Our memory has evolved not only to guide our future decisions but also to prepare us for interactions with the environment, including the social domain. Previous research has demonstrated that mutual cooperation relies on (source) memory. However, the hypothesis of a highly specialized cheater detection module, designed to identify uncooperative individuals, lacks empirical support. Instead, the literature suggests the flexible use of cross-domain inference and source memory mechanisms, which help to predict the future behavior of others by considering both situational and person-specific factors.
Nevertheless, typical research designs have used the same visual material in both the learning and test phases, raising questions about the validity of the results when more realistic materials are used. To address this, the lack of realism was manipulated in various experiments. I discuss the theoretical implications of these findings in relation to the mechanisms underlying source memory for social interaction partners.