory applies from the other type of people to whom the theory does not apply. No matter how biased the sample is, we can tell at least that the theory is not complete if the result turns out to be different from what the theory predicts. Conversely, this also means that we should not consider that a theory is proved to be true just because results from one or two experiments support the theory. The most fruitful way to use experimention in the social sciences is to conduct multiple experiments using various samples and various settings, and to try to distinguish what is consistent across the results of multiple experiments and what is dependent on a particular sample and setting. If results differ from sample to sample and setting to setting, they give us precious clues for making our theory more detailed and valid. This is because, if results vary, it means that there are other factors that were not considered in the previous theory. If we can specify those factors, the theory becomes better and more valid. In this sense, experimental results are the whip that drives us to pursue further theoretical developments. We become more and more confident in the validity of the part of the theory that yields consistent results regardless of sample and setting. For example, the proposition that social uncertainty facilitates commitment formation was supported among both American students and Japanese students. Moreover, consistent results were obtained when the experimental setting was similar to the actual trade (Kollock’s experiment), when all of the actors except one real participant were simulated actors (the first experiment), and when the setting was more abstract so that there was no corresponding situation in the real world (the second experiment). On the basis of these consistent results, we can conclude that the proposition that social uncertainty facilitates commitment formation is a considerably general proposition. The above discussion may be summarized as follows. What experiments attempt to generalize to the real world is not the results themselves but the theories. Considering the logic of experimentation, there is no question about the fact that we cannot generalize results themselves beyond the laboratory. The logic of experimentation requires us to eliminate all possible influences other than the factors in which we are interested. Because real world phenomena are influenced simultaneously by many factors, the situation created in an experiment in which all sources of influence except the ones in which we are interested are excluded cannot exist in the real world. Paradoxically, an experiment whose results per se can be generalized to a real life situation is a failed experiment since the experimental setting of such an experiment must involve all potential sources of influence existing
113 in the real world, implying that the experimenter failed to exert sufficient experimental control. I hope that the reader has understood that the purpose of an experiment is not the generalization of results per se. When we conducted the second experiment, we were not trying to acquire general knowledge about how often people exploit a partner when an extortion chance is given alternately. Rather, the objective of an experiment is to generalize theory. To generalize a theory is to make it clear under what conditions a theory holds and under what conditions it does not hold. What is necessary to achieve this type of generalization is not to conduct an experiment using a representative sample but to conduct multiple experiments using various samples and various settings to test the same theory. The most efficient way to generalize a theory is to conduct a replication experiment that differs in samples and settings from the original one. This is, of course, beyond one researcher’s hands. It is often beyond one research team’s hands. It becomes possible only when a whole group of researchers tries to accomplish this goal as a group. That is why it is necessary to have cross-fertilization among researchers who criticize each other for the development of a science.
Chapter 6 Trust as Social Intelligence
In this chapter, first I reexamine the theory of trust discussed in the previous chapters from an evolutionary game theoretic perspective. In so doing, I show that the theory has a theoretical missing link. Finally, I present results of a series of experiments that show that this missing link can be filled with “social intelligence.”
Selection by Consequence According to the theory of trust presented in this book, people trust general others because doing so is adaptive, or is advantageous, in a certain type of social environment—i.e., an environment in which both social uncertainty and opportunity costs are high. Whether or not people have a high level of general trust is explained ultimately by the advantage it provides in a particular social environment. However, that does not necessarily mean that people use a high level of general trust consciously as a means of pursuing self-interest. It is no secret that people unconsciously acquire psychological traits that are beneficial to them. For example, suppose a girl meets a boy who always treats her tenderly. It would not be unusual for her to come to love him after a while. She would not make a conscious decision to love him in order to get the benefit she values, i.e., tender care, after deliberate considerations such as: “I will be able to secure tender care from him by falling in love with him.” Although such cases may exist, people do not fall in love as a result of a conscious choice. They fall in love involuntarily. And yet, it is often the case that one falls in love with a person who provides resources (including psychological resources, such as tenderness, love, and understanding, as distinct from physical resources) that one values. The mechanism that endows people with psychological traits or behavioral tendencies that serve their self-interest, without their consciously pursuing those traits or tendencies, is the mechanism of “selection by consequence,” broadly including learning and evolution. Your dog’s learning to put his leg on your hand is contingent on his receiving food when he puts his leg on your hand. Your dog learns to put his leg on your hand not because he thinks: “I’d better put a leg on my master’s hand to make him give me food.” Without making such a conscious calculation, your dog puts his leg on your hand insofar as the behavior brings a desirable consequence, i.e., food. The same can be said for evolution. Male peacocks have gorgeous feathers. However, this is not a consequence of their efforts to obtain the gorgeous feathers that make them attractive to peahens. They have gorgeous feathers because peacocks
115 with gorgeous feathers attracted many peahens and thus left many offspring who inherited the genes for wonderful feathers. When I argue that people come to trust others in general in an environment in which general trust brings beneficial outcomes, this principle of “selection by consequence” is implied. I do not argue that people come to trust others in order to pursue self-interest. People whose level of general trust is high will tend to believe, without giving it much thought, that others are trustworthy. However, if people are embedded in an environment in which trusting others does not bring outcomes that further their self-interest, they will not come to believe unwittingly that others are trustworthy. This is the logic used in the emancipation theory of trust.
Social Environment The emancipation theory of trust proposes that high levels of social uncertainty and opportunity costs, when they co-exist simultaneously, make general trust an advantageous psychological trait to acquire. It is very important here to note that this environment is the social environment, that is, the environment that is created and maintained by humans. When the principle of “selection by consequence” operates, which outcome is desirable usually is determined by constant characteristics of the environment. Insofar as the characteristics of the environment are unvarying (e.g., the master always gives food to his dog when the dog puts a leg on his hand), the same behavior (e.g., putting a leg on the master’s hand) is selected. When the social environment does the selection, however, the same behavior may or may not prove to be adaptive. This is because the social environment itself changes. An example is whether having a high level of general trust produces desirable outcomes. In a society in which the level of general trust is low, people will tend to limit their interaction partners to those who are known to be trustworthy and to form commitment relations with them. In such a society, the enhanced tendency to remain in commitment relations further lowers opportunity cost of remaining in current relations. When people interact only with partners to whom they are committed, those who exit commitment relations and seek out new relations will have difficulty finding new partners who will welcome them and be willing to interact with them. For example, in a job market characterized by the practice of lifetime employment, it is difficult for a person who quits a job in the middle of his or her career to find a new job. This is one example of how the prevalence of commitment relations reduces the opportunity cost of staying in current relations. In contrast, in a society in which the level of general trust is high, the tendency to stay in current relations is low, and thus the general level of opportunity cost is high in the society. Those who exit current relations can find new partners easily who will welcome and accept them. In sum, whether general trust operates as an advantageous or disadvantageous trait depends on the nature of the social environment, which in turn depends on what
116 kind of psychological traits people have. The social environment is distinct from the physical environment in that it endows a psychological trait with adaptive advantage or disadvantage while it itself is a product of the very psychological traits it promotes or discourages, and varies depending on their distribution.
The Game Approach Most social relations are intertwined such that one person’s behavior prompts another person’s response, which in turn prompts the first person’s or a third person’s response. Given such complexity, knowledge of a person’s psychological traits or personality is not sufficient for understanding why she engaged in a certain behavior. We need to know how her behavior affects people around her and how they affect her. Game theory is an analytical tool for analyzing what the final outcome of a process of mutual influence will be among a set of interdependent people. In game theory, the principle describing a person’s behavior usually is assumed to be stable. In most cases, the principle of rational choice is assumed; that is, people select the behavior that is expected to bring the most desirable outcome to them. Following Heath (1976), let us call this “forward-looking rationality.” It is “forward-looking” in the sense that a person predicts outcomes of his behaviors based on knowledge about the environment and the ways other people react, and then consciously chooses the behavior that brings him the most desirable outcome. This also may be called “conscious or deliberate rationality.” Game theory predicts how a group of people who affect each other, who are rational themselves and who know that the others are also rational, eventually behave. Although game theory usually assumes that the behavioral principle adopted by “players of the game” is forward-looking rationality, we can replace that principle by another principle and still use the same method to analyze situations. In this sense, forward-looking rationality does not constitute an indispensable component of the theory. Game theory can be applied to predict how a group of mutually interdependent people will behave when they individually follow a certain behavioral principle. The
Evolutionary Game Approach
A typical game theoretic analysis stops there. However, in evolutionary game theory, which emerged in a marriage between game theory and evolutionary theory, there is another phase. This is because evolutionary games include the feature that people’s traits or behavioral principles themselves are selected by outcomes. The emancipation theory of trust includes an element of evolutionary games in this sense. Before explaining this point, I describe evolutionary games in a more general sense. Let us take a closer look at the evolution of the peacock’s feathers that I briefly introduced earlier. The peacock’s gorgeous feathers are a trait that is
117 disadvantageous to their possessor’s survival. They make easier for a predator to detect it. When their possessor is found and chased by a predator, gorgeous feathers may prevent him from escaping because they may get caught in the brush. Thus, the evolution of the peacock’s feathers cannot be explained by their benefit for the survival of individuals. We cannot explain the peacock’s feathers by their survival value, as we can explain the lion’s teeth or the zebra’s speed. Then, how can we explain the evolution of the peacock’s feathers? The explanation is this: the more gorgeous a peacock’s feathers are, the more attractive he is to peahens. The same mechanism works in a modern human society. In Japan, everybody’s hair is naturally black, but some young men dye their hair blond. This is a disadvantage when they seek a job, but they do this because “blond hair looks cool” and attracts young women. However, why does a disadvantageous trait such as gorgeous feathers attract females? Just as middle aged men may not understand why young women are attracted to a young guy with blond hair who does not seem to get a good job, it is a puzzle that having gorgeous feathers is attractive when it is disadvantageous for survival. The key to solving this puzzle is that the environment that was important for the evolution of feathers was not a simple physical environment but a social environment. As in the case of the peacock’s plumage, a species acquires certain traits because genes responsible for those traits spread in the species. The peacock’s feathers evolve to be gorgeous because a peacock that has gorgeous feathers can produce more offspring than a peacock who has less gorgeous feathers. This is possible because a peacock with gorgeous feathers attracts more peahens. Why do peahens that are attracted to peacocks with gorgeous feathers leave more offspring? That occurs because there are other peahens that are attracted to peacocks with gorgeous feathers. Here, first let us consider a situation in which all peahens do not care at all about whether peacocks’ feathers are gorgeous or not. In this situation, if there is a mutant peahen that is attracted to peacocks with gorgeous feathers, it is unlikely that this peahen will produce more offspring than the other peahens. However, if there are other peahens that are attracted just a little bit to peacocks with gorgeous feathers, things will change dramatically. In this case, a peacock that is the offspring of the peahen that was attracted to peacocks with gorgeous feathers attracts peahens in the next generation by a small increment. Then, the number of grandchildren of the first peahen who produced the sons of the gorgeous feathered husband is a little bit greater than that of the other peahens. Assuming that half of the grandchildren are peahens, the proportion of peahens that are attracted to peacocks with gorgeous feathers increases just a little bit in the generation of the grandchildren. As more peahens are attracted to peacocks with gorgeous feathers, peacocks with gorgeous feathers become attractive to a larger number of peahens. In the generation of the grandchildren of the grandchildren, the proportion of peahens
118 that are attracted to gorgeous feathers will increase further. Through repetition of the same cycle for thousands of generations, the peahen trait of being attracted to gorgeous feathers spreads throughout the population. The first trigger may have been a small bias in preference. Evolution of the trait proceeds through mutual amplification such that peacocks’ feathers become more gorgeous and peahens become more attracted to gorgeous feathers. The first trigger was, perhaps, that brighter feathers showed a peacock’s good health. This trait, bright feathers, in itself is desirable to endow to offspring. As this original trait is amplified, it becomes a trait undesirable for the survival of individual peacocks. Nevertheless, this trait (gorgeous feathers) remains attractive to peahens. It remains attractive not because this characteristic itself is advantageous for survival but because other peahens are attracted to it. In Japan, for example, let us suppose that there is a mother who does not want to be obsessed with educating her children competitively and thinks that putting too much pressure on her children to study to pass the entrance exam is not desirable. Nevertheless, against her will she has to force her children to study hard because other mothers are obsessed with educating their children competitively. This mother is in the same position as that of a peahen who is attracted to gorgeous feathers. In both cases, if other mothers stop doing the “silly” behavior, this mother does not have to engage in the “silly” behavior. However, as long as other mothers’ behavior stays the same, this mother cannot change her behavior by herself. If she does so alone, her children will be left in a disadvantaged position. Thus, through such a mechanism, a trait that is undesirable for an individual (e.g., gorgeous feathers, studying too hard) can be reinforced while no one really wants it. In this example of the peacock’s feathers, the distribution of peahen preferences constitutes the social environment that selects the peacock’s trait (i.e., gorgeous feathers). Each individual peahen acquires another trait, preference for peacocks with gorgeous feathers, because there is an environment in which having this trait leads to having greater number of grandchildren. However, this environment itself is created by peahens as a group who are attracted to peacocks with gorgeous feathers. Here, possession by many individuals of a certain trait produces the social environment that makes this trait adaptive. It is important in this peacock example to note that the social environment is not given as fixed in the first place, but is created through the development of the peacock’s and peahen’s traits. At the same time, it is also important to note that selection of a trait is not done consciously through exercise of an individual’s forward-looking rationality.
