This is a challenging book to review for an economic history audience. The bulk of it is highly technical, consisting of a detailed exposition, development, and critique of game theoretic models as they apply to the social and behavioral sciences, interwoven with discussions of the findings of experimental/behavioral economics. As the title suggests, Herbert Gintis argues that game theoretic methods must lie at the core of forward progress. The most accessible material is in chapter 12, a provocative and original synthesis of Gintis's views on the unification of the behavioral sciences, which he defines as including economics, political science, psychology, sociology, and anthropology, along with biology insofar as it applies to human behavior. This book can be and should be read by anyone working in these areas.

And it should be discussed. There are few whose views will go unchallenged, yet almost all will find, sometimes surprisingly, that the value of their approaches are, at least in part, endorsed. Gintis identifies economic, sociological, psychological, and biological frameworks that have associated with them distinct methodological traditions that are in many respects incompatible. They are incompatible because each makes claims about how and why people make decisions that are denied by others. The situation was tolerated if not tolerable so long as each discipline focused on separate and circumscribed areas of behavior or interaction (economists, for example, concentrating on market exchange). This intellectual division of labor, however, can no longer be sustained. If we are all students of human behavior, he argues, we need a unified framework within which to approach our subject. Gintis aims to modify these different approaches, removing their incompatibilities and allowing them to complement each other. The enterprise is to be built around game theoretic methods.

Yet the broad endorsement of the value of game theory is qualified. He is critical of the state of research in microeconomics (of which game theory forms part), contrasting the degree to which truly scientific disciplines have advanced in the past half century with what has happened in economics. At the same time, while he sees great value in the sociological approach, he is also critical of how it has evolved. When he writes, delicately, that “the behavioral sciences currently consist of some fields where theory has evolved virtually without regard to facts and others where facts abound and theory is absent” (p. xvi), it is clear that he is Catholic in his critical targets, and no one can easily or quickly claim that his or her own field is not in his sights.

Thus Gintis argues, in a way that will by turns infuriate and encourage, that while “game theory without broader social theory is merely technical bravado…social theory without game theory is a handicapped enterprise” (p. xiii). Regardless of where one stands with respect to the value of these methods, his ambitions will raise eyebrows. He has clearly set for himself a daunting challenge, one which some readers will probably compare to that of squaring the circle.

To make the case for the centrality of game theory, Gintis must address a number of important problems. There are two which are central. First, even though the famous folk theorem demonstrates that, within a context of indefinitely repeated interaction, cooperative outcomes can be sustained as a Nash equilibrium, there are almost always multiple equilibria (a Nash equilibrium is defined as an outcome where no one can choose a better strategy given what everyone else is doing; everyone is playing best response). Once arrived at, a particular equilibrium can be sustained as the result of the behavior of self-regarding actors. But why one rather than another? There is no clean analytic mechanism to explain or determine why a particular equilibrium is arrived at or coordinated upon.

Because of this lacuna, much game theoretic literature has involved the development and articulation of equilibrium selection criteria. Subgame perfection, due to Selten, is perhaps the best known, but there are many others. Gintis focuses, for example, on the work of Aumann on correlated equilibria. But while these different criteria can, with greater or lesser success, appeal to our intuition in terms of their plausibility, that is all they can do. There are not scientific criteria for choosing among them. No one who believes that game theory should be central to the understanding of social outcomes can be comfortable with this situation, or believe that it is healthy. As Gintis notes diagnostically, “the economic theorists' overvaluation of rigor is a symptom of their undervaluation of explanatory power” (p. xviii).

Here is where Gintis identifies, in a straightforward fashion and without embarrassment or shame, the value of social theory and the sociological approach. Not to beat around the bush: many game theorists are actively hostile to some of the concepts, such as social norms, the value of which Gintis, to his credit, embraces. We cannot understand equilibrium selection, he maintains, if we do not appreciate the etiology and causal significance of norms, and in particular the fact that evolutionary processes have endowed us with other regarding preferences.

That point of view is also important in addressing a second and related issue. As noted, in indefinitely repeated games, there are many equilibria, and we need to explain why one rather than another is selected. But we are not always and necessarily embedded in repeated interaction. Indeed, to get to repetition, we must often negotiate situations that might well end up one shot. Yet in many one shot strategic interactions, the challenge to game theory is even more direct: people simply don't behave the way theory says that self-regarding agents should.

Gintis draws a nice distinction between decision theory, which involves optimizing against (a nonstrategic) nature, and game theory. In the former, when one points out to subjects that their behavior deviates from the normative prescriptions of theory, and why, they will generally alter their behavior.

Not so in comparable circumstances involving strategic interaction. When push comes to shove, there is simply no reason why, in a one shot Prisoner's dilemma, any self-regarding individual will cooperate. But cooperate we do, both inside the lab and outside of it, and there is a range of robust experimental results involving trust, dictator, and other games where a similar phenomenon manifests itself. Explaining to subjects why their behavior deviates from that prescribed by theory and why, will in these instances, generally not cause the behavior to change.

The reason, says Gintis, is that as humans we approach problems with a combination of self-regarding and other regarding inclinations or preferences, a reality that he explains as the consequence of gene-culture coevolution. He deserves plaudits for questioning the widely accepted view that mechanisms of kin selection (Hamilton) and reciprocal altruism (Trivers) are sufficient to dispose of the “problem” of apparently biologically altruistic behavior. I would have preferred a more full throated endorsement of the importance of biological group selection in producing this outcome (he mentions the “highly interesting debate” but says, rather enigmatically, that it has been “clarified but not completely resolved” (p. 244). But I am broadly sympathetic to what Gintis suggests, and the need for confronting these results head on, rather than trying to dismiss them as artifacts of the experimental method.

There are many provocative aspects of the book I can only touch on. These include his criticism of behavioral economist's dismissal of decision theory as emotionally satisfying, but “immature, shortsighted, and scientifically destructive (p. 246). Nor have I discussed the author's rejection of methodological individualism, and his embrace of complexity theory, agent-based simulation methods, and the value of the concept of emergent properties.

Gintis intends this book not as the end of a conversation, but as the beginning of one. He is open to and indeed receptive to methods and concepts drawn from a variety of disciplines, rather than starting with narrow and sometimes ill-founded prejudices against one or another. This is a conversation in which many can productively participate. He writes, for example, in a fashion that will appeal to many readers of this Journal, that “the hypthetico-deductive methods of game theory, the BPC model, and even gene-culture evolution must be complemented by the work of behavioral scientists who adhere to more historical and interpretive traditions” (p. 243) (BPC refers to behavior-preferences-constraints, a term he likes better than rational choice, because that terminology, he believes, is too often identified as requiring strictly self-regarding preferences).

Statements such as these are not simply throw away lines, designed to construct a big tent for his audience. Yet there are many unresolved tensions in what Gintis articulates, perhaps none more striking than the juxtaposition of such sentiments with eleven chapters of game theoretic analysis and results. Still, the exposition in these chapters is not uncritical, and he challenges us, by example, to confront these tensions, to try and come to terms with them, to overcome “interdisciplinary ethics that value comfort and tradition over the struggle for truth” (p. xvi).