Learning Reduces Ingroup Bias More with Perceived Losses than Gains Across Cultures

Impressions of individuals from their own and different groups are shaped by culture and learning. While previous evidence suggests stronger ingroup biases in collectivistic compared to individualistic cultures, the effect of such cultural differences on learning-related changes in intergroup impressions remains poorly understood. Here we use reinforcement learning models to investigate changes in intergroup impressions in gain and loss frames among individualistic and collectivistic cultures. Participants from a collectivistic culture (East Asian: Chinese) and an individualistic culture (Western: Swiss) repeatedly interacted with ingroup and outgroup individuals who increased (Gain frame) or reduced (Loss frame) their earnings. Importantly, the possible net outcomes were identical for the two frames. We collected impression ratings before and after these interactions. Results showed a higher ingroup identification score in the East Asian sample and an initial ingroup bias in impression ratings in both samples. Western participants learned to reduce this initial ingroup bias, but