In Search of a Missing Link I stated earlier that the emancipation theory of trust contains an element of the evolutionary game approach. The adoption of the evolutionary game approach there involves one hidden critical problem. That is, trust cannot be explained by genes. In
119 light of the fact that historical changes in levels of trust occur much more quickly than do changes in genes, we cannot attribute the difference between the U. S. and Japan in levels of trust to genetic differences between Americans and Japanese. Most evolutionary games assume that selection by consequence occurs through the selection of genes. If we cannot use genes as an explanatory tool, is it meaningful to use the principle of selection by consequence in the study of trust? There are two solutions that allow us to answer this question positively. The first is to adopt the principle of reinforcement in learning theory. A dog comes to put his front leg on his master’s hand if the master gives food to the dog only when he puts his leg on his master’s hand. This is the principle of reinforcement. More precisely, this is the principle of operant conditioning. In this case, as the case of evolution, a certain trait is selected by the outcome from having that trait. If putting his front leg on his master’s hand produces nothing, a dog does not come voluntarily to put his leg on his master’s hand. It is only when putting his front leg on his master’s hand produces positive outcomes that the dog comes to put his leg on his master’s hand. Researchers who use the evolutionary game approach to explain various human behavioral tendencies, such as Axelrod whom I introduced in Chapter 3, often think that we should use the principle of reinforcement to explain behavioral tendencies that cannot be explained by genes. The principle of reinforcement may explain in part the genesis of trust. However, I believe that the principle is not appropriate for the explanation of trust. This is because it takes a while for the positive outcome produced by general trust to emerge, and thus it is difficult to understand that having high trust produced that outcome. In general, reinforcement is not likely to occur unless there is a clear connection between behavior and reward. The reward from having general trust is an indirect one that is obtained as a result of utilizing outside opportunities through expanding relations. It is not easy to understand intuitively the connection between having general trust and the reward. Thus, it is difficult to explain the acquisition of general trust with the principle of reinforcement. Researchers who are trying to explain various behavioral tendencies using an evolutionary game approach sometimes also use the principle of imitation in combination with the principle of reinforcement. This principle states that if people see that a person succeeds in something by behaving in a certain way, they will copy that behavior. As a result of imitation, a behavior that brings a desirable outcome will be adopted by surrounding people. Although the principle of imitation may be suitable for explaining other behaviors, again it is difficult to use it to explain general trust. The reason is the same as in the case of the reinforcement principle. If the correspondence between general trust and the reward is unclear, it is difficult even for people who see high trusters being successful to speculate that the cause is general trust. Some insightful people might notice this correspondence. However, assuming such an elaborate
120 insight in many people means that we leave the principle of selection by consequence and return to the assumption of forward-looking rationality. To explain assurance mechanisms, as opposed to trust, it is sufficient to assume that humans have forward-looking rationality. Assurance of security is the expectation that my partner will not exploit me if he or she is a rational person. However, as I explained in Chapter 2, the trust on which this book focuses exists beyond rational expectations, and thus it is difficult to explain trust itself as the production of rational decision making. Now, we have seen that the first solution, the principles of reinforcement and imitation, at best is not sufficient. Let me turn to the second solution. The second solution is to introduce an auxiliary theory. The emancipation theory of trust specifies the conditions—high social uncertainty and opportunity costs—under which general trust is advantageous. However, neither the spread of genes, nor the principle of reinforcement, nor the principle of imitation can explain how the people who live in such an environment come to acquire general trust that is advantageous to have. [**Toshio: The original of the previous sentence was positive (no “neither” and “nor”s), but I assume it should be negative.] As long as we cannot explain the acquisition mechanism, we have to say that the theory is incomplete. This is the missing link that remains in the emancipation theory of trust. The second solution argues that to fill in this missing link we need to have an auxiliary theory focusing on this particular issue. In this chapter, I present a series of experiments we have conducted to find this missing link. While conducting this series of experiments, I came to realize that general trust may be acquired as a byproduct of an adaptive behavior that people consciously engage in based on forward-looking rationality. In other words, results of the series of experiments that I present in this chapter suggest that general trust can be explained indirectly, that is, as a byproduct of a consciously adaptive behavior. As I explained earlier, treating traits that cannot be explained by genes or reinforcement within the framework of evolutionary game theory requires additional theory that explains the process through which advantageous traits are acquired. Results of the series of experiments presented in this chapter suggest an “investment model of trust acquisition,” which I explain below as the needed auxiliary theory. According to this theory, general trust emerges as a byproduct of behavior, consciously adaptive to the environment, that makes general trust an advantageous trait to acquire.
Advantage of Being Credulous? I define general trust in this book as the default expectation of another person’s trustworthiness in the absence of information about that person. Trust in this sense, that is, trust as the tendency to overestimate a partner’s trustworthiness based on incomplete information, can be considered to constitute a bias in human information processing. If we consider trust a bias, we should conclude that having a high level
121 of general trust is adaptively disadvantageous. Accepting this argument means that, for example, a person who has had a super microcomputer implanted in his brain to boost his calculation power far beyond the limit of biological information processing capacity should have a smaller bias in his estimation of others’ trustworthiness than ordinary people. This superhuman will have hardly any trust. The above argument is consistent with the “rational choice” approach according to which trusting others beyond “assurance of security” is a “stupid” thing to do. In this argument, trust as a bias is judgments of others’ trustworthiness without any ground. Therefore, having a higher level of trust simply means being “credulous.” I have been countering this argument by saying that people trust others not because of limitations in our cognitive capacity to process information but because there are social environments in which having trust as a positive bias itself leads to a beneficial outcome. In other words, the emancipation theory of trust has emphasized “the advantage of being credulous.” The advantage is that being credulous facilitates forming relations with new people. However, results of the series of experiments that I describe below showed that this conception of “the advantage of being credulous” is wrong. I already mentioned some of these studies in Chapter 1 in reference to one of the paradoxes of trust: high trusters are more sensitive to information about others’ trustworthiness than low trusters, and high trusters predict others’ trustworthiness more accurately than low trusters. As I noted briefly in Chapter 1, numerous studies have shown that high trusters whose level of general trust is high are not necessarily more naïve and more gullible than low trusters. For example, Julian Rotter reviewed studies showing that high trusters are no more likely to be deceived by others than low trusters. He defined “gullibility” in his review articles as trust in the face of evidence showing lack of trustworthiness (Rotter, 1980a, 1980b).* Also, according to those studies, high trusters are no more likely than low trusters to be regarded by people around them as more gullible (Rotter 1967); the difference between high trusters and low trusters in degree of trust in the experimenter disappears when they are given the information that the experimenter lied to the participants in the previous experimental session (Geller, 1966); the difference between high trusters’ and low trusters’ trust in the partner disappears when they experience deception during the game (Hamsher, 1968; Wright, 1972); and there is no difference between high trusters and low trusters in judgments when they are asked to judge whether a person is guilty or innocent after reading the court record. These studies show that even high trusters come to distrust a partner when they are given information revealing that the partner is not trustworthy. In other words, even high trusters do not trust proven-to-be-untrustworthy people. Findings from those studies thus suggest that *
In these studies, participants’ level of trust was measured by the Interpersonal Trust Scale developed by Rotter (Rotter, 1967, 1971). The contents of this scale are similar to those of our trust scale. Several studies have shown that there is a high correlation between these two scales.
122 belief in human benevolence, that is, general trust as the default estimation of trustworthiness of others without information, on the one hand, and estimation of trustworthiness of specific people when information about them is available, on the other, are mutually independent. One study that Rotter cites (Lajoy, 1975) shows that high trusters are more likely than low trusters to use characteristics of the target person as clues when they are asked to estimate trustworthiness of a target person shown on a videotape. Another study examines the relationship between level of trust and interpersonal problems in everyday life, and shows that high trusters do not have more interpersonal problems than low trusters nor do they have more experience of being deceived than low trusters. Looking at these studies, it is clear that it is wrong to believe that high trusters are naïve and gullible. Results from our series of experiments went beyond those earlier experiments. They indicate that high trusters are more concerned about the trustworthiness of others than low trusters, and that high trusters distinguish trustworthy people from untrustworthy ones more accurately than low trusters. Although I already presented some of these results as one of the paradoxes of trust in Chapter 1, I present them below in more detail and include results not presented in Chapter 1. High Trusters Are More Sensitive to Information In Chapter 1, I presented two experiments showing that high trusters respond to information revealing other people’s lack of trustworthiness more sensitively than do low trusters. Here, let us briefly revisit the design of the experiments. In these experiments, participants read several scenarios and predicted in each scenario whether the target person would behave in a trustworthy manner or not. Results were that high trusters made higher estimates of the target person’s trustworthiness than low trusters when no information about the target person was available. However, when they were given information, especially negative information indicating that the target person was not trustworthy, high trusters changed their estimation of the target person’s trustworthiness more quickly than did low trusters. These results suggest that high trusters respond more sensitively to information potentially indicating others’ lack of trustworthiness. However, we should remember that participants in those experiments read hypothetical scenarios and estimated trustworthiness of hypothetical people, and that the outcome of the estimation had no consequence for the participants themselves. Thus, it is not so clear what we can learn from the responses to such hypothetical scenarios. It is true that high trusters responded to information more sensitively than did low trusters when they estimated trustworthiness of the target person in the scenario. However, low trusters may not have responded less sensitively to information if the outcome of their estimation had had real consequences for themselves. Low trusters may be more concerned than high trusters with their own benefit, and if so, they may behave differently when their own benefits depend on estimation of others’ trustworthiness.
123 Therefore, as I suggested in Chapter 1, to conclude that high trusters respond to information suggesting lack of others’ trustworthiness more sensitively than do low trusters, we need to create a situation in which participants’ rewards directly depend on the estimation of others’ trustworthiness and see how participants estimate others’ trustworthiness and behave based on this estimation. Although the experiment that I show you below was conducted for a different purpose, in the process of analysis we obtained results that show the difference between high trusters and low trusters relevant to this issue. The experimental setting was such that participants’ reward depended on the trustworthiness of the partner. High trusters in this experiment were shown to respond more sensitively than low trusters to information suggesting the partner’s trustworthiness or untrustworthiness and to behave accordingly.
Prisoner’s Dilemma with Variable Dependence I conducted this experiment with one of my graduate students, Riki Kakiuchi (Kakiuchi & Yamagishi, 1997). There were 80 Japanese participants. They were led individually to a small cubicle without seeing or talking to other participants and played a Prisoner’s Dilemma game repeatedly. More specifically, participants played a Prisoner’s Dilemma game 48 times with the same partner using the payoff matrix shown in Figure 6.1. However, participants were not informed in advance how many times they would play the Prisoner’s Dilemma game.
