Title:Resolving social dilemmas with minimal reward transfer

Abstract:Multi-agent cooperation is an important topic, and is particularly challenging in mixed-motive situations where it does not pay to be nice to others. Consequently, self-interested agents often avoid collective behaviour, resulting in suboptimal outcomes for the group. In response, in this paper we introduce a metric to quantify the disparity between what is rational for individual agents and what is rational for the group, which we call the general self-interest level. This metric represents the maximum proportion of individual rewards that all agents can retain while ensuring that achieving social welfare optimum becomes a dominant strategy. By aligning the individual and group incentives, rational agents acting to maximise their own reward will simultaneously maximise the collective reward. As agents transfer their rewards to motivate others to consider their welfare, we diverge from traditional concepts of altruism or prosocial behaviours. The general self-interest level is a property of a game that is useful for assessing the propensity of players to cooperate and understanding how features of a game impact this. We illustrate the effectiveness of our method on several novel games representations of social dilemmas with arbitrary numbers of players.