Social dilemmas: what if not everybody knows that everybody knows that everybody is rational?

摘要

In this paper social dilemmas are modelled as two-player games. In particular we modelthe Prisoner’s Dilemma, Chicken and Stag Hunt. When modelling these games weassume that players adapt their behaviour according to their experience and look foroutcomes that have proved to be satisfactory in the past. These ideas are investigated byconducting several experiments with an agent-based simulation model in which agentsuse a simple form of case-based reasoning. It is shown that cooperation can emerge fromthe interaction of selfish case-based reasoners. In determining how often cooperationoccurs, not only what Agents end up doing in any given situation is important, but alsothe process of learning what to do can crucially influence the final outcome. Agents’aspiration thresholds play an important role in that learning process. It is also found thatcase-based reasoners find it easier to cooperate in Chicken than in the Prisoner’sDilemma and Stag Hunt.