Proceedings of the Annual Meeting of the Cognitive Science Society

Social biases can negatively impact our sense of belonging, achievement, and social relationships. However, it is unclear what inferential processes underlie how people detect biases. We propose that people infer social biases by positing prior beliefs to account for potential gaps between what someone observed (e.g., seeing you succeed on a challenging task) and how they responded (e.g., recommending you try something easier). We present a computational model formalizing this process, and validate it with two experiments. We find a strong quantitative fit between model predictions and participant judgments across a range of inferences, namely which prior belief the coach held (i.e., which team the coach thought the player was on, or which bias the coach has). This work bridges computational methods with social psychological research on social biases, by showing how mental state inferences contribute to our ability to rapidly detect biases.