Proceedings of the Annual Meeting of the Cognitive Science Society

There is widespread agreement across disciplines that morality has evolved to facilitate cooperation among agents with conflicting interests. Traditional approaches to the study of morality and cooperation either abstain from formal models (e.g., moral psychology in social psychology) or model cooperation without much concern for underlying psychological processes (e.g., evolutionary models of reciprocity, kin selection etc.). Computational cognitive scientists are starting to enrich classic computational frameworks (Bayesian inference, reinforcement learning, game theory, expected utility) with psychological mechanisms, traits, and constraints (Theory of Mind, metacognition, trust, joint planning, resource rationality) to model behavior, learning and evolutionary processes, and develop a finer understanding of the cognitive underpinnings of morality and cooperation. This progress is accelerated by current advances in modeling (probabilistic programming languages specialized for social cognition, large-language models, multi-agent RL) and computational efficiency (GPUs, numerical computation). Building formal models of moral cognition and cooperation that are both computationally precise and psychologically rich opens the door for cumulative progress in our understanding of their principles, origins, and functioning, and for the engineering of cooperative and ethical capabilities in artificial systems.