Partner’s Choice Ａ 10yen
Ａ 10yen -30yen
Figure 6.1: The payoff matrix used in Kakiuchi & Yamagishi’s (1997) experiment
124 The payoff matrix shows that each participant has two choices, A or B. A is the cooperative choice, while B is the defecting choice. An important feature of this payoff matrix is that one’s own reward is negative regardless of one’s own choice when the partner chooses B (i.e., defection). Since we could not force participants to pay money if the cumulative payment became negative, we gave them 500 yen before we started the experimental session. When the payoff was negative, we subtracted it from the initial endowment of 500 yen. The basic experimental setting was the Prisoner’s Dilemma game shown in Figure 6.1 repeatedly played by two participants located in separate cubicles. The following two conditions were superimposed on the basic setting. In one of the conditions (i.e., the constant payoff condition) participants repeatedly played the Prisoner’s Dilemma game as described above. In the other condition (i.e., the variable payoff condition), participants were able to change the payoff matrix. Although the original purpose of the experiment was to compare these two conditions, here I discuss only the latter, variable payoff condition, because it was only in this condition that level of general trust was relevant to sensitivity to information. In the variable payoff condition also, participants played a Prisoner’s Dilemma game repeatedly as in the constant payoff condition. The variable payoff condition differed from the ordinary Prisoner’s Dilemma in that participants could change the size of their own payoffs in the matrix shown in Figure 6.1. Suppose a player starts from the payoff matrix in Figure 6.1. When a participant chooses to increase, the payoff values in all cells for that participant are increased by 10 percent of the value in the initial matrix. This means that the absolute values of payoffs in cells that have a negative value also increase. In the initial payoff matrix shown in Figure 6.1, a participant receives 10 yen when mutual cooperation is achieved (AA), -10 yen when mutual defection occurs (BB), -30 yen when the participant is exploited by the partner (AB), and 30 yen when the participant exploits the partner (BA). If a participant decides to increase payoffs, in the next trial this participant receives 11 yen in the AA cell, -11 yen in BB, -33 yen in AB, and 33 yen in BA. If this participant decides to increase payoffs again, in the next trial he or she receives 12 yen in AA, -12 yen in BB, -36 yen in AB, and 36 yen in BA. If this participant continues to increase payoffs in 10 consecutive trials, in the 11th trial he or she receives 20 yen in AA, -20 yen in BB, -60 yen in AB, and 60 yen in BA. If a participant chooses to increase payoffs after every trial, in the 48th and final trial he or she receives 57 yen in AA, -57 yen in BB, -171 yen in AB, and 171 yen in BA. Participants did not have to increase payoffs every time. They could maintain or reduce payoffs. When participants reduced payoffs, payoffs changed in the reverse direction, toward zero. After every trial, participants decided whether to increase, maintain, or reduce payoffs. Therefore, they could keep choosing to increase for a while and then choose to reduce again. Please note, however, that each participant could change his or her own payoffs only. The partner’s payoffs could be
125 changed only by the partner. Likewise, the participant’s own payoffs could not be changed by the partner. Now, let us think about the implications of changing one’s own payoffs. Increasing one’s own payoffs means that the profit that one gets when the partner chooses cooperation increases, and, at the same time, the loss that one incurs when the partner chooses defection increases as well. Altogether, the impact of the partner’s behavior on one’s outcome increases. For example, in the initial payoff matrix shown in Figure 6.1, whether a participant’s partner chooses cooperation or defection makes a difference of 40 yen to the participant. (When the participant chooses A, he or she gets 10 yen if the partner chooses A, while getting -30 yen if the partner chooses B. The difference of 40 yen is produced by the partner’s behavior. Likewise, when the participant chooses B, he or she gets 30 yen if the partner chooses A, while getting -10 yen if the partner chooses B, resulting, again, in a difference of 40 yen.) On the other hand, if a participant increases his or her own payoffs one step, the influence of the partner’s behavior on his or her own outcome increases from 40 yen to 44 yen. This means that the degree of dependence of the participant’s outcome on the partner’s behavior increases from 40 yen to 44 yen. Increasing one’s own payoffs in this experiment means increasing dependence on one’s partner, that is, increasing the extent to which one places his or her fate (one’s own outcome) in the hands of one’s partner. It is easy to see that this choice is directly related to trust in the partner. If one thinks that the partner will not cooperate, one should reduce the size of one’s own payoffs. If one thinks that the partner will cooperate, one should increase the size of one’s own payoffs. Whether one increases or decreases the size of one’s own payoffs should be directly related to one’s prediction about the partner’s behavior. Results of the Prisoner’s Dilemma Experiment with Variable Dependence Although various interesting results were obtained in this experiment, here I limit my discussion to a comparison between high trusters’ and low trusters’ sensitivity to information. Toward this objective, let us first see how high trusters and low trusters responded to the partner’s behavior. During the first 16 periods, both high trusters and low trusters increased their payoffs similarly when the partner cooperated in the previous period. Figure 6.2 shows results of the first 16 trials out of 48 total trials. I calculated the mean change in payoff size by assigning +1 when the participant increased own payoffs, -1 when he or she reduced them and 0 when he or she did not change own payoffs. When the partner cooperated in the previous period, the mean change was 0.71 among high trusters while it was 0.68 among low trusters; there was little difference between them. However, after the partner defected, there was a difference between high trusters and low trusters. High trusters reduced their payoffs (the mean was -0.17) while low trusters still increased their payoffs (the mean is 0.32). This difference
126 between high and low trusters is statistically significant.
High Trusters Low Trusters
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3
After Partner’s Cooperation
After Partner’s Defection
Change in payoff size among high trusters and low trusters
An increase in one’s own payoffs means a greater loss when faced with a defecting partner. Nevertheless, during the first 16 trials low trusters on average continued to increase their payoffs even after the partner defected. In contrast, high trusters increased their payoffs following the partner’s cooperation and reduced them following the partner’s defection. Low trusters kept increasing dependency on the partner despite the partner’s untrustworthy behavior, whereas high trusters adjusted their dependency on the partner depending on the behavior of the partner. This means that low trusters failed to respond to the partner’s behavior in an appropriate way, while high trusters did, implying that high trusters responded to information revealing the partner’s trustworthiness more sensitively than did low trusters. The above result concerns the first 16 periods, or the first third of the total of 48 periods. Interestingly, during the next third of the trials (i.e., the 17th trial to the 32nd trial), low trusters became too sensitive and too cautious about the partner’s behavior. As I explained above, initially low trusters kept increasing payoffs regardless of the partner’s behavior. As a consequence, they suffered painful loss. Then, during the next third of the trials, they came to not increase their payoffs strongly even after the
127 partner had cooperated. Specifically, low trusters’ average change in payoff size after the partner’s cooperation was 0.44, while high trusters’ average was 0.82; low trusters hesitated to increase payoff size even after the partner had cooperated. After the partner had defected, low trusters’ average was -0.27, while high trusters’ average was -0.04. There was not a big difference, then. During the last third of the trials (from the 33rd trial to the 48th trial), neither after the partner had cooperated nor after the partner had defected was there much difference between low trusters and high trusters. In sum, low trusters initially were insensitive to information revealing the partner’s untrustworthiness, but after they incurred painful loss they became oversensitive and hesitated to increase the payoff size even after the partner’s cooperation. Low trusters came to regard others unconditionally as untrustworthy (i.e., they came to have a low level of trust) because they were insensitive to information revealing the partner’s untrustworthiness. This directly supports the “investment in cognitive resources model” that I propose at the end of this chapter. On the other hand, high trusters responded appropriately to information revealing the partner’s trustworthiness or lack of it. They did not increase payoffs when the partner had defected, and increased them when the partner had cooperated. Results of this experiment, in which trust in the partner directly affected the choice of the participant, thus show that low trusters and high trusters behave in a manner consistent with results of the scenario experiment presented earlier. So far, we have seen that high trusters are more sensitive than low trusters to information suggesting others’ lack of trustworthiness. This was observed both in the scenario experiment and in the experiment in which participant’s rewards were directly affected by their evaluation of the partner’s trustworthiness. We cannot, of course, draw this conclusion with complete confidence from results of these experiments alone. However, considering the fact that the same conclusion has been obtained from many experiments using different procedures, we can have some confidence in the conclusion. Further taking into account the fact that this conclusion is consistent with results of experiments presented below, that high trusters can predict the partner’s trustworthiness more accurately than low trusters, the persuasiveness of this conclusion becomes even stronger.
The First Detection Experiment I next present experiments, results of which show that high trusters detect the partner’s trustworthiness more accurately than low trusters. I already discussed one of the experiments in Chapter 1. Let us start with a brief review of this experiment. Participants in this experiment first engaged in a 30-minute discussion of garbage collection issues in 6-person groups. Then, each participant chose whether to cooperate or defect in Prisoner’s Dilemma games with some of the other five participants. Finally, each participant judged whether each of the two partners with
128 whom he or she actually was paired cooperated or defected. There were two experimental conditions, and different types of information could be used for prediction of the partner’s behavior in the two conditions. In the partner-known condition, each participant played a two-person, one-shot Prisoner’s Dilemma game with each of two specific partners, knowing who his or her partners were. The Prisoner’s Dilemma game was constructed in the following manner. Each participant was asked to give 100 yen to the partner or to take 100 yen from the partner. In the partner-unknown condition, when each participant decided whether to give or take 100 yen, he or she did not know who the partners would be. After all 6 participants had decided whether to give or take without knowing who the partners would be, they were told who were their actual partners. Although I did not explain this in detail earlier, the two conditions differed in what information was relevant to predicting a partner’s behavior. In the partner-known condition, each participant predicted how each partner would behave toward him or herself. In this condition, participants knew that the partner knew that he or she was playing with them. The partner’s behavior that the participant predicted was the partner’s behavior directed toward the participant. In contrast, in the partner-unknown condition, participants predicted how the partner would behave toward an unspecified target rather than how the partner would behave toward them specifically. In predicting the partner’s behavior, whether or not the participant could use information about how the partner felt about him or her differed between the two conditions. Now, let me remind you of the distinction between “character-based trust” and “relational trust” discussed in Chapter 2. Character-based trust is based on one’s belief about the other person’s integrity in general. The expectation that “because he is a good man, he would do no harm to other people (including me)” corresponds to character-based trust. On the other hand, relational trust is based on one’s belief about the kind of attitudes and feelings the other person has toward one. The expectation that “because he likes me, he would do no harm to me (although I don’t know how he would behave toward other people),” corresponds to relational trust. What is theoretically different between the partner-known condition and the partner-unknown condition is which kind of trust, relational or character-based, plays the important role. In the partner-known condition, the partner’s behavior should depend on how the partner thinks of the participant. Thus, relational trust based on belief about how the partner feels about the participant should be important in predicting the partner’s behavior toward the participant. On the other hand, in the partner-unknown condition, the participant has to predict how the partner will behave toward unspecified others. Then, information about how the partner feels about the participant is useless. In this condition, the information that is meaningful for the prediction of the partner’s behavior is information about how the partner will behave toward unknown others. Here, character-based trust based on estimation of
129 the partner’s integrity in general is the key. I included these two conditions in the experiment because the kind of trust that operates is expected to be different between the two. Results showed that accuracy in predicting the partner’s behavior was associated with participants’ level of general trust only in the partner-unknown condition in which character-based trust plays the important role. In the partner-known condition in which relational trust plays the important role, accuracy in predicting the partner’s behavior was not associated with participants’ level of general trust. High trusters were more accurate in estimating the partner’s behavior than either low trusters or medium trusters only in the partner-unknown condition; in the partner-known condition, there was no statistically significant association between accuracy in estimating the partner’s behavior and participants’ level of general trust. These results suggest that high trusters are more sensitive than low trusters to information suggesting the partner’s character in general rather than to information suggesting the kind of attitudes or feelings the partner has toward them. In the partner-known condition, using Japanese singer/songwriter Miyuki Nakajima’s verse, “a notorious rascal or a hatchet man, I care only if he loves me tenderly or not.” Whether one can trust his personality or not (whether he is a notorious rascal or not, or whether he is a hatchet man or not) is not very important. Thus, the result that general trust is meaningful only in the partner-unknown condition means that high trusters are sensitive to information revealing others’ trustworthiness in general, but that they are not particularly sensitive to information about how others feel toward them specifically. I discuss why this is so and what it means theoretically after I present another experiment.
