Evolving Strategies for the Prisoner’s Dilemma

This paper investigates the use of Genetic Algorithms (GA’s) to evolve optimal strategies to the Prisoner’s Dilemma, a classic game theory problem. The problem was heavily studied in the 1980’s, but using more advanced computing techniques, this research extends the existing body of research. My hypothesis is that populations evolve to exhibit two traits: the ability to defend against defectors, and the ability to cooperate with other cooperators. Two successful, well studied strategies which embody these traits, Pavlov and Tit for Tat, are used as controls. Populations that do not possess these traits a priori are evolved and then compared to the performance of the controls. The results presented here strongly indicate that the hypothesized traits are present in all evolved populations. Key-Words: genetic algorithms, Prisoner’s dilemma, game theory, evolutionary computation