Problems with Individualist Social Theory

Economists have engaged in a concerted attempt to improve the theory of rational action in order to make it an adequate philosophical framework for their discipline. In doing so they have developed the most systematic theories of rationality today. No discussion of rationality and/or rational action can avoid coming to terms with the sophisticated studies they have carried out and the enormous influence they have enjoyed. They are also of special significance for this inquiry because the dominant tradition in economics is individualist. But in spite of the wide acceptance of individualist theories of rationality, research by economists reveals most clearly the severe limits of individualist theories of explanation in the social sciences. They have been excellent critics of their own efforts, repeatedly pointed out weaknesses in their theories, and regularly tried alternative ways of overcoming them. A review of some of these efforts and how they have failed to overcome difficulties offers the opportunity of explaining how the weaknesses of individualist theories are based on traditional views of rationality. Individualist theories in economics regularly resort to the same logic which we find in Weber’s approach, that is, they use individualist assumptions to develop functionalist views. They presume that methodological individualism is true, that is, that individuals act to achieve their aims given the logic of their situation. They further presume that rational action of individuals in pursuit of their own goals when uninhibited by political control always produces functioning social systems. These systems are coherent and justified on the grounds that they satisfy the aims of individuals. The traditional demands that rational thought and actions be systematic and justified can be met both in regard to individuals and to social systems: individuals have coherent belief systems and act on them whereas social systems are organic, each part of a system contributes as if it were part of a coherent plan for the well-functioning of the system. The fundamental ideas of this approach are above all due to Adam Smith. But in more recent times strenuous efforts in so-called neoclassical economics have been made to make them more precise so that their explanatory and predictive power could be increased. These efforts have led to great tension because the dramatic increase in precision achieved through the use of mathematical models has led to the inability of these theories to describe the world. They are defended with the claim that even with the use of false assumptions economists are able to construct models which sufficiently mirror the world, that true and useful predictions can be made by them. Some economists are worried about the shaky

framework which is being used. Others are not. I side with the former because I presume that a good philosophical anthropology is needed for good social science. The philosophical anthropology of modern individualist theories-above all rational choice theories-assumes that only individuals exist and that society is a mere combination of their rational actions. Rational actions are presumed to require justification and calculation of the best alternative. Rational action requires coherent thought. The development of theories of action which are in accord with these strong and traditional assumptions requires further and quite dubious hypotheses. These philosophical anthropological hypotheses are introduced ad hoc as they are needed to explain how individuals are rational. They are introduced and used independently of discussions of whether they describe the world. Harsanyi presumes, for example, the existence of meaningful subjective probabilities, of subjectively determined but objectively comparable utilities, and of rational social systems at least. The defense of these assumptions seems to rest on a transcendental argument which goes something as follows. We know that individuals are rational and/or that social science requires us to assume that they are. And we know that the assumptions made in rational choice theory are necessary for rational action. Social scientists are, therefore, justified in making these assumptions. If social scientists did not make them, they would reject the possibilities of both rational action and of social scientific explanations. But rejecting these possibilities is absurd. The prime functions of the use of this technique-also employed in political theory by such thinkers as Rawls and Nozick-is to block discussions of alternatives and conceal the fact that one is engaged in old-fashioned speculative philosophical anthropology, that is, that one cannot get along with common sense and/or trivial assumptions about rationality. Old-fashioned speculative philosophical anthropology can be valuable. But it can be harmful when it is done surreptitiously. Bad philosophical anthropology, as in the case of the philosophical anthropology of rational choice theories, is the result. We need to do better. In this and the following three chapters I will discuss the problems which the philosophical anthropology of individualist theories, above all in economics, have faced and argue that they have failed to overcome them, because of their adherence to traditional standards of rationality. In this chapter I will introduce the ideal with a discussion of Harsanyi and central problem areas with this approach, that is, individuals are not rational in the way the theory says they should be, and rational action of individuals alone cannot explain the role of institutions in steering social events, with discussions of Schumpeter and Elster. In the succeeding three chapters I will discuss (failed) attempts to account for institutions within the program, the overestimation of the rationality of individuals which make the program so problematical, and the sympathetic criticism and program of Albert. Harsanyi’s dream In spite of serious difficulties facing individualist and rational choice theories, defenders of it seek to show that the dream of a society which moves toward

