Group for Research in Decision Analysis

Learning to cooperate on climate policy facing uncertain damages

Sonja Köke University of Hamburg, Germany

Cooperation on many environmental problems like climate policy, forest fire or hurricane prevention can be characterized as a social dilemma in which cooperation over time may reduce the size of potential damages or the probability of their occurrence. The evolution of cooperative behavior in such dynamic but stochastic settings has received limited attention in the literature. In this paper, we investigate behavior in a repeated fourperson prisoner’s dilemma game with probabilistic damages. We compare the case where cooperation yields certain damage reduction to the situation where it reduces the damage size of a potentially occurring damage and to the case where it reduces the probability of an adverse event. The results show that cooperation levels are highest in the probability reduction and lowest in the certain treatment. Differences particularly occur over time. Our results indicate that a jointly experienced extreme event may trigger the evolution of cooperative behavior. We explain our results by proposing a model of positive and negative reinforcement learning.