Rationality and Society

Is Prisoner's Dilemma a natural framework for understanding human interaction? The author's view is that the rationality of the prisoner's dilemma, like rationality in general, is context dependent; without a supportive cultural context, no strategy makes sense. Different strategies will be rational for prisoners within different cultural contexts. The solution favored by most game players, namely, silence, for instance, would be anathema in a hierarchical culture that seeks to inculcate respect for authority. But silence would make sense for egalitarians who wish to undermine authority as inegalitarian on its face. “Distrust everybody” is a strategy that is supported only by people who adhere to fatalistic cultures. Thus the cultural context through which preferences are formed and reformed is central to who will place which values on diverse strategies. It is not rationality per se, as if one form of rationality was reason itself, but diverse rationalities that vary with the objectives of their adherents that should be the mainstay of game theory.

Game theory has had a considerable impact on the way we talk about problems and on our reasoning. But this has been more or less a philosophical effect. When it comes to formal reasoning, the calculation of a mixed strategy is characteristically either impossible or misleading.

Game-theoretical analyses of the iterated Prisoner's Dilemma (PD) supergame aim to deduce the existence of wholly endogenous solutions to collective action problems. So far, however, game theory has been unable to identify a unique equilibrium solution to the PD supergame. This lack of success, together with the assumption of complete and perfect information, suggests that these purely deductive searches for endogenous solutions are insufficient for understanding collective action problems in the real world. To be of use in the social sciences, applied game theory must be oriented toward specifying the links between its abstract categories and actual social structures.

This article comments on the works by Michael Hechter, Gordon Tullock, and Aaron Wildavsky that are critical of game theory. Although most of the points raised by the authors have merit, some are not relevant to game theory (as opposed to the misuse of game theory within social science). Those points pertaining to game theory vary in importance from quite central to exceedingly peripheral.

In this article, the author analyzes and argues against the claims made by Michael Hechter, Gordon Tullock, and Aaron Wildavsky that game theory has limited value for the social sciences.

This article discusses the value and limitations of game theory's use in the social sciences. The role of game theory is discussed and contrasted with exaggerated expectations of the subject. The importance of a modeling dialogue between theorists and empiricists is reviewed. The basic limitations of game theory are discussed, including the rationality and intelligence assumptions and the problem of multiple equilibria. The appropriate interpretation of randomized equilibria is illustrated.

A sharp distinction is made between normative and descriptive theories of decision. The former is purely deductive and thus can be regarded as belonging to a branch of mathematics. The results of a normative decision theory are not applicable to descriptions or predictions of how people will make decisions in given situations. Given a sufficiently precise definition of a “rational decision,” a normative theory can prescribe one in a given situation. Descriptive decision theory attempts to predict decisions of people in given situations sufficiently precisely described. In his critique of Prisoner's Dilemma, Aaron Wildavsky regards the theory underlying this game as a descriptive (or predictive) theory. For this reason, although his discussion calls attention to many interesting questions relating to people's decision behavior, his conclusions have no relevance to the theory of games.

Game theory has been criticized as neglecting key aspects of individual behavior and as relying too heavily on special assumptions. It can, in fact, handle individual heterogeneity if the modeler is willing to carefully specify how people are different, but to the extent that such things as heterogeneity and culture are important, the desire for a single unified model is impossible to satisfy. At the same time, game theory's approach is very useful for building specialized models.

An important task in the application of game-theoretic rational choice models to collective action problems consists in linking elements of the social structure to abstract concepts used in the modeling of game situations. Michael Hechter correctly identifies three major research areas in game theory (equilibrium selection, incomplete information, imperfect information) where such links are needed. However, we disagree with most of his conclusions on the problem situation in these areas.

This article focuses on selected claims made by three authors about game theory: “What games?” (Aaron Wildavsky), “Why equilibria?” (Gordon Tullock), and “Which equilibria?” (Michael Hechter). The intention is to bring some balance to the debate by pointing to the fact that most of the arguments discussed are conditional; they are true under some assumptions but false under others. The power of game theory stems from the fact that it enables us to seek solutions no matter what assumptions we believe to be true about the modeled phenomenon.

The Forum consists of comments on previously published papers and responses by authors. Unlike the Comment section in many other academic journals, the Forum includes solicited as well as unsolicited contributions. We encourage debates by actively seeking points of view contrary to those expressed in articles published in the journal. The Forum is intended to promote an open and critical debate that contributes to the intellectual vitality and further development of rational-choice-based theory and research.

Forum contributions should follow the form used by Rationality and Society and should not exceed five double-spaced pages.