A generalization and partial reconceptualization of the theory of non-cooperative games

The question of rational behavior in dynamic non-cooperative games of incomplete information is examined. Two formulations of rationality are presented. In the first approach a player reduces his game to an ´adaptive´ optimal control problem. The second approach amounts to a generalization of Nash equilibrium theory. The author discusses the applicability of the two models in light of both philosophical and technical differences between them. Both approaches are shown to accommodate Me concept of ´strategic´ learning as an aspect of rational behavior. However the two differ in their ability to model or to prescribe the appealing notion of ´tacit cooperation´ familiar in prisoner´s dilemma games. In the last section an optimality criterion is proposed to help clarify the circumstances under which cooperation is rational.