The Second “Detection” Experiment In the previous experiments, we asked participants their estimation of the target person’s trustworthiness in the scenario, and we investigated how participants estimate trustworthiness of other participants whom they have not seen before. The experiment that I discuss next differs from the previous ones in that it investigates estimation of trustworthiness among people who have known each other for some time. Specifically, participants were sophomores who belonged to the same academic program (a smaller unit within an academic department). They had spent about a year in the program and knew each other fairly well. Two groups of participants were involved, one consisting of 19 sophomores who belonged to academic program A and the other consisting of 14 sophomores who belonged to academic program B at Hokkaido University. Almost everyone in each academic program participated in the experiment. Except that participants were acquaintances, the experimental setting was almost identical to that of the partner-unknown condition in the first detection
130 experiment. First, participants were told that they would play a Prisoner’s Dilemma game with one randomly selected classmate without knowing who the partner would be. Then, they were asked to decide whether to cooperate or defect. Finally, they judged whether each of the other classmates in the same academic program had chosen to cooperate or defect. In addition, they answered a post-experimental questionnaire including items related to general trust and other psychological traits. Results of this experiment were surprising, even to us. First, we calculated the accuracy of each person’s judgments of other participants’ choices in the Prisoner’s Dilemma.* Overall accuracy of judgment was 0.48. That is, overall, participants’ judgment of their classmates’ behavior was only as good as random guessing. From this result alone, we may conclude that participants could not predict the behavior of their classmates. However, when we divide participants into three categories based on level of general trust (i.e., low trusters, medium trusters, and high trusters) as we did in Kikuchi et al.’s experiment presented in Chapter 1 (Figure 1.2), we can see that the relationship between general trust and accuracy of judgment is similar to what appeared in Figure 1.2 in Chapter 1 (see Figure 6.3). As we saw in the case of Kikuchi et al.’s experiment, high trusters predicted their classmates’ behavior more accurately than low or medium trusters. In addition, when we treated level of general trust as a continuous variable without dividing it into three categories, the correlation between accuracy of judgment and general trust score was fairly high, r = 0.48. This correlation coefficient is statistically significant. The correlation coefficient did not change much when we controlled for participant’s program and the number of people whom the participant judged to be cooperative (r = .41). In addition to the relationship between accuracy of judgment of trustworthiness and general trust, relationships between accuracy of judgment of trustworthiness and other psychological traits were examined in this experiment. Particularly noteworthy is the strong correlation between the Honesty/Fairness Scale and the accuracy score (r = 0.43). For the items included in this scale, please see Table 4.5 in Chapter 4. As I explained in Chapter 4, this scale measures the respondent’s belief in the personal significance of being honest and fair to others. The higher the score is, the more important the respondent considers being honest and fair to others to be. I discussed in Chapter 4 the Japanese-American difference on this scale. Americans believe more strongly than Japanese that being honest and fair is important. *
In this experiment, the overall cooperation rate was 0.47, that is, close to 0.5. However, the two groups differed considerably in the cooperation rate and the level of general trust. The cooperation rate was 0.38 and the average level of general trust was 3.02 in Program A, while they were 0.53 and 3.64, respectively, in Program B. Then, we had to take care of the following potential problem. Suppose high trusters expected that other people would be cooperative. Then, high trusters would have a high accuracy score in Program B where a majority actually had cooperated even when high trusters had predicted indiscriminately that everyone would be cooperative. In order to avoid this problem, accuracy of judgment was calculated as the unweighted average of the proportions of correct judgments for the actual cooperators and for the actual defectors.
Accuracy of judgment
Accuracy of prediction by low-, medium, and high-trusters
In addition to the Honesty/Fairness Scale, the Sense of Interdependence Scale was also related to accuracy of judgment. This scale was developed recently by one of my former graduate students, Nobuhito Jin (Jin, 1997). It measures the belief that establishing and maintaining mutually cooperative relations is in one’s own self-interest. Specifically, it consists of four items. “Society is created by mutual dependency between people.” “In order to be successful in society, mutual cooperation is necessary.” “Being kind to others eventually will help oneself.” “Those who care only about their own benefit eventually will lose out.” Results showed a high correlation between the score on this scale and the accuracy score (r
132 = 0.55). Finally, the proportion of a participant’s classmates who judged the participant to have cooperated was correlated with the participant’s accuracy score (r=.32), although statistically the correlation was only marginally significant. This correlation means that those who were considered by classmates to be cooperative and trustworthy were more accurate in judging classmates’ behavior. These results suggest the following. Those who believe in human benevolence and who trust others in general, those who consider being honest and fair to others important, those who believe that cooperation is a gainful strategy, and those who are considered by people around to be trustworthy, are more accurate in detecting the trustworthiness of people around them than are their counterparts—those who consider everyone a thief, those who believe that honesty and fairness are meaningless, those who believe that being kind to others brings no good, and those who are considered by people around them to be untrustworthy.
Social Intelligence and General Trust The results of the series of experiments presented above consistently show that high trusters are actually not naïve, gullible or credulous. This implies that “the advantage of being credulous” assumed in the emancipation theory of trust is in fact wrong. The experimental findings presented above imply that high trusters are not only less gullible but also more clever than low trusters. What I mean by “clever” here is not that this person is smart in the sense that he is good at mathematics or having a high GPA. What I mean is that this person is clever in the sense that he is good at understanding the minds and characters of other people. Since the inception of scientific research on intelligence, many people have pointed out that there are various types of intelligence. In fact, one aspect of intelligence research is a history of debates over what kinds of intelligence exist and whether it is meaningful to think about general intelligence. Social intelligence has its own place in this history of intelligence research. The argument that social intelligence should be distinguished from other types of intelligence has been made since the beginning of intelligence research. For example, Thorndike argued that social intelligence or the ability to understand self and others and successfully use such understandings in interacting with others, is distinct from either abstract intelligence as the ability to manipulate language or numbers, or concrete or practical intelligence as the ability to handle things by a physical motion. According to Thorndike’s definition, social intelligence is the ability to understand own and other people’s internal state and use that understanding in social relations. Then, due to Cronbach and others’ criticism that social intelligence had never been defined precisely nor had been measured properly even after 50 years of research, research on social intelligence stagnated for some time. Recently, however, social intelligence again has begun to attract the attention of many researchers. For example, Gardner (1983) proposed a multiple intelligences
133 theory that argues that intelligence is divided into linguistic, logical-mathematical, spatial, musical, bodily-kinesthetic, and personal intelligence. According to this theory, which type of intelligence is considered important and thus is called simply “intelligence” depends on what is necessary for adaptation within each culture. Gardner’s “personal intelligence” is the ability to understand oneself and others, and thus it can be considered social intelligence in the context of this chapter. Sternberg (1986) furthermore argues that neither the intelligence as academic problem solving skill that is measured by IQ tests, nor creative intelligence to handle new problems adaptively, but intelligence as skill at treating oneself and others successfully within everyday social relations is needed under certain environments. Emotional intelligence proposed by Goleman (1995)—ability to control emotion, to understand other people, and to make social relations smooth—may be considered social intelligence in a broad sense. Here, I draw the reader’s attention to the possibility that the findings from the series of experiments presented in this chapter—that high trusters are more sensitive than low trusters to information suggesting untrustworthiness and can judge trustworthiness more accurately—may be interpreted as an indication of high trusters being more clever or socially more intelligent than low trusters. Of course, being sensitive to information suggesting untrustworthiness, or being able to use such information to discern trustworthy from untrustworthy people successfully is only a part of social intelligence as a more general ability to manage social relations successfully in order to achieve one’s own goal. However, at least it would be an important part of social intelligence. Accepting this view that high trusters are socially more intelligent than low trusters, a question arises immediately. Why is it so? Why do high trusters have higher social intelligence? Or, conversely, why do socially intelligent people come to have a higher level of general trust? This relation between social intelligence and general trust can be analyzed in terms of three potential causal mechanisms. The first possibility is that social intelligence increases as a consequence of acquiring general trust. However, it is implausible that having a belief that people are generally benevolent would increase one’s social intelligence. If there is any chance that this holds, the only conceivable process is as follows. Those who overestimate the trustworthiness of others about whom they have no information are likely to be taken advantage of by others. For them to avoid such disasters, they need to improve their social intelligence. In other words, social intelligence is not needed by low trusters, but it is highly needed by high trusters. Therefore, high trusters voluntarily “invest” cognitive resources in the acquisition of social intelligence, such as always paying attention to other people’s trustworthiness. As a result, compared to low trusters high trusters become more sensitive to information suggesting others’ trustworthiness, and they develop the ability to detect signs of trustworthiness. Furthermore, because high trusters are more likely to venture
134 voluntarily into social relations in which they can be deceived or exploited, they have more opportunities to develop the ability that is needed in such situations. The second possibility is in exactly the opposite causal direction—i.e., those who have high social intelligence come to acquire general trust. Those who are sensitive to information suggesting others’ trustworthiness (especially the lack of it), and have the ability to detect others’ trustworthiness, will not suffer serious damage even if they simply assume that all people are trustworthy (i.e., they have high default expectations of the trustworthiness of others). In contrast, socially unintelligent people are more likely to suffer severe loss when they enter into social relations assuming that people around are trustworthy. The best way for socially unintelligent people to avoid such suffering is to presume that strangers are untrustworthy and to avoid contact with strangers as long as possible. For a social genius who can read other people’s minds perfectly (although such a person does not exist in reality), social uncertainty does not exist. Thus, there is no need to form yakuza-type commitment relations in order to reduce social uncertainty. In contrast, a social idiot who is insensitive to other people’s minds faces a high level of social uncertainty once he or she gets out of commitment relations in which security is assured by the nature of social relations. For such a social idiot, it is a wise strategy to simply assume that everyone is untrustworthy, or using a Japanese saying, to assume that “everyone is a thief,” except when he or she is in commitment relations in which benign behavior from partners is assured. A social idiot would be advised to keep away from novel relations, and would do so by keeping the level of general trust or default expectation of others’ trustworthiness low. As a consequence, socially unintelligent people acquire a low level of general trust. This is the second possibility. The third possibility is that both general trust and social intelligence are produced by the same source. Previously, I argued that general trust is promoted in an environment characterized by high social uncertainty and opportunity costs. If the same environment also promotes social intelligence, high trusters naturally would be socially intelligent. To examine this possibility, we have to investigate whether an environment in which both social uncertainty and opportunity costs are high promotes development of social intelligence. An environment in which both social uncertainty and opportunity costs are high is essentially the environment in which “I want to eat fugu (globefish), but I don’t want to die.”* Here, delicious globefish represents opportunities outside commitment relations, and dying represents suffering in disastrous social encounters. If one is able to tell whether a globefish filet served at the table is poisonous or not by just looking at it, one can enjoy a delicious meal without losing one’s life. The *
Translator’s footnote. Although globefish is a delicacy many Japanese love to eat, it contains a deadly poison. Every year in Japan several people die from eating inappropriately cooked globefish.
135 same thing can be said about social intelligence. If one develops social intelligence and learns to detect trustworthiness in other people with whom one would like to have relations, one can enjoy the fruit of great opportunities outside one’s commitment relations without being deceived or exploited. In such a social environment, developing social intelligence yields great benefit. Let us compare this environment with another environment in which at least either the level of social uncertainty or the level of opportunity costs is low. An environment in which social uncertainty is low corresponds to the case in which somehow it is guaranteed that the globefish filet served at the table has no poison. In this case, a diner does not need the ability to identify a poisonous filet from its appearance. In the same way, nurturing social intelligence to distinguish trustworthy from untrustworthy people yields no benefit in an environment in which benign behavior from other people somehow is guaranteed. An environment in which opportunity costs are low while social uncertainty is high is like the situation in which the diner dislikes globefish. Since he does not eat globefish, whether or not he has the ability to detect a poisonous filet does not make a difference. In the same way, in an environment in which there are no greater opportunities outside one’s commitment relations, and thus there is no need to seek out new relations by leaving one’s commitment relations, the ability to detect trustworthiness in other people is a white elephant. This discussion leads us to the conclusion that an environment in which social intelligence to detect trustworthiness in other people serves one’s self-interest is also one in which a high level of general trust serves one’s self-interest. It is an environment in which both social uncertainty and opportunity costs are high.
