Procedural Rationality in Repeated Games

This paper considers finitely repeated games played by procedurally rational players, who sample their available alternatives and use realized payoffs as a basis for strategy selection. The corresponding solution concept is that of (payoff) sampling equilibrium, which is a distribution over strategies that is self-replicating under the sampling procedure. Sampling equilibria are rest points of a disequilibrium dynamic process, and stability with respect to this process can be used as an equilibrium selection criterion. The structure of stable sampling equilibria in symmetric, finitely repeated games is characterized, and illustrated with applications to cooperation and coordination over time