equilibrium, a society in which no change can make any person better off without making someone else worse off, and which is based on the free rational actions of individuals is realistic. They hope to move toward its realization with the aid of social scientific theories which describe how it functions. Among other things these theorists seek to show (1) how rational action conceived of as the pursuit of the maximization of expected utility in even complex situations is possible, (2) how such action will lead to coordination of efforts to produce the general good and thereby account for morality, and (3) how individual action guided by rational choice principles can account for institutions, that is, those rules and procedures which arise when individuals follow rational choice principles. In one such attempt John Harsanyi has made highly regarded contributions to technical problems facing this program. One needs to surmount the difficulties he takes on in order to say how one can act rationally in some complicated decision making situations. He has also offered a comprehensive view of how this program may be carried out, which includes a moral theory. Following a standard rational choice research program, Harsanyi assumes that a proper theory of rationality should provide a definite move which any rational person should make in any situation. This action should be determined by the individual’s aims, valuations and the relative probabilities of the success of each proposed alternative course of action. Still further, when the action of each individual in some society is rational, the best results for all individuals should result. If it does not, the theory of rational action cannot provide the proper framework for the theory of economic systems. The rational action of any number of actors should, then, lead to a system which is an unintended but desirable consequence of individual rational action and which is, as far as possible, beneficial for all. This demand for a beneficial system does not reflect good will, but is an internal requirement of the theory: if such a point could not be reached, usually conceived of as a system in which no one can improve their lot without harming someone else, all individuals could not be rational, that is, act so as to maximize expected utility in the same situation. The situation would be constantly changing so that no one could calculate their expected benefits. A theory of rationality should provide, then, the framework in which explanations of how a state of affairs in which no member can improve his lot without harming someone else is and/or can be achieved. It cannot pay to act irrationally nor can rational action lead to poor results for all, if the dream of rational choice economic theory is to be realized. By and large Harsanyi’s work research is not empirical. It consists of philosophical investigations into the possibility that any such ideal, one might say utopian, system such as that envisaged by economists could exist in some world, in which all individuals were free, intelligent and rational enough to pursue the maximization of their own utilities in effective ways. In order to pursue this ideal, Harsanyi has to make assumptions about rationality and morality which need to be applicable, if the utopian system were to operate as he dreams it should. We may then ask two questions. The first is: What gaps are there between the dream of a rational society and the reality which it is supposed to model? The second is, Given these gaps, of what use is the ideal? I will first summarize Harsanyi’s view, secondly give an overview of five gaps between Harsanyi’s dream and the reality it

is supposed to model, thirdly discuss them in detail, and finally contend that, whether one adopts either his standards or my standards for explanation in the social sciences, the pursuit of this ideal is not a reasonable course to take in social science today. Harsanyi carries further the program of von Neuman and Morgenstern according to which rational action can be studied with game theory, that is, various types of situations are analyzed as consisting of games which follow certain rules. The development of this program requires that increasingly complex situations be formally analyzed, such as those in which each player is ignorant of what other players will do. Harsanyi made an advance in game theory, by explaining how one player’s ignorance of the intentions and/or utilities of other players could be quantitatively built into his considerations. His idea is that one can reach acceptable results by assigning probability estimates to the various courses of action one’s opponents might take. Harsanyi uses a Bayesian, that is, subjective, approach to probability estimates, even while admitting that a good theory or method of probability assignments is lacking. Harsanyi thinks that the idea that functionalist and individualist explanations in the social sciences are incompatible is absurd and wonders why such a highly regarded social scientist as Ralf Dahrendorf could even suggest such a thesis. Individualist explanations can use game theory to take account of institutional factors, he says, by treating institutions as players. Moral factors, Harsanyi argues, may be an integral part of the rational man’s calculations, if he chooses to be rational, as most individuals do. Harsanyi seeks to base his