APPROXIMATIONS OF COOPERATIVE EQUILIBRIA IN MULTIPERSON PRISONERS-DILEMMA PLAYED BY CELLULAR AUTOMATA

Two-dimensional cellular automata are used to model Schelling's 'multiperson prisoners' dilemma' as a repeated N-person game where player payoffs are determined in a 'neighborhood', a field for a subgame of fewer than N players. Neighborhoods, which are also local information sets, overlap to form the full N-person society. A vanishingly small proportion of bounded-rationality rules for agent action results in a composite society with sufficient complexity to enforce the equivalent of cooperative play. All societies with less (qualitative) complexity can only support uniform defection.