Breaking out : essays on partner selection in prisoner's dilemma games

Abstract

People seem to be much more cooperative than game theory predicts. Especially in situations that can be characterized as the prisoner’s dilemma. Cooperative outcomes are possible, only if the game is infinitely repeated. Does this mean that there is something wrong with game theory? Or does it mean that there is something wrong with the way the theory is applied? I believe the latter to be the case. In many real life applications people do not play the dilemma as such, even if the payoffs are revealed by the well- known payoff matrix. This is because the matrix only reflects the actual payoffs achieved once you are involved in a game, and choosing between the two strategies cooperate and defect (confess). The matrix does not reveal a further strategic choice, namely the option with whom to play. The role of partner selection will differ in the one-shot or repeated game. In the one-shot game partner selection signals willingness to play. In the repeated case, partner selection has an additional function: punishment. Instead of punishing non-cooperative behaviour by defection, a match with a defector can be broken apart. This might be the better disciplining device. As time passes, defectors risk to meet only fellow defectors, while cooperators do not forego cooperative opportunities by becoming defective themselves. That cooperation might be due to voluntary interactions is already captured in Adam Smith's term discipline of continuous dealings.