Markov Simulation of an Iterated Prisoners' Dilemma Experiment

We study cooperation and its development over the time in a repeated prisoner's dilemma experiment with unknown length and unknown continuation probability. Subjects are re matched with new partners several times. The study examines the influence of past decisions to future behavior. In the simulation we raise the question whether the experimentally observed time pattern of cooperation can be reconstructed with a simple Markov model of individual behavior. The model parameters are inferred from the experimental data. The resulting Markov model is used to simulate experimental outcomes. Our model indicates that a simple Markov model can be used as a reasonable description of transition probabilities for successive states of cooperation.