Abstract:

Stereotypes are generalized beliefs about social groups that shape interpersonal interactions and intergroup relations. Traditional theories, however, tend to remain at the descriptive level, lacking an integrative computational explanation of how stereotypes are acquired and applied. Drawing on reinforcement learning and Bayesian theory, this review synthesizes the computational cognitive mechanisms of stereotyping along two core processes: social learning and social generalization. For social learning, we describe how Bayesian structure learning enables the inference of latent group categories, distinguish between experiential and linguistic pathways through which group-trait associations are established, and analyze how these associations are updated and consolidated through the interplay of prediction error, prior bias, and the explore-exploit dilemma. For social generalization, we examine how the brain categorizes novel individuals into known groups based on perceptual and functional cues, and how learned associative knowledge is retrieved. Future research could advance by broadening generalization cues, integrating multi-cue representations, and computationally simulating the process of linguistic transmission.

Key words: stereotype, social generalization, reinforcement learning, Bayesian theory, computational modeling