In lieu of an abstract, here is a brief excerpt of the content:

CHAPTER 1 Evidence and Logic This chapter develops broad contours of the book's argument, adduces additional experimental and observational evidence, and establishes foundations for the more detailed treatments ofchapters 2 through 6. It extends the prologue's discussion of the limits of rational choice theory as well as the sociological/anthropological approach in understanding the empirical evidence, themes developed in detail in chapters 3 and 4. It addresses the relevance to these questions of the heuristics and biases research program, the subject ofchapter 6. And it contextualizes the key elements of the alternate route proposed to the solution of the Hobbesian problem, the concepts of multilevel selection and modularity, the subjects of chapters 2 and 5. The emphasis is on the relationship ofthese concepts to, respectively, traditional evolutionary biology and behaviorist psychology. On the empirical side, the chapter summarizes experimental results beyond those involving PD games, including those concerned with the voluntary provision of public goods, ultimatum, centipede, and dictator games. It probes the claim that failure to harm is the canonical form of human altruism, and thus that the phenomenon is more widespread than commonly appreciated, through the historical consideration of what appears to be the strongest countercase: war and genocide, in particular the war of extermination against the Jews conducted by the Third Reich. It discusses the interactions of nation-states, which operate in the absence of the coercive ruler Hobbes identified as necessary to solve the PD. Finally, in addressing why democracies do not fight each other, the chapter reaffirms that variations in political culture will influence the expression of essential predispositions. Born though we are with similar biological scaffolding, our continuing efforts to fashion, refashion, maintain, and change political structures have real consequences. The limits of Game Theory As the prologue suggested, game theory, the branch of microeconomic theory addressing strategic interaction, faces severe difficulties in accounting for observed behavior in Prisoner's Dilemmas. In the single play game 29 30 Altruistically Inclined? the unique Nash equilibrium is for each player to defect, but many players do otherwise. When games are repeated a fixed and known number of times, defecting at every stage remains the unique Nash equilibrium, although the strategy is no longer strictly dominant. If there is some probability one's counterparty might not be entirely rational, or might tremble a bit, it could pay (be rational) to employ other strategies (Kreps et al. 1982). This conclusion does not, however, follow ifboth players are assumed rational. Nevertheless, many sophisticated subjects cooperate throughout much of a finitely repeated game, the Alchian-Williams pairing a case in point. Models of single play and offinitely repeated PD games engaged in by rational players therefore imply unique equilibria and lead to clean, unambiguous predictions of behavior, predictions consistently at variance with the experimental evidence. More complex models do raise the possibility of explaining the persistence of cooperative outcomes within a rational choice framework. When PD games are repeated indefinitely, there may exist Nash equilibria involving some degree of cooperation, provided players do not discount future payoffs too much (Fudenberg and Maskin 1986; Fudenberg and Tirole 1991, chap. 5). The discount rate matters because the higher it is, the less important are later stages of the game, and the closer the interaction comes to being equivalent to a single play game. If we accept these assumptions, it is possible to "rationalize" how cooperative outcomes are sustained as an outcome of egoistic choice. Unfortunately, these models feature multiple equilibria, and not all of them involve cooperation. If we are interested in prediction, what grounds do we have for knowing which one will obtain? In what sense has theory predicted or explained a particular outcome? These conclusions, widely understood, help account for the strong predilection of game theorists (e.g., Binmore 1994, 1998a) to begin by assuming an environment of indefinitely repeated interaction. If one is interested in squaring theory with data, having a model that at least allows for the possibility ofcooperation as an outcome is surely preferable to one that precludes it. One must keep in mind, however, that what is being accounted for is the maintenance of an equilibrium, not its origin. Moreover, the applicability of the underlying assumption to real world situations is often questionable . A marriage may be modeled as involving indefinitely repeated interaction, but since in principle either party may end it, each has the option, at his or her own initiative, of converting the interaction into one of fixed and known duration. Moreover...

Share