Stereotypes at the intersection of perceivers, situations, and categories: analyzing stereotypes from storytelling using natural language processing

Stereotypes are often viewed as cognitive shortcuts that simplify expectations about how individuals in a group may behave. Consequently, prior work primarily examined stereotype content as a function of group labels, overlooking how stereotypes vary across perceivers, situations, and intersecting categories. To address this gap, we conducted two pre-registered studies with U.S. representative participants (N = 1,721) who freely wrote stories about how other people generally view 9 groups across 6 representative situations and 30 intersectional categories. We analyzed these stories using large language models. We found that although stereotypes are considered products of social learning about groups, they varied considerably across perceivers and situations. Intersectional stereotypes showed evidence for dominance effects of “Asian” and “unattractiveness”, potentially due to their lower representativeness in the U.S. population and uniform stereotypes across situations. These findings highlight the complex variability of stereotype content in naturalistic contexts, opening new doors to reducing intergroup biases.