Evolution and Game Theory

TLDR: This paper summarizes recent approaches to noncooperative game theory that have been based on evolutionary models on how to expect equilibrium play in games.

Abstract: Introduced by John von Neumann and Oskar Morgenstern (1944), energized by the addition of John Nash’s (1950) equilibrium concept, and popularized by the strategic revolution of the 1980s, noncooperative game theory has become a standard tool in economics. In the process, attention has increasingly been focused on game theory’s conceptual foundations. Two questions have taken center stage: Should we expect Nash equilibrium play—that is, should we expect the choice of each player to be a best response to the choices of the other players? If so, which of the multiple Nash equilibria that arise in many games should we expect? In the 1980s, game theorists addressed these questions with models based on the assumptions that players are perfectly rational and have common knowledge of this rationality. In the 1990s, however, emphasis has shifted away from rationalitybased to evolutionary models. One reason for this shift was frustration with the limitations of rationality-based models. These models readily motivated one of the requirements of Nash equilibrium, that players choose best responses to their beliefs about others’ behavior, but less readily provided the second requirement, that these beliefs be correct. Simultaneously, rationality-based criteria for choosing among Nash equilibria produced alternative “equilibrium refinements”—strengthenings of the Nash equilibrium concept designed to exclude implausible Nash equilibria—with sufficient abandon as to prompt despair at the thought of ever choosing one as the “right” concept. A second reason for the shift away from rationality-based game theory was a change in the underlying view of what games represent. It was once typical to interpret a game as a literal description of an idealized interaction, in which an assumption of perfect rationality appeared quite