The Cognitive Investment Model of Trust Development The three possibilities discussed above can be integrated into the point that the ability to detect trustworthiness in others is needed only in an environment in which both social uncertainty and opportunity costs are high. As the emancipation theory of trust argues, when social uncertainty and opportunity costs are both relatively high, one can expect, as a consequence of leaving current commitment relations in search of new relations, to get a benefit greater than what is obtainable in the current relations. At the same time, such exploration invites exploitation from new potential partners. In such a social environment, detecting trustworthiness in new potential partners directly affects one’s own benefit. Let me introduce one assumption here. That is, the ability to respond sensitively to information suggesting others’ lack of trustworthiness and the ability to detect others’ lack of trustworthiness can be acquired through training at least to some extent. There is no direct evidence that such abilities, revealed in the experiments presented in this chapter, can be acquired through training. However, often it is considered that it is possible to acquire emotional intelligence, including
136 sensitivity to the internal state and emotions of oneself and others, to a certain extent by training. Let me thus assume, for the present, that people can acquire the ability to respond sensitively to information suggesting others’ lack of trustworthiness and the ability to detect others’ lack of trustworthiness at least to some extent by training. While to many people the word “intelligence” means a stable trait that cannot be changed by training, I assume here that social intelligence can be trained. According to the above assumption, training is needed to acquire social intelligence as we need training to acquire other abilities. Social intelligence, needed to discern other people’s trustworthiness, will not come naturally unless one has paid attention to various features of surrounding people, such as their gestures, facial expressions, speech, and dress, since early childhood. Let me call the allocation of cognitive resources such as attention to features of surrounding people “investment of cognitive resources.” I adopt such an economic term here because it makes it intuitively clear that mobilization of cognitive resources incurs cost, and that such a costly investment should be based on the expectation of some return. Specifically, by investment of cognitive resources I mean behavior such as being cautious and paying attention to information suggesting others’ lack of trustworthiness. From this perspective, it is in an environment in which detecting other’s trustworthiness yields great profit that such investments of cognitive resources are facilitated. The previous argument about the relationship between general trust and social intelligence then can be reformulated in the following way. In an environment in which both social uncertainty and opportunity costs are high, investment of cognitive resources to collect information regarding other people’s trustworthiness is more likely to take place, and as a result of the investment social intelligence needed to detect others’ trustworthiness is more likely to develop. Investment of cognitive resources to acquire social intelligence needed to detect others’ trustworthiness is more likely to occur in an environment in which both social uncertainty and opportunity costs are high and thus the investment potentially can yield greater return. This argument is consistent with the first possible causal mechanism examined earlier. The first possibility explaining the relationship between general trust and the ability to detect others’ trustworthiness is that high trusters become more socially intelligent because those who trust others need to have social intelligence. This explanation is consistent with the argument here in the sense that they both argue that those who feel the need for social intelligence invest cognitive resources and acquire the necessary ability, assuming that social intelligence can be acquired by training. Furthermore, if we regard high trusters as those who actively think about leaving commitment relations, the first explanation is basically the same as the investment argument presented above. That is, cultivating social intelligence for detecting others’ trustworthiness is a good investment toward securing large profits for those who have a high level of general trust and are willing to leave commitment relations in search of better alternatives;
137 thus, high trusters who want to leave commitment relations will invest cognitive resources in the cultivation of social intelligence. This argument that investment of cognitive resources takes place to cultivate social intelligence in an environment in which both social uncertainty and opportunity costs are high does not contradict the second possible explanation, either. The second explanation was that socially unintelligent people who cannot discern the trustworthiness of others will have a lower level of general trust because by assuming that everyone is a thief they will avoid socially risky situations and thus will be better off. This explanation does not contradict the argument that cognitive investment to develop social intelligence is promoted in an environment in which both social uncertainty and opportunity costs are high. Finally, this argument is naturally consistent with the third explanation because it emerged from the third explanation.
The Missing Link Revisited Now, we can see that the argument that investment of cognitive resources in the development of social intelligence is promoted in an environment in which both social uncertainty and opportunity costs are high does not contradict any of the three explanations for why high trusters have higher social intelligence. However, one problem must not be forgotten. That is the problem of the missing link that I discussed earlier in this chapter. That is, it is true that an environment in which both social uncertainty and opportunity costs are high is a favorable environment for high trusters. However, unless each person’s level of general trust is determined genetically, we cannot specify the process through which people actually acquire general trust. Because we assume that general trust is a trait difficult to control consciously, we cannot explain the acquisition of general trust by the conscious investment of cognitive resources as we have discussed in the case of social intelligence. This means that the third possible explanation of the relationship between general trust and social intelligence is actually insufficient. The first account that starts with different levels of general trust as a given is insufficient unless we can explain why different people have different levels of general trust in the first place. Then, the only remaining possibility is the second explanation—people with high social intelligence can afford to have a high level of general trust, whereas those with low social intelligence cannot. Reversing the second account in fact provides the missing link, i.e., the mechanism by which general trust is produced in an environment in which both social uncertainty and opportunity costs are high. So far, I have approached trust from an evolutionary game perspective and have specified the conditions under which general trust serves one’s self-interest, but so far I have failed to specify the mechanism by which general trust is promoted in such an environment. That is what
138 is meant by the missing link in the emancipation theory of trust. The same approach can explain acquisition of social intelligence to discern others’ trustworthiness. It is like acquiring the ability to detect poisonous globefish filet for people who “want to eat globefish, but don’t want to die.” It is rational for them to acquire that skill if at all possible. Anyone can see this point. Then, those who live in an environment in which social intelligence for detecting others’ trustworthiness is essential surely will try to acquire such social intelligence. That is, they will invest the cognitive resources necessary for the cultivation of social intelligence. This explains why social intelligence is promoted in an environment full of social uncertainty and opportunity costs. According to the second account summarized above, socially unintelligent people simply assume that everyone is a thief, whereas socially intelligent people expect as a default that others are not so bad. This means that general trust develops as a byproduct of social intelligence cultivated through the investment of cognitive resources, which is adaptive in an environment full of social uncertainty and opportunity costs. Cultivation of social intelligence as a conscious adaptive behavior thus provides the missing link in explaining how general trust is promoted in an environment high in opportunity costs and social uncertainty.
Is Trust Needed? Now we have found the missing link in the emancipation theory of trust. The missing link is provided by social intelligence, and general trust can be treated as a byproduct of acquiring social intelligence. Then, a new question comes up: is trust really needed for the emancipation theory of trust? Isn’t it a redundant concept? Isn’t it social intelligence, not general trust, that is needed for emancipating people from yakuza-type commitment relations? Isn’t social intelligence, or the confidence based on social intelligence that one can deal with social risks, sufficient to get people to leave the security of commitment relations to seek out new opportunities? In fact, those who have acquired social intelligence will be confident that they can avoid the worst scenario even if they trust others until proven otherwise. This confidence is essentially the same as general trust. If general trust develops as a byproduct of social intelligence, general trust is nothing but confidence based on social intelligence that one can deal with social relations successfully. From this perspective, the arguments in this book can be stated as either arguments about “trust” or arguments about “social intelligence.” Whichever label we apply, the arguments will be the same. If we adopt this view, we may need to change the definition of general trust. The definition of general trust as the default expectation of others’ trustworthiness may be a definition based on the “phenotype” of general trust, which does not represent the true nature of general trust. From the “investment of cognitive resources model of trust development,” the true nature of general trust may lie in
139 awareness of the ability to deal with risky social relations. However, at this point in the research we are not ready to draw any conclusion about this issue, and we do not need to do so, because the research has not been completed yet. The core message of this book that collectivist society destroys trust stands as it is using either definition. I leave this question open until through future research we come to understand the implications of the new definition of trust as social intelligence.
Chapter 7 In Search of a Foundation for an Open Society
As I wrote in the beginning, this book is written around the central message that collectivist societies produce security and destroy trust. In addition to this explicit message, this book has another, implicit message. The implicit message of this book is that we cannot think about characteristics of human minds without considering their relations to the social environment. Unlike the explicit message, this message is not the focus of discussion in this book. However, this constitutes the “meta theory” for my theory of trust, the main theme of this book. The understanding that humans are social beings, and that we cannot think about characteristics of human minds without considering their relations to the social environment, is certainly not an original idea of mine. It would be no exaggeration to say that social psychology, my specialty, itself started from this understanding. However, while social psychology was being developed as a sub-discipline of psychology, the importance of this understanding was forgotten. Currently, it is no more than a “sutra” that any textbook of social psychology recites in the introduction. While many researchers work on the problem of how human minds are affected by the social environment, interest in the reverse problem of how human minds affect the social environment has disappeared from social psychology. I have adopted an evolutionary game approach as the “meta theory” that guides my research endeavor because I believe that this is the only approach that provides us with a framework for analyzing the dynamic effects of human minds on the social environment and the effects of the social environment on human minds simultaneously. In this chapter, I bring the implicit message constituting the backbone of this book to the surface. Then, I explain the relationship between the explicit message and the implicit message or the “meta theory.” At the end, I discuss how my messages including both the explicit and the implicit are related to problems that we face in the contemporary world.
Evolutionary Game and Co-evolution As I explained earlier, the implicit message of this book is that we cannot analyze human minds without considering their relations to the social environment. As I explained in Chapter 6, the social environment differs from the physical environment in that it is formed and maintained by humans’ who have minds with specific characteristics. In Chapter 6, I explained that the evolutionary game is a useful framework for analyzing dynamic interdependency between characteristics of
141 human minds and characteristics of the social environment. An evolutionary approach or evolutionary game approach was used in a very broad sense there: the goal of this approach is to uncover the “adaptive value” that a particular characteristic has in a particular environment. When we try to apply an evolutionary approach in this sense to the social sciences, probably the most important key concept is “co-evolution.” I believe that this concept of co-evolution is central to understanding society and culture from an evolutionary point of view. In short, co-evolution means the principle that a certain trait that does not have an adaptive value by itself acquires an adaptive value when it co-exists with other characteristics. In biology, the concept of co-evolution is used specifically for situations in which combinations of characteristics across different species generate adaptive values. However, in this book, I use it in a much broader sense. Thus, I use co-evolution not only to describe the combination of traits across species, but also the combination of traits across individuals or within individuals. In the natural world, we see many examples of symbiosis. They are typical examples of co-evolution. For example, a cleaner fish eats parasites in the mouth of a host fish. In this case, the cleaner fish’s disposition to enter the mouth of the host fish is disadvantageous for its survival unless the host fish has a disposition not to eat the cleaner fish. The disposition to enter the mouth of a host fish and the disposition not to eat easy prey are by themselves adaptively disadvantageous. However, when the host fish has the disposition to allow cleaner fish to clean its mouth freely, the cleaner fish’s disposition acquires an adaptive advantage. The disposition of the cleaner fish to enter the mouth of the host fish to eat parasites, on the one hand, and the disposition of the host fish to allow the cleaner fish to clean its mouth freely, on the other, each are advantageous for their respective fish’s adaptation as long as the other fish has the appropriate disposition. Such co-evolution, through which one trait acquires an adaptive value due to the existence of another trait, occurs not only between different species but also between individuals of the same species. The peacock’s plumage that I introduced previously is a typical example. The peacock’s gorgeous feathers, by themselves, have no adaptive value, or even would be disadvantageous for survival. They are conspicuous and prevent the owner from moving quickly, and thus make their owner easy prey for predators. In contrast, the same gorgeous feathers acquire an adaptive advantage when peahens have a disposition to prefer peacocks with gorgeous feathers. Please note that adaptation in this case concerns “inclusive fitness,” the term used in biology to refer to the spread of the relevant genes, rather than individual survival. As I mentioned above, co-evolution in the sense that a trait that has no adaptive advantage by itself acquires an adaptive value when another trait exists occurs not only between different species but also between individuals in the same species or within an individual. Although biologists do not use the term co-evolution for those
142 cases, I use the same term, co-evolution, even in those cases because I want to emphasize the commonality between biologically defined co-evolution and “co-evolution” in those cases. That is, each trait, of no use by itself, acquires adaptive advantage when it is combined with a set of other traits. The structure and organization of organisms including humans evolved as a result of co-evolution in this sense. For example, an eye functions effectively only when all of its various parts exist simultaneously. Further, no matter how intricately various parts of an eye are created, it has no adaptive value unless appropriate nervous circuits exist in the brain to process visual information. Thus, when we think about human traits, we should not examine each trait separately. We should think about the significance of each trait in relation to other traits.
