STOCHASTIC EVOLUTIONARY DYNAMICS IN THE REPEATED PRISONERS' DILEMMA

This paper explores boundedly rational behavior in the infinitely repeated prisoners' dilemma. Strategy choices are modeled as finite automata, and the evolution of populations through asexual reproduction with mutation is explored. Evolutionary success is based on how well the strategy does in an infinite game against randomly matched opponents, and mutation schemes based on a stylized model of these strategies are employed. The results demonstrate when either defecting or cooperating strategies tend to evolve, depending on how important the future is. Interestingly, when the strategies are allowed to become tolerant of defections, the population cycles between defecting and cooperating strategies.