Promotion of robust cooperation among agents in complex networks by enhanced expectation-of-cooperation strategy

We present an interaction strategy with reinforcement learning to promote mutual cooperation among agents in complex networks. Networked computerized systems consisting of many agents that are delegates of social entities, such as companies and organizations, are being implemented due to advances in networking and computer technologies. Because the relationships among agents reflect the interaction structures of the corresponding social entities in the real world, social dilemma situations like the prisoner’s dilemma are often encountered. Thus, agents have to learn appropriate behaviors from the long term viewpoint to be able to function properly in the virtual society. The proposed interaction strategy is called the enhanced expectation-of-cooperation (EEoC) strategy and is an extension of our previously proposed strategy for improving robustness against defecting agents and for preventing exploitation by them. Experiments demonstrated that agents using the EEoC strategy can effectively distinguish cooperative neighboring agents from all-defecting (AllD) agents and thus can spread cooperation among EEoC agents and avoid being exploited by AllD agents. Examination of robustness against probabilistically defecting (ProbD) agents demonstrated that EEoC agents can spread and maintain mutual cooperation if the number of ProbD agents is not large. The EEoC strategy is thus simple and useful in actual computerized systems.