Trust and Social Intelligence Now, let us consider the adaptive value of trust, as we have discussed it in this book, from a co-evolution point of view. In previous discussions, I argued for the possibility that having general trust increases adaptive fitness for each individual in an environment in which both social uncertainty and opportunity costs are high. However, that possibility is only a possibility. In a high-uncertainty-high-opportunity-cost environment, it is possible that an individual who leaves yakuza-type commitment relations will enjoy fruitful opportunities. At the same time, however, the possibility of being exploited and abused is also high for those who have left commitment relations. Therefore, in general we cannot determine whether having general trust increases or decreases an individual’s adaptive fitness. This potential difficulty faced by the emancipation theory of trust was addressed in Chapter 6 when I introduced the idea of investment of cognitive resources for nurturing social intelligence. There, I argued that those who have a higher level of social intelligence and are not exploited easily in a high-uncertainty situation can afford to maintain a higher level of general trust, while those who have a lower level of social intelligence cannot. In fact, this argument presented in Chapter 6 is an argument that two psychological traits, social intelligence and general trust, co-evolve. First, for those who have higher social intelligence, the risk of being exploited once they leave yakuza-type commitment relations is relatively low. Therefore, having general trust tends to work to increase their adaptive fitness. On the contrary, for those who have lower social intelligence, potential gain from utilizing opportunities outside commitment relations will be outweighed by the potential loss from being exploited. The same relationship exists in the other direction. Certain cognitive costs are incurred in developing social intelligence. Thus, paying the costs of developing social intelligence is a waste for those who stay in the security of commitment relations and do not face social uncertainty. For low trusters who do not
143 leave the security of yakuza-type commitment relations, the possibility of being abused by others is low and thus investment of cognitive resources to nurture social intelligence will produce no dividend. This is in sharp contrast to the case of high trusters who are likely to leave commitment relations and thus are likely to face social predators. Investing cognitive resources in order to acquire social intelligence that will keep them from being the prey of predators will yield a high dividend for high trusters..
Trust, Trustworthiness, and Social Intelligence Looking at trust from a co-evolutionary point of view, general trust may co-evolve not only with social intelligence, but also with trustworthiness. In the second detection experiment, presented in Chapter 6, those who detected others’ trustworthiness correctly were shown to be the ones who had a high level of general trust. Furthermore, their own level of trustworthiness (Honesty/Fairness Scale score) also was high. Assuming for now that the findings have a certain validity, although they need to be confirmed repeatedly in future research before being accepted as facts, they suggest that social intelligence co-evolves with trustworthiness. An association between trust and trustworthiness (that those who have a higher level of general trust tend to be trustworthy as well) has been observed in numerous studies including the above experiment, suggesting the possibility that general trust co-evolves with trustworthiness. The empirically observed correlation suggests this possibility. At the same time, it may be derived theoretically by assuming a high-uncertainty-opportunity environment. Now, we should consider two possibilities in a high-uncertainty-opportunity environment: that trustworthiness enhances the adaptive value of trust, and that trust enhances the adaptive value of trustworthiness. Let us start with the first possibility. In a high-uncertainty-opportunity environment, leaving commitment relations in search of outside opportunities is advantageous as long as one can keep the likelihood of being exploited low. However, this is only one side of the coin. We have to think about the other side of the coin as well. That is, those who leave commitment relations to seek new opportunities need to be chosen by new partners. At the same time, being chosen by new partners is gainful only for those who have left commitment relations in search of new partners. Those who stay in the security of current commitment relations do not need to be chosen by new partners. This means that being chosen by new partners has adaptive value only for high trusters who are actively seeking new partners outside their current relations. Trustworthiness, that is, a disposition to behave in a trustworthy manner, can be the most important trait for which potential partners from outside the commitment relation look. Thus, acquiring a disposition to behave in a trustworthy manner improves one’s chances of being chosen by new partners. (Of course, even one who has not paid the costs of acquiring that disposition can pretend to be a trustworthy
144 person. This book does not go into the details of this signal detection problem. I recommend Frank, 1988, to readers who are interested in this issue.) Now, we have seen that acquiring trustworthiness helps one’s chance of being chosen as a new partner in an environment involving social uncertainty. From earlier discussions, we can infer that the adaptive advantage of having a trustworthy disposition exists only for high trusters who actively seek out new relations. Acquiring trustworthiness will not increase the adaptive fitness of low trusters who do not want to leave commitment relations. In short, being trustworthy assumes an adaptive value only for high trusters. This relationship can be stated from the opposite perspective—trust has an adaptive value only for trustworthy people. This is because being willing to leave a current relation in search of new partners is of no use unless one is likely to be chosen by a new partner. Untrustworthy people, that is, those who exploit others whenever it is possible to do so, will not be chosen by anybody who knows that about them. Leaving commitment relations and sacrificing the benefit accruing there produces no benefit for untrustworthy people because they cannot form new relations. Therefore, the disposition to trust others and to seek new opportunities outside current commitment relations is a disadvantageous psychological trait. In contrast, for a trustworthy person whom everyone wants as a partner, acquiring general trust, being willing to leave commitment relations, and being willing to seek out new relations is an advantageous trait. Thus, we have seen that the relationship between trust and trustworthiness is one in which one trait enhances the other trait’s adaptive value. And yet, as I explained earlier, we should note that the possibility that high trusters will earn large benefits outside commitment relations is the other side of the risk of being abused outside committed relations. Then, in order for general trust to have an adaptive value, one needs to acquire social intelligence as well as trustworthiness. In a high-uncertainty-opportunity environment, acquiring an honest and fair disposition worthy of the trust of others, believing in the benevolence of others in general, and developing social intelligence to detect others’ trustworthiness work to one’s own advantage, as long as those three traits exist simultaneously as a set, not in isolation.
Equilibrium within Individuals and between Individuals To put it in a slightly different way, having each trait furthers self-interest more than not having it as long as this set of traits exists as a set. We can say of this situation that an “equilibrium” exists among traits within each individual. In game theory, when the current choice produces the most desirable outcome in the current situation for all players, it is called the Nash equilibrium. When an equilibrium exists, nobody who pursues self-interest wants to change her choice voluntarily since it produces the best outcome for her in the given situation. However, this does necessarily mean that the equilibrium combination of players’ choices produces the most desirable outcome for each player. For example, in a Prisoner’s Dilemma, a
145 situation in which two “rational” players both choose defection is in equilibrium. If one player unilaterally changes her choice from defection to cooperation, she ends up with a less desirable outcome. The same can be said about the relationship among traits within individuals. Let us think about a situation in which certain traits are advantageous for one’s own self-interest when they exist as a set, but not when one of them is missing. As an example, let us suppose that general trust and social intelligence mutually enhance the other’s adaptive value as I discussed earlier. Then, there is no “incentive” in this environment to induce one to change one trait while the other exists. In this sense, given a high-uncertainty-opportunity environment, the three traits—general trust, social intelligence, and trustworthiness—form an equilibrium of traits, not of behavioral choices. One point requires attention here: an equilibrium of behavioral choices is produced by players’ conscious decisions, while the equilibrium of traits does not assume players’ conscious decisions. The other point to which we should pay attention is that the equilibrium exists not between players but between traits within a player. From my perspective, the biggest difference between an evolutionary game and an ordinary game, although the difference is not widely discussed in the literature, is that an ordinary game involves relations among players whereas an evolutionary game involves a multi-level structure so that it can handle relations among traits within players as well as relations among players. In an evolutionary game, as in the case of the “uni-level game” that is assumed in ordinary game theory, each player’s payoffs are determined by the combinations of players’ behavioral choices. The choice of behaviors by a player in an evolutionary game is determined by the player’s trait. So far, nothing is new if we interpret a player’s trait to refer to the strategy typically used in ordinary game theory. Where an evolutionary game differs from an ordinary uni-level game is that the trait (= strategy) itself is selected. In an ordinary uni-level game, each player’s behavior is determined by her strategy. The possibility that the strategy itself is selected or changes based on a certain principle is not considered. In contrast, in an evolutionary game, the trait (= strategy) itself can change in response to the outcome of players’ behavioral choices. For example, whether a high-trust player who is willing to leave commitment relations earns a high payoff or not depends on whether she can form new relations with more desirable partners or not. Moreover, whether she can form more desirable relations or not depends on whether other players stay confined to commitment relations or not. Leaving a commitment relation when other players are confined to commitment relations is like quitting a job in a labor market in which all companies have adopted the lifetime employment system. Leaving current commitment relations only produces an undesirable result. Under such conditions, it is unlikely that general trust, promoting exit from commitment relations, will develop, and it is unlikely that people will invest cognitive resources in the development of social intelligence.
146 One important point to be noted here is that one feature of recent research on experimental games is a choice of partners (with whom to interact) in addition to a choice of behaviors (to cooperate or defect). Following Orbell & Dawes (1991, 1993), Hayashi (1995), and Hayashi & Yamagishi (1997), let us call ordinary game research that deals with a relation between designated partners the “forced play paradigm,” and new research that also deals with a choice of partners the “selective play paradigm.” In an ordinary, forced play game, the social environment consists only of the distribution of behavioral choices (i.e., cooperation or defection) by the player. In contrast, in the selective play paradigm the social environment also includes the distribution of relations among players. The extent to which people form commitment relations with specific partners is an important aspect of the social environment in a selective play situation. In this book we have seen that this aspect of the social environment determines the set of psychological traits of people who live in that environment, and that the set of the psychological traits that people acquire plays an important role in producing the social environment. For example, people who live in a social environment in which strong yakuza-type commitment relations are formed will acquire a set of psychological traits that promote in-group favoring behavior. Conversely, the more people engage in in-group favoring behavior, the stronger yakuza-type commitment relations will become. This implies that we can think of an equilibrium between traits at two different levels of an evolutionary game (at the level of the social environment where the trait consists of the distribution of players’ behaviors, and at the level of the player where the trait consists of the player’s trait), in the sense that one reinforces the other. When I argue that collectivist society destroys trust, as the explicit message of this book, I see the core of collectivist society in the equilibrium between the closed nature of social relations, on the one hand, and psychological traits promoting in-group favoring behavior, on the other.
Assurance of security and Network Extension Now, let us move back to the explicit message again. Humans probably have spent more than 99.9 percent of the human history of evolution over millions of years in a collectivist society. One of the most established findings in social psychology is that people have a tendency to prefer others with whom they interact frequently. Considering the generality and robustness of this finding, we may want to make a wild claim that this is a genetically embodied trait that evolved through human history in order to make humans treat in-group members in a preferential manner. What first induced humans to leave such commitment relations was probably opportunity costs, especially opportunity costs with regard to the availability of spouses not easily available within commitment relations. Insofar as people try to find spouses within the narrow range of commitment relations, and given the incest taboo, completely maintaining commitment relations and closing
147 relations to outside groups would have resulted in a situation in which too many people could not have spouses. In other words, maintenance of commitment relations incurred opportunity costs in the form of not having spouses. Various marriage systems are speculated to be responses of humans to the challenge of maintaining stable commitment relations, on the one hand, and avoiding the opportunity cost of doing without spouses, on the other. The purpose of many such marriage rules is to secure a stable supply of spouses through forming commitment relations with other groups. Using this method, that is, by extending networks of commitment relations, humans have succeeded in lowering opportunity costs while keeping the transaction costs of finding spouses low. We can imagine that when humans developed civilizations and started stocking surplus foods, the opportunity cost of staying in small commitment relations increased dramatically, since exchanges of surplus resources for scarce ones generate profits. In such an environment, staying in a commitment relation means paying opportunity cost since one who stays forgoes benefit that could have been obtained through exchanging resources. The method used in many societies to reduce this kind of opportunity costs again is to extend commitment relations, the same method as in the case of finding spouses. The method is to build and extend the network of commitment relations through which resources are exchanged with specific partners. Throughout the long history of humanity, people have kept social uncertainty low by reducing newly emerged opportunity costs through extending networks of commitment relations. This strategy has been still prevalent even in the modern world and even in the area of business relations until very recently. One example in the area of business showing that this strategy, that is, to reduce opportunity costs by extending committed relations while keeping transaction costs low, functioned very well is the Japanese style of management characterized by keiretsu relations and lifetime employment. However, the Japanese style of management emphasizing commitment relations has come under critical scrutiny recently. The reason for loss of confidence in that management practice among Japanese is, as I have mentioned several times, that the commitment extension strategy no longer can handle rapidly increasing opportunity costs. If this tendency continues and opportunity costs continue to increase rapidly, sooner or later, even in Japan, the strategy of extending networks of commitment relations will yield to the alternative strategy of utilizing open-market-like relations, since that alternative strategy is more effective at reducing opportunity costs. The important point to remember here is that the reduction in social uncertainty and the assurance of security provided by commitment relations will disappear in every corner of Japanese society as a consequence of this change. At the same time, the myth of safety in Japanese society also will disappear.
