Kristian Lindgren

Paper #: 96-06-043

A number of evolutionary models based on the iterated Prisoner's Dilemma with noise are discussed. Different aspects of the evolutionary behavior are illustrated (i) by varying the trickiness of the game (iterated game, mistakes, misunderstandings, choice of payoff matrix), (ii) by introducing spatial dimensions, and (iii) by modifying the strategy space and the representation of strategies. One of the variations involves the finitely iterated game, that has a unique Nash equilibrium of only defecting strategies, and it is illustrated that when a spatial dimension is added, evolution usually avoids this state. The finite automaton representation of strategies is also revisited, and one model shows an evolution of a very robust error-correcting strategy for the Prisoner's Dilemma game.