Spatial evolution of automata in the prisoners' dilemma

We apply the idea of evolution to a spatial model. Prisoners' dilemmas or coordination games are played repeatedly within neighbourhoods where players do not optimise but instead copy successful strategies. Discriminative behaviour of players is introduced representing strategies as small automata, which can be in different states against different neighbours. Extensive simulations show that cooperation persists even in a stochastic environment, that players do not always coordinate on risk dominant equilibria in 2 x 2 coordination games, and that success among surviving strategies may differ. We present two analytical models that help understanding of these phenomena. (C) 2000 Elsevier Science B.V. All rights reserved.