Breakdown of Trust and Breakdown of Assurance of Security Zucker (1986) discusses social changes similar to those we currently face in Japan, social changes that occur when increasing opportunity costs entice people out of fixed networks of stable commitment relations. She discusses social changes that occurred in American society in the late 19th century and the early 20th century, especially changes in American business around the turn of the century. She argues the following. When the assurance of security provided by yakuza-type commitment relations (although she does not use these terms) was disappearing in American society around the turn of the century, various institutions that facilitated open business relations based on general trust were developed, and this change built the foundation for business practices in modern America. In Japan, we are facing a similar situation in which it is getting more and more difficult to rely on the assurance of security provided by commitment relations. In discussing how the future of Japanese society should look, we should listen to Zucker’s argument that fair and effective social, economic, and political institutions are the precondition for nurturing social and business practices based on trust in a more open society. Zucker’s argument that American society around the turn of the century successfully dealt with the same kind of problem that Japanese society currently faces—transformation from a commitment based society that produces assurance of security to an open society based on general trust—and built the foundation of America’s current prosperity, however, does not mean necessarily that general trust continues to function effectively as a lubricant of social and economic activities in contemporary American society. Currently, voices are being raised that warn against a collapse of trust in American society. The fact that research on trust has become popular in recent years, including the book by Fukuyama (1995) I referred to earlier, and that academic conferences on trust are being held frequently seems to reflect the view widely shared among social scientists that a breakdown of the trust that has supported American society is a serious problem that needs urgent treatment. While a breakdown of the general trust that traditionally has provided the foundation of an open society in some highly advanced industrial societies like American society has come to be realized as an urgent problem, another kind of breakdown, that is, breakdown not of trust but of assurance of security, constitutes an urgent problem in other parts of the world. Japanese society, on which this book has focused, is one example. More extreme examples are found in the former Eastern Bloc countries where assurance of security collapsed due to the weakening of the central authority. Until the collapse of the former Soviet Union, not commitment relations among individuals or families but the central authority had provided social order and assurance of security in those countries. With the collapse of the central authority, those countries currently face the problem of how to provide a source for the needed security. Generally speaking, when the central authority weakens and the assurance of
149 security that was provided by the central authority disappears, other means of providing needed assurance appear. We have discussed repeatedly that the most common means of achieving this goal is the formation and maintenance of yakuza-type commitment relations. Interestingly enough, in Russia where the central authority weakened and assurance of security is lacking, real yakuza known as the Russian Mafia has shown dramatic growth. Gambetta (1988, 1993) studied the real Mafia in Italy and argued that the Mafia is an “industry” that provides assurance of security or private protection that is needed to reduce social uncertainty in Southeast Italy. When the central authority weakens and the assurance of security that was provided by the central authority becomes unavailable, such as in Japanese society immediately following her defeat in World War II, private power organized along commitment relations such as the yakuza and Mafia expands its reaches. The expansion of such commitment-based provision of protection is in a sense natural considering the importance of yakuza-type commitment relations as a means of dealing with social uncertainty. However, of course, it does not mean that providing assurance of security by such means is desirable.
Subjective Transformation of Assurance of security to Trust This book shows two ways to deal with social uncertainty. One is to form yakuza-type commitment relations or, I would say, to form collectivist relations. Security is assured within yakuza-type commitment relations. In many cases, people call the assurance of security provided within yakuza-type commitment relations “trust.” The stereotypic view that Japan is a “trust-based society” is based on this confusion between assurance of security and trust. The confusion between assurance provided by yakuza-type commitment relations and trust is not limited to the stereotypic characterization of Japanese society as a trust-based society. We almost can say that this confusion creates the “social reality” of people who live in yakuza-type commitment relations. That is, when security is provided by yakuza-type commitment relations, it is often the case that, in the subjective world of people involved, they believe that they are bonded by trust as typically is depicted in yakuza films. This is shown clearly in responses to the post-experimental questionnaire in the experiments introduced earlier. According to responses to the post-experimental questionnaire in the first and second experiments, participants who formed a commitment relation with a specific partner thought that the partner they chose was trustworthy and had a desirable character. It is also clear from responses to the post-experimental questionnaire that they formed commitments with specific partners because they thought that the partners were trustworthy. In the subjective “reality” of those participants, the partner’s trustworthy character is the reason for forming a commitment with the partner. However, results of those experiments clearly show that that “reality” is only a part of the “real” reality. Participants themselves did not realize the reality
150 that they were induced to form commitment relations by social uncertainty. No matter how precisely we investigate the subjective “reality” of people, we cannot reach reality of which they themselves are not aware. If we had started from the subjectively “reality” of the participants, we would not have reached the insight that their behaviors were produced by experimental manipulations. Similarly, no matter how precisely we understand the subjective meanings of the world in which people live, we will never approach full insight into the mechanisms that produce their minds and behaviors. These experimental results clearly show one problem with the subjective approach adopted by a certain type of social scientists.
Modern Society, Assurance of security, and Trust In concluding this book, let me return to the problem of trust that we are facing currently. As I discussed earlier, we are facing two types of general problems concerning trust. The first is the problem of a “breakdown of trust” that advanced industrial societies in the North America and Europe are facing. Regarding this problem, I believe that a breakdown of trust has not occurred yet at least in current American society. According to an analysis I presented in another article (Yamagishi, Kikuchi & Kosugi, 1999), what is happening in American society is not a breakdown of trust but an increase in caution. To be cautious in an environment in which caution is needed is an aspect of social intelligence; it does not imply necessarily a lack of general trust. Whether general trust collapses or not ultimately will depend on whether or not the feature of the environment that gives trust an adaptive advantage continues to exist. In this regard, the high levels of both social uncertainty and opportunity costs in advanced industrial societies will not change easily in the future. However, whether or not these two features of the social environment—high social uncertainty and high opportunity costs—are sufficient to keep the level of general trust high among citizens of those societies needs further examination. An especially necessary task for future analysis is to determine how to establish and maintain social institutions that work efficiently to prevent social uncertainty from becoming excessively high. Zucker argues that American society created the foundation necessary for a trust-based open society by establishing various social institutions and laws around the turn of the century. This suggests that a minimum level of assurance of security needs to be provided for activities that take place outside commitment relations in order for general trust and trustworthiness that is backed up by social intelligence, as a set, to acquire adaptive advantage. If the increase in caution in American society reflects a decrease in the efficiency of social institutions that provide such a minimum level of security, it should be recognized as a serious problem. The second type of problem regarding trust that we are facing is the breakdown of assurance of security in former communist countries in which a strong central authority previously had provided assurance to its people. The problems deriving
151 from this are shown most typically in the increased influence of organized crime in Russia, the so-called Russian Mafia. To deal with the breakdown of the assurance of security that was provided by the strong central authority, what is urgently needed in these countries is provision of the needed assurance at the grass roots level based on commitment relations. As I have emphasized many times in this book, the simplest and easiest way to deal with this problem for people who cannot rely on the central authority for assurance is to form yakuza-type commitment relations with specific partners. In this book, I have emphasized the negative aspect of yakuza-type commitment relations, namely, opportunity costs. Now, I emphasize its positive side as well: it provides assurance of security. Countries that are facing the breakdown of mechanisms for assuring security need most to establish networks of commitment relations to provide the assurance needed and to expand the reach of those networks. Organized crime, represented by the Russian Mafia, is expanding rapidly using exactly this strategy. Establishing networks of commitment relations among legitimate businesses and organizations would provide a source for assuring security throughout the whole society, and not only would lead to restraint of Mafia activities but also would build foundations for economic development in these countries. Here I would like to remind you of Fukuyama’s argument that too strong family ties keep trust from expanding beyond the boundaries of the family. I have pointed out repeatedly in this book that the same problem exists with Japanese collectivism. That is, the closed nature of groups that is produced by the principle of in-group favoritism prevents the development of general trust. However, the principle of in-group favoritism in Japanese collectivism is applied to a group larger than the family, and because networks of commitment relations between groups are established and expanded, we can say that Japanese collectivism has succeeded to a certain extent in reducing opportunity costs. The strategy adopted by Japanese collectivism of expanding the reach of assurance bases through network expansion may be the most promising direction to go, at least for now, for countries currently suffering from a lack of mechanisms for assuring security due to weakening of the central authority. Finally, let us return to the problem that Japanese society currently faces. Japanese society has been relatively successful until recently in providing assurance of security (or, using economic terms, in reducing transaction costs), on the one hand, and in keeping opportunity costs minimal, on the other, by expanding networks of commitment relations. Moreover, as testified by the chorus of applause for Japanese style management, this strategy has worked quite nicely until quite recently. That is probably because the opportunity costs of staying in commitment relations were not too high, and thus they were dealt with somehow by extending network relations. However, Japanese society currently faces an increase in opportunity costs too rapid for network extensions to accommodate successfully. This creates a situation in which pursuing open-market type relations is likely to bring better outcomes than
152 maintaining extended commitment-type relations. Facing this situation, the needed transformation of Japanese society to a more open society that allows for effective utilization of various opportunities will depend on whether we can promote the development of general trust based on social intelligence. I believe this is where the central message of this book becomes socially relevant. General trust is grounded in social intelligence. We need to realize the importance of this idea, which I have tried to convey throughout the book, for anyone who wants to raise the level of general trust in society. At the same time, let me emphasize at the end of this book that social intelligence is not sufficient, by itself, for the development of general trust. In order to nurture general trust as a foundation for an open society, as Zucker argued, we need to establish effective and fair social, economic, and political institutions that do not depend on specific relations but follow universal principles. Given such effective and fair institutions, the set of traits consisting of social intelligence, trust, and trustworthiness should become advantageous for individuals in society to have as the overall level of opportunity costs for staying in commitment relations increases. It is only in such a situation that the foundations of an open society, which cannot be achieved by sermons or slogans, will emerge. That is a society in which honest people who trust others are not exploited but prosper. The reason I appreciate an evolutionary game approach highly is that it gives us hope that there is a possibility that such a utopia can emerge by itself if an appropriate environment is arranged, rather than simply being created in people’s minds. I believe that an evolutionary game approach has the potential to become a tool for showing us the way to make that happen. It is my hope to contribute, through my research, to making the possibility greater, by however small an increment.
