Moralized Stereotypes Are More Persistent Against Disconfirming Evidence: A Computational Reinforcement Learning Account

Moralized stereotypes are more persistent than non-moral stereotypes (2 experiments, N = 187). Participants completed a reinforcement learning task with players stereotyped as positive or negative and in moralized or non-moralized terms. Computational modeling showed that moralized stereotypes produced more potent effects on the formation and generalization of group-based preferences.