Bounded learning and planning in public goods games.

Abstract

A previously developed agent model, based on bounded rational planning, is extended by introducing learning, with bounds on the memory of the agents. The exclusive impact of learning, especially in longer games, is investigated. Based on our results, we provide testable predictions for experiments on repeated public goods games (PGG) with synchronized actions. We observe that noise in player contributions can have a positive impact of group cooperation in PGG. We theoretically explain the experimental results on the impact of group size as well as mean per capita return (MPCR) on cooperation.