References Akerlof, G. A. 1970. The market for ‘lemons’: Qualitative uncertainty and the market mechanism. Quarterly Journal of Economics 84, 488-500. Axelrod, R. 1984. The evolution of cooperation. New York: Basic Books. Bacharach, M., & Gambetta, D. 1997. Trust in signs. Mimeo. Oxford University. Barber, B. 1983. The logic and limit of trust. New Brunswick: Rutgers University Press. Cash, T. F., Stack, J. J., & Lune, G. C. 1975. Convergent and discriminant behavioral aspects of interpersonal trust. Psychological Reports 37, 983-986. Chun, K., & Cambell, J. B. 1974. Dimensionality of the Rotter Interpersonal Trust Scale. Psychological Reports 35, 1059-1070. Coleman, J, S. 1990. Foundations of social theory. Cambridge: Harvard University Press. Cook, K. S., & Emerson, R. M. 1978. Power, equity and commitment in exchange networks. American Sociological Review 43, 721-739. Dasgupta, P. 1988. Trust as a commodity. Pp. 49-72 in D. Gambetta(Ed.), Trus: Making and breaking cooperative relations. Oxford: Basil Blackwell. Emerson, R. M. 1976. Social exchange theory. Annual Review of Sociology 2, 335-362. Erikson, E. H. 1963. Childhood and society, 2nd ed. New York: Norton. Frank, R. H. 1988. Passions within reason: The strategic role of emotions. York: W. W. Norton.
Fukuyama, F. 1995. Trust: The social virtues and the creation of prosperity. Glencoe, IL: The Free Press. Gambetta, D. (Ed.) 1988. Trust: Making and breaking cooperative relations. Oxford: Basil Blackwell Gambetta, D. 1993. The Sicilian Mafia: The business of private protection. Cambridge: Harvard University Press. Gardner, H. 1983 Frames of mind: The theory of multiple intelligences. New York: Basic Books. Geller, J. D. 1966. Some personal and situational determinants of interpersonal trust. Unpublished doctoral dissertation, University of Connecticut, cited in Rotter, 1980a. Goleman, D. 1995. Emotional intelligence: Why it can matter more than IQ. Bantam Books. Granovetter, M. 1974. Getting a job: A study of contacts and careers. Harvard University Press.
154 Gurtman, B. M. 1992. Trust, distrust, and interpersonal problems: A circumplex analysis. Journal of Personality and Social Psychology 62, 989-1002. Hamsher, J. H., Jr. 1968. Validity of personality inventories as a function of disguise of purpose. Unpublished doctoral disseration, University of Connecticut, cited in Rotter, 1980a. Hardin, R. 1991. Trusting persons, trusting institutions. In R. J. Zeckhauser (ed.), Strategy and choice (pp.185-209). Cambridge, MA: MIT Press. Hardin, R. 1992. The street-level epistemology of trust. Politics and Society 21, 505-529. Hayashi, N. 1993. From TIT-FOR-TAT to OUT-FOR-TAT. Sociological Theory and Methods 8, 19-32. (In Japanese with an English abstract) Hayashi, N. 1995. Emergence of cooperation in one-shot prisoner’s dilemmas and the role of trust. Japanese Journal of Psychology 66, 184-190. (In Japanese with an English abstract) Hayashi, N., Jin, N., & Yamagishi, T. 1993. Prisoner’s Dilemma network: A computer simulation of strategies. Research in Social Psychology 8, 33-43. (In Japanese with an English abstract) Hayashi, N., & Yamagishi, T. 1997. A socio-relational basis of “irrational” cooperation: An experimental study. Japanese Journal of Psychology 67, 444-451. (In Japanese with an English abstract) Hayashi, N., & Yamagishi, T. 1998. Selective play: Choosing partners in an uncertain world. Personality and Social Psychology Review 2, 276-289. Heath, A. 1976. Rational choice and social exchange: A critique of exchange theory. Cambridge: Cambridge University Press. Jin, N., Hayashi, N., & Shinotsuka, H. 1993. An experimental study of Prisoner’s Dilemma network: The formation of commitment in selective dyads. Japanese Journal of Experimental Social Psychology 33, 21-30. (In Japanese with an English abstract) Kakiuchi, R., & Yamagishi, T. 1997. General trust and the dilemma of variable interdependency. Japanese Journal of Social Psychology 12, 212-221. (In Japanese with an English abstract) Kaplan, R. M. 1973. Components of trust: Note on use of Rotter's scale. Psychological Report 33, 13-14.. Kikuchi, M., Watanabe, Y., & Yamagishi, T. 1997. Judgment accuracy of other’s trustworthiness and general trust: an experimental study. Japanese Journal of Experimental Social Psychology 37, 23-36. (In Japanese with an English abstract) Kiyonari, T., & Yamagishi, T. 1996. Distrusting outsiders as a consequence of commitment formation. Japanese Journal of Experimental Social Psychology 36, 56-67. (In Japanese with an English abstract) Kollock, P. 1994. The emergence of exchange structures: An experimental study of uncertainty, commitment, and trust. American Journal of Sociology 100,
155 313-345. Kosugi, M., & Yamagishi, T. 1995. Trust as a cognitive trait. Proceedings of the 43rd Annual Meetings of the Japanese Group Dynamics Association, 150-151. (In Japanese) Kosugi, M., & Yamagishi, T. 1996. Trust and gullibility. Proceedings of the 37th Annual Meetings of the Japanese Social Psychological Association, 288-289. Lajoy, R. J. 1975. The effects of arrogance and expertise on the communications of physicians and auto repairmen. Unpublished doctoral dissertation, University of Connecticut, cited in Rotter, 1980a. Lewis, J. D., & Weigert, A. 1985. Trust as a social reality. Social Forces 63, 967-985. Luhmann, N. 1979. Trust and power. Chishester, U.K.: Wiley. Nakatani, I. 1990. The origin of the Japan problem. Tokyo: Kodansha. (In Japanese) Orbell, J. M., & Dawes, R.M. 1991. A “cognitive miser” theory of cooperators’ advantage. American Political Science Review 85, 515-528. Orbell, J. M., & Dawes, R. M. 1993. Social welfare, cooperators’ advantage, and the option of not playing the game. American Sociological Review 58, 787-800. Pruitt, D. G., & Kimmel, M. J. 1977. Twenty years of experimental gaming: Critique, synthesis, and suggestions for the future. Annual Review of Psychology 28, 363-392. Putnam, R. D. 1993. Making democracy work: Civic traditions in modern Italy. Princeton, NJ: Princeton University Press. Putnam, R. D. 1993. The prosperous community: Social capital and public affairs. The American Prospect, Spring, 35-42 Rempel, J. K., & Holmes, J. G. 1986. How do I trust thee? Psychology Today February, 28-34. Rosenberg, M. 1957. Occupations and values. Glencoe, IL: Free Press. Rotter, J. 1967. A new scale for the measurement of interpersonal trust. Journal of Personality 35, 651-665. Rotter, J. 1971 Generalized expectancies for interpersonal trust. American Psychologist 26, 443-452. Rotter, J. 1980a. Interpersonal trust, trustworthiness, and gullibility. American Psychologist 35, 1-7. Rotter, J. 1980b. Trust and gullibility. Psychology Today. 102, 35-42. Snijders, Chris. 1996. Trust and commitments. Utrecht, the Netherlands: Interuniversity Center for Social Science Theory and Methodology. Sternberg, R. J. 1988. The triarchic mind: A new theory of human intelligence. New York: Viking Press. The Institute of Statistical Mathematics (Ed.) 1982. The Japanese national character,
156 Volume 4. Tokyo: Idemitsu Shoten. Watabe, M., Haruna, Y., & Kitada, A. 1994. The structure of trust in the safety of nuclear power generation. INSS Journal 1, 69-92. Watanabe, S. 1991. Job-searching: Social networks and mobility outcomes. Japanese Sociological Review 165, 1-16. (In Japanese) Wright, T. L. 1972. Situational and personality parameters of interpersonal in a modified Prisoner’s Dilemma game. Unpublished doctoral dissertation, University of Connecticut, cited in Rotter, 1980a. Wrightsman, L. S. 1974. Assumptions about human nature: A social-psychological analysis. Monterey, CA: Brooks/Cole. Yamagishi, M., Kosugi, M., & Yamagishi, T. 1996. A method to control for acquiescence responses in cross-cultural questionnaire studies. Japanese Journal of Social Psychology 12, 33-42. (In Japanese with an English abstract) Yamagishi, M., & Yamagishi, T. 1989. Trust, commitment, and the development of network structures. Paper presented at the Workshop for the “Beyond Bureaucracy Research Project,” December 18-21, Hong Kong. Yamagishi, T. 1986. The provision of a sanctioning system as a public good. Journal of Personality and Social Psychology 51, 110-116. Yamagishi, T. 1988a. The provision of a sanctioning system in the United States and Japan. Social Psychology Quarterly 51, 265-271. Yamagishi, T. 1988b. Seriousness of social dilemmas and the provision of a sanctioning system. Social Psychology Quarterly 51, 32-42. Yamagishi, T. 1990a. Major theoretical approaches in social dilemmas research. Japanese Psychological Review 32, 262-294. (In Japanese). Yamagishi, T. 1990b. Social dilemmas. Tokyo: Saiensu Sha. Yamagishi, T. 1992. Group size and the provision of a sanctioning system in a social dilemma. Pp.267-287 in W. B. G. Liebrand, D. M. Messick, & H. A. M. Wilke (eds.), A Social Psychological Approach to Social Dilemmas. Pergamon Press. Yamagishi, T. 1995a. Social dilemmas. Pp.311-335 in Karen S. Cook, Gary Fine, and James House (eds.), Sociological Perspectives on Social Psychology. Boston: Allyn and Bacon. Yamagishi, T.1995b. Have Americans really become distrustful ? Paper presented at the American Sociological Association Annual Meetings. Washington, D. C., August 19-23. Yamagishi, T. 1997. Trust and formation of relations: A cross “cultural” perspective. Pp. 321-342 in K. Ohbuchi (ed.), Social psychology of conflict resolution. Nakanishiya Shuppan. (In Japanese). Yamagishi, T. In Press. Trust and cooperation. In A. Kuper & J. Kuper (eds.), The Social Science Encyclopedia, 2nd Ed. London: Routledge. Yamagishi, T., & Cook, K. S. 1993. Generalized exchange and social dilemmas.
157 Social Psychology Quarterly 56, 235-248. Yamagishi, T., Cook, K. S., & Watabe, M. 1998. Uncertainty, trust and commitment formation in the United States and Japan. American Journal of Sociology 104, 165-194. Yamagishi, T., & Hayashi, N. 1996. Selective play: Social embeddedness of social dilemmas. Pp. 363-384 in W. B. G. Liebrand & D. M. Messick (Eds.), Frontiers in social dilemma research. Berlin: Springer-Verlag. Yamagishi, T., Hayashi, N., & Jin, N. 1994. Prisoner’s Dilemma network: Selection atrategy versus action strategy. Pp.233-250 in U. Schulz, W. Alberts & U. Müeller (Eds.) Social dilemmas and cooperation. Berlin: Springer-Verlag. Yamagishi, T., Jin, N., & Miller, A. S. 1998. In-group favoritism and culture of collectivism. Asian Journal of Social Psychology 1, 315-328. Yamagishi, T., & Komiyama, H. 1995. Significance and the structure of trust: Theoretical and empirical research on trust and commitment relations. INSS Journal 2, 1-59. (In Japanese) Yamagishi, T., Kikuchi, M., & Kosugi, M. 1999. Trust, gullibility, and social intelligence. Asian Journal of Social Psychology 2, 145-161. Yamagishi, T., & Sato, K. 1986. Motivational bases of the public goods problem. Journal of Personality and Social Psychology 50, 67-73. Yamagishi, T., Watabe, M., Hayashi, N., Takahashi, N., & Yamagishi, M. 1996. Trust and commitment under social uncertainty. Research in Social Psychology11, 206-216. (In Japanese with an English abstract) Yamagishi, T., Yamagishi, M., Takahashi, N., Hayashi, N., & Watabe, M. 1995. Trust and commitment formation. Japanese Journal of Experimental Social Psychology 35, 23-34. (In Japanese with an English abstract) Yamagishi, T., & Yamagishi, M. Trust and commitment in the United States and Japan. Motivation and Emotion 18, 129-166. Yamagishi, T., & Yamagishi, M. 1998. Trust and commitment as alternative responses to social uncertainty. Pp.109-123 in Mark Frunin (ed.), Network and markets: Pacific Rim investigations. Oxford: Oxford University Press. Zucker, L. 1986. Production of trust: Institutional sources of economic structure, 1840-1920. Research in Organizational Behavior, 8, 53-111.