Stereotypic beliefs dominate reversal learning.

Counterstereotypic knowledge acquisition may undermine discriminatory thinking. However, researchers do not yet know how readily people procure stereotype-related information (i.e., stereotypic vs. counterstereotypic). Accordingly, we employed a probabilistic reversal-learning task in combination with computational modeling to explore the dynamics of person learning when targets exhibited both stereotypic and counterstereotypic behavior (i.e., food choices; women prefer salads, men favor burgers). We expected that stereotypic (vs. counterstereotypic) outcomes would dominate reversal learning, reflecting the potency of stereotype-based presumptions. Behavioral and computational analyses corroborated this prediction. First, participants learned stereotypic outcomes faster and more accurately than counterstereotypic outcomes. Second, participants were more likely to repeat stereotypic than counterstereotypic responses following positive reinforcement. Third, model estimates revealed that participants adjusted their choices following contingency reversals more slowly when newly reinforced outcomes were counterstereotypic rather than stereotypic. Collectively, these findings underscore the power of stereotypes and speak to the challenges of belief updating in daily life.