utilitarian moral theory on appraisals of the greatest good for the greatest number. Morality requires that in any situation one take account of the expected utilities of all individuals. These can be determined, Harsanyi says (in principle?), by viewing any situation from the point of view of each individual and by taking the sum of the realization of utilities as the moral end. One should then act to maximize the mean value of these utilities, but not directly. One should ask which rule should be followed to achieve this end. But one cannot provide rules which one should follow in any situation, since in differing contexts following the same rule may have differing kinds of consequences. One needs to ask which system of rules will achieve this end of maximizing the sum of utilities in any given circumstance. Harsanyi views the theory of rationality as a work-in-progress. We have, on the one hand, the view that all social events should be explained within a framework offered by the rational choice theory of rationality. But, on the other hand, we have the fact that, in its present form, this theory can only be applied to a small range of social events. Harsanyi, then, seeks to narrow this gap by extending the range of social situations for which we can construct rational choice models. Harsanyi offers the outlines of a utopia of rational individuals in which (1) the rational action of each individual in each situation is specified uniquely, in which (2) all factors any individual needs to take into account in making his decisions, including institutional ones are taken into account, which (3) establishes equilibrium, that is, no individual can improve his situation without harming someone else, which is (4) also based on moral considerations, that is, on the consideration that the most number of individuals will achieve the most good

possible. He does not present this view as a vision of a utopia, but simply as what any reasonable person would assume about the nature of rational action and a rational society. This vision is very attractive indeed, since it would unite rational action in the sense of the planned pursuit of personal aims, with the maximum possible success for all, with morality. The rejection of it, on the other hand, would throw into question the aims of much economic research, which could no longer be that of showing how economic systems move toward equilibrium and how this process was driven by the rational action of individuals. It would mean that rules of rational action could not be reduced to pursuing the realization of the highest level of personal or general utility. One would have to take other kinds of considerations into account which would conflict with such action. Let us look at some gaps between Harsanyi’s dream of the rational society and the model of it available today. When the model is filled out, could it possibly describe and/or bring about a real world? Or, are gaps between the situation described by the model and reality so big that it makes no sense to use the former to talk about the latter? Does the pursuit of this dream bring us forward to a better understanding of man and to better social scientific theories? Or, does this pursuit do harm, by pursuing the unattainable and the useless while neglecting real possibilities? In order to answer these questions we need to look at gaps between the dream and the reality it is supposed to model more closely. I will first summarize the five gaps and then discuss each in detail. Firstly, in order to find any definitive approach in any situation one needs to have estimates of the probabilities of the outcomes of specific courses of action. But at the present time we cannot find any sensible way of doing this. Such estimates are made subjectively; it is simply presumed that subjective is not so arbitrary as to be worthless. Secondly, we not only need numerical preference ratings of preferred outcomes, but we need numerical estimations of the degree of preference of one outcome over another. But these seem quite arbitrary. We have no explanation of how such estimates can be made precisely. They are also said to be personal and not arbitrary. But we have no explanation of this capacity. Thirdly, the calculations which are needed to determine the ‘rational’ responses in many situations are so complex that hardly anyone is capable of making them at all and no one in actual situations, under the pressure of making decisions quickly. The analysis of decision-procedures do not, then, describe how people either do or can think. Fourthly, in order to model any social situation calculations need to be made by each person in this situation, which are based on his own subjective probabilities and utilities and which tell him which course of action is the best for him in that situation. In order to formulate such a problem-situation, an analysis of which alternatives are live possibilities-as James would put it. But any situation can be analyzed in many different ways from many perspectives. The most important aspect of making a decision is analyzing the problem. And this aspect of the problem has nothing to do with decision theory. Many different results, then, are always possible within the bounds of decision-theory. And, the analysis of any

situation in terms of game theory requires that different players agree on which game is being played. But for the most part, they do not. Elster reports that his attempt to develop such an analysis for collective bargaining in Sweden failed because everything was up for grabs and fluid (Elster 1989a p. vii). There is no stable game over periods of time. Fifthly, Harsanyi’s moral theory requires each individual to make estimates of the utilities of other individuals. Such estimates require empathy and understanding, which we hardly possess in regard to any individual and which are completely fictitious when it comes to large numbers. After we have examined these gaps between the dream of the rational society and the model available today, we may ask whether it makes any sense to try to narrow and lessen the gaps, or whether some entirely different approach is called for. Does research on the rational choice model of society move us forward toward the utopian ideal of an economic system which is in equilibrium and which is a product of the rational actions of individuals as conceived of by this theory? Let us look at these five gaps in more detail. The first gap between Harsanyi’s dream of the rational society and the model of it available today is that between the assumption that the probability of specific outcomes may be used in the calculation of which course of action is the best and the fact that no such probability estimates can be made in any clearly reasonable way. Harsanyi is careful not to say that we know how to make probability estimates, even though he calls himself a Bayesian, that is, he presumes we may use subjective probabilities in the determination of which course of action to follow, that is, in computing expected utility, which is a function of the (subjective) probability of some event occurring and the comparative desirability of this event vis-à-vis other events for some person determined by the place of this event on some personally determined scale. The reason that Bayesians have been called by this name, however, is not merely that they use subjective probabilities, but that they use these probabilities to appraise the probability of the truth of theories. They claim to be able to do this, because they revise initial subjective probabilities of the truth of some theory in accord with new evidence for or against it with the use of Bayes’ theorem. This use of Bayes’ theorem has been effectively criticized: Popper and his followers have demonstrated that the probability calculus, including Bayes’ theorem, cannot be used to calculate the probability of the truth of propositions. The assumption that they can ignores the asymmetry between confirmation and refutation. Harsanyi avoids the question of whether we can use Bayes’ theorem to adjust probabilities of the truth of theories in accord with positive and negative evidence, since he merely uses subjective estimates of the probability of events-a probability which may be objective, as Popper’s propensity theory of probability demonstrates-to calculate expected utility. It is, then, somewhat misleading to call him a Bayesian, simply because he uses subjective estimates of the probabilities of events. We may also note that Harsanyi rejects Popper’s view according to which the simplest empirical hypothesis is the least probable, because it is open to the most risk, that is, it entails the largest class of potentially refuting basic statements (Harsanyi 1976). Harsanyi argues that, when faced with a choice between two hypotheses, we should presume that the simplest one is most probable. He follows

Jeffreys who proposes that the theory with the fewest number of parameters should be deemed to be the simplest. This view of simplicity is also adopted by Popper. Harsanyi suggests that Popper’s criticism of the views of Jeffreys as well as of other inductivists is not convincing because it is external. It merely shows that these views contradict Popper’s. And there are independent reasons to reject Popper’s views. In particular, Harsanyi suggests, Popper accepts the (Wittgensteinian) view that each event is independent, that the occurrence of one event has no influence on the probability of some other event. Harsanyi rejects this view on the grounds that scientists presume the world to be law-like. Any combination of discreet events is not as probable as any other. It is, therefore, reasonable to assume that a simpler theory, one with fewer parameters, would be more probable than a more complex one, because it describes the world as simpler than its competitor. It is not, however, true that Popper’s criticism of Jeffreys’ theory is merely external. Popper argues that Jeffreys’ own view requires him to find some way of assigning prior probabilities but he cannot do this in any coherent way, that is, in any way that accords with the probability calculus. The sum of the probabilities of all alternatives cannot be greater than one. But we may always construct an infinite number of theories with the same number of parameters. For example, we may have ax1, ax2, ax3, etc. All members of this series have the same number of parameters, but constitute an infinite list. Since the sum of alternatives cannot be greater than one, we can only assign zero probability to each individual theory. Haransyi’s reference to Popper’s acceptance of Wittgenstein’s view might give the impression that Popper’s view here rested on some dubious metaphysical assumption. But this is not the case. Popper does not assume that the probability of each event is independent from other events because Wittgenstein’s metaphysics demands it, but because the probability calculus does. Instead of discussing how prior probabilities can be sensibly assigned in terms of the calculus of probability, Harsanyi merely discusses the limited case in which we have two alternatives of differing parameters and assign the higher probability to the one with fewer parameters. Jeffreys attempted, and failed, to explain how this could be done systematically without contradicting assumptions made by the calculus of probabilities. But Harsanyi does not attempt to overcome the difficulties of assigning prior probabilities. He merely asserts that we may do this in various ways, and that we may reject the assumption that we can only assign the a priori probability zero to theories, with the mere assertion that we can construct metrics which do not make this assumption. He dodges the problems of using the probability calculus to judge the probability of theories with no more than appeal to intuition. His theory of prior probabilities seems to be little more than mere appeal to inductivist dogma. He offers no theory of how the probability of events may be assessed in terms of the probability calculus such as those offered by Jeffreys and Popper. Instead he offers a vague theory of subjective feeling. Why should we take so-called subjective probabilities so seriously that we make decisions dependent on them? The normal explanation, one also adopted by Harsanyi, is something like the following. Such a procedure is a mere extension of ordinary decision-making procedures, that is, when we make any decision we

ordinarily ask what likelihood of success some possible course of action may have. This is true whether our decision is a very minor one about catching a bus or a major one about getting married or choosing a profession. We can hardly avoid estimating the likelihood of events when making decisions. It is reasonable to do that. And the best estimates of success we have are, unfortunately, subjective. So, we had better use them or we are left with nothing. The fact, however, that we assess our chances of being successful in any specific case does not mean that these assessments can be portrayed as probabilities in any reasonable way. If we do this and deceive ourselves, it would be better to take this fact into account, rather than to continue the (self-)deception that we have useful knowledge about the probability of some event happening. One question to ask, then, is whether subjective probabilities have anything to do with reality. In making assessments, Harsanyi says, we should take all the facts into account and be neutral. Why should we do that, if that will not help us make reasonable judgments in any known way? And, if it does help, how can we explain and show that? Harsanyi has no answers to these questions. He needs subjective probabilities for decision and game theory. But that does not show that it is sensible to make such quantitative probability judgments and to rely on them as knowledge in making decisions. It might be said that we have no alternative. But this is also not the case. Instead of directing our concern at the appraisal of expected utility, we can direct our concern at the criticism of decisions and institutions. We would, thereby, sacrifice the aim of a decision theory, which explains what we ought to do in any situation. But at the mere price of accepting responsibility for our decisions, we may decrease bad ones, which are based on probability estimates which in so many cases have so little to do with reality that we may deem them arbitrary. It should be stressed here that Harsanyi seeks to extend decision and game theory to as wide a range of events as possible, that is, to those events that go far beyond our ordinary assumptions that there is a high probability that the bus will be at least five minutes late. The wider the scope of the use of these theories the less defensible the use of subjective probabilities becomes, because the more they have to apply to situations which are not repetitive. We lack any testable probabilistic hypotheses. The second gap between the dream of the rational society and model of it we now have available is the use of numerical assessments of utility, that is, of the desirability of particular outcomes. These are also subjective. They do not correspond to anything which is comparable between individuals, that is, one person’s scale of utilities is not comparable with another person’s. Harsanyi does claim however that the utilities assigned by one person to various events can be compared. Our utilities are what we say they are, neither more nor less, to paraphrase Lewis Carrol’s Humpty Dumpty. They may change from moment to moment, from mood to mood. Now, Harsanyi explains that the central point is that they create some order in our appraisal of the desirability of various events taking place for some person. But even the construction of this order will not enable us to calculate which result we want to bring about, for we must also be able to say that one outcome has, say, ten times the value of some other, as Harsanyi also presumes in endorsing cardinal and not merely ordinal utilities. Otherwise calculations of

expected utility could not be made in any sensible way: we could not assess to what degree we would be willing to encounter some greater risk to reach a more desirable outcome. But how can we assign such numbers? We have no method. So, why should we take such assignments so seriously that we plug them into formulas and accept the results as an accurate reflection of our wishes? Comparative utility assignments seem to be a rather flimsy basis for any decision. Yet they are crucial for any rational choice program. When we have assigned a numerical value for some outcome and have plugged it into our formula, we are committed to the result if we are, on this view, rational. If we reject it and adjust our utilities, we can hardly be said to have any serious basis for our calculations. It would seem better to have conjectures which we test and correct when we can. The third gap between the dream of the rational society and the model of it which is available now is that between, on the one hand, the way the theory says rational individuals should calculate in order to determine which course of action in any situation maximizes expected utility, and, on the other hand, the way individuals in fact think and are able to calculate. It is clear and admitted by developers of rational choice theories-Harsanyi included-that at least in complicated cases these are different. Yet, the whole program is based on its ability to account for complicated cases of decision-making as real individuals do and/or should make them. As the formal theory is developed to take account of such complicated cases, it describes actual decision-making as individuals engage in it less and less. Does this matter? It would seem so. But without any clear statement of what this theory seeks to accomplish, it is nearly impossible to judge. I have not found such a statement. Some think this will not do. Others say that insofar as the thought processes of humans are rational they will lead to virtually the same results. Chess players do not ‘think’ like computers do, but good chess players make rational moves, as de Groot has shown (de Groot 1965). Harsanyi and his collaborators on this program do claim that their results are useful. They claim that models of specific kinds of situations developed within the framework of rational choice theory can be profitably used in explaining the real world because they have predictive power. A number of variables, they say, can be deduced from their models. They help us, then, in understanding what is going on in specific situations by showing how factors which determine outcomes are related. The use of any model requires that we know under what conditions it may be used and when not. It is an empirical question. The model itself does not supply us with such information. The success which is achieved, then, is measured by finding that given some model of a specific situation, various factors may be predicted to have some effect and they do have this effect. This confirmation does not necessarily help us in some new situation, since we cannot presume that it will be the same without testing anew. (Friedman suggests that only economists can intuitively judge when a model is applicable or not (Friedman 1953).) In each new case we have to see if a specific model is applicable. If it is, we may then use it for this case, if that should help us to develop responses to some specified situation(s).

This kind of knowledge can be useful. It provides a social technology, which may be used in negotiations, for example. But even if there have been uses found for such models, this use is always problematical, as the Long Term Capital Management case shows, and narrow. The sophisticated technical procedures require well-defined situations. But in modern societies this is just what we do not have. ‘Opponents’ in social games regularly change conditions by introducing new ideas. In this process continual corrections are needed and complicated calculations are useless. Secondly, we may ask whether models constructed with the tools of the rational choice program are of more use than those constructed without them. In the case of rational choice theory, we see that much of the results which are claimed for this theory are also achieved by methodological individualist theories such as those of Weber, Popper and Hayek without the new technical developments. We may find that the increase in the technical development which is designed to increase the precision of application may have diminishing returns because rationality postulates have to become so complex to solve the theoretical (philosophical or metaphysical, that is, non-empirical) problems of the theory. Simpler ones will do. The fourth gap between the dream of the rational society and model of it we now have available is the widest and the most telling. Making any decision requires that one describe a problem situation and possible responses to it. These decisions are the most important factor in determining what is to be done. Yet rational choice theory has nothing to say about them. (Esser seeks to overcome this problem, see below.) Still further, they are steadily revised. It is only by taking very narrow and arbitrary descriptions of situations that the calculus of decisionmaking can even be used. Simon has tried to lessen the difficulty but, as I argue below, cannot remove it. The fifth gap between the dream of the rational society and the model of it we now have available is found in Harsanyi’s moral theory. This theory requires each individual to make estimates of the utilities of other individuals. Such estimates require empathy and understanding, which we hardly possess in regard to any individual and which are completely fictitious when it comes to large numbers. In order to judge the morality of any action, Harsanyi says, we should construct a welfare function which is the mean of the expected utilities of all individuals. He has no theory of which individuals one should take into account, perhaps all of mankind and perhaps animals too. This welfare function would enable us to decide which act should be favored from a moral point of view. A moral judgment is still more complex, however, because it is not enough to choose that action which maximizes the mean of the expected utilities of all individuals. Rather, we have to act in accord with the rule which will do that in the most number of cases. We also cannot ask which individual rule we should follow, but must ask which system of rules we should follow and apply that system to the individual case. This is poor metaphysics with no relation whatsoever to making moral judgments in the real world. Harsanyi’s theory should show the possibility of a moral theory in the context of rational decision making. It does nothing of the kind, because it gives no reason why we should think that the kind of decisions this theory requires could even be meaningfully posed, much less answered.