Learning reduces ingroup bias more with perceived losses than gains across cultures

Cultural background shapes intergroup impressions. While previous evidence suggests collectivistic cultures show stronger ingroup bias, cultural effects on impression formation processes remain unexplored. Here, we used reinforcement learning models to examine changes in intergroup impressions within gain and loss frames across individualistic (Western) and collectivistic (East Asian) cultures. Participants interacted with ingroup or outgroup individuals who increased (Gain) or decreased (Loss) their earnings, with identical net outcomes. Impression ratings were taken pre- and post-interaction. Results revealed higher ingroup identification in East Asians and initial ingroup bias in both cultures. Westerners learned to reduce this initial ingroup bias based on a learning signal (negative prediction error) generated if an ingroup individual reduced their earnings (i.e., Loss frame). East Asians showed the same learning mechanism, but only with low ingroup identification. Together, we show that learning from negatively perceived ingroup interactions can decrease ingroup bias across cultures, modulated by individual ingroup identification.