SFB 504 discussion paper No. 00-11:

Local and Group Interaction in Prisoners' Dilemma Experiments—Imitate locally, think globally

Oliver Kirchkamp and Rosemarie Nagel


We investigate with the help of experiments a hypothesis that is common in the literature on evolution and local interaction (e.g. Axelrod 1984, May and Nowak 1992, Eshel, Samuelson, Shaked 1998), namely that players imitate successful neighbours. This hypothesis is particularly interesting in the context of spatial prisoners' dilemmas, since there it gives an elegant explanation for the survival of cooperation. A cluster of cooperative players obtains on average a higher payoff than a nearby cluster of noncooperative players. If imitation of successful players works, then members of the noncooperative cluster would start imitating the cooperative cluster which would, hence, grow larger and larger.

In our experiments we study different setups where players have different information available. For all setups we find that the amount of learning from successful neighbours is very small—much too small to allow the conclusions typically drawn in the literature. The dynamics of the game is mainly driven by reinforcement and reciprocity. We then introduce computerised players who form a cluster of always cooperating players, which should, if players would imitate successful neighbours, immediately start to grow. But again, the effect of imitation is too small to elicit cooperation in the population.

(JEL C72, C92, D74, D83, H41, R12)

You can download the paper as a PDF File