A central tenet of the adaptive-memory framework is that memory has not merely evolved to help us relive the past but to prepare us for the future. In reciprocal social exchange, for instance, people must learn from previous experiences to approach cooperators and to avoid cheaters. In this sense, adaptive memory is inherently prospective. The present research is the first to test this central assumption of the adaptive-memory framework. In Experiments 1 and 2, participants played a Prisoner's Dilemma game and encountered cheating, cooperating, and neutral partners. The faces of these partners later reappeared during an event-based prospective-memory task. Participants showed better prospective-memory performance for cooperator and cheater faces than for neutral control faces. Multinomial processing-tree modeling served to separate the prospective component (remembering that an action needs to be performed) from the retrospective component (recognizing the target faces) of prospective memory. Superior prospective-memory performance for cooperator and cheater faces was attributable to a stronger prospective component, whereas the retrospective component remained unaffected. Experiment 3 showed that emotional descriptions of targets were ineffective in increasing prospective memory, suggesting that emotional valence alone cannot account for the prospective-memory advantage found in Experiments 1 and 2. The results suggest that cooperating with someone or being cheated by someone has a strong impact on future-oriented cognition. Enhanced prospective memory for cooperator and cheater faces may have an important function for maintaining reciprocal relationships and for avoiding cheaters. (PsycInfo Database Record (c) 2022 APA, all rights reserved).