Reciprocal Social Strategy in Social Repeated Games and Emergence of Social Norms

In an artificial society where agents repeatedly interact with one another, effective coordination among agents is generally a challenge. This is especially true when the participating agents are self-interested, and that there is no central authority to coordinate, and direct communication or negotiation are not possible. Recently, the problem was studied in a paper by Hao and Leung, where a new repeated game mechanism for modeling multi-agent interactions as well as a new reinforcement learning based agent learning method were proposed. In particular, the game mechanism differs from traditional repeated games in that the agents are anonymous, and the agents interact with randomly chosen opponents during each iteration. Their learning mechanism allows agents to coordinate without negotiations. The initial results had been promising. However, extended simulation also reveals that the outcomes are not stable in the long run in some cases, as the high level of cooperation is eventually not sustainable. In this work, we revisit the problem and propose a new learning mechanism as follows. First, we propose an enhanced Q-learning-based framework that allows the agents to better capture both the individual and social utilities that they have learned through observations. Second, we propose a new concept of "social attitude" for determining the action of the agents throughout the game. Simulation results reveal that this approach can achieve higher social utility, including close-to-optimal results in some scenarios, and more importantly, the results are sustainable with social norms emerging.