Recommendation extension to the iterated prisoner's dilemma game

Abstract

Optimism, pessimism, and realism are human dispositions that are used in everyday life to describe the typical behavior of a certain personality type. In this study, we investigate the effects of these dispositions on trust estimations from an agent's perspective. We present an extended version of the Iterated Prisoner's Dilemma game in which each player receives recommendations about other players to decide to play with unknown opponents. In our model, each agent keeps the trust values of the agents they have played before in their memory. If an agent has no information about an opponent, it consults neighbor agents of it. If neighbor agent knows that opponent, it gives a recommendation. An agent can be received more than one recommendations about the same opponent. How do agents evaluate recommendations? The evaluation process of recommendations varies according to the agent's disposition. We aim to observe which disposition shows the best performance in di erent forgetting mechanisms. We examine the e ect of receiving the recommendation. Our results show that it is more advantageous to receive recommendations, but that varies depending on memory usage. Realist agents perform better than other dispositions. We conclude that the best forgetting mechanisms for agents receiving recommendations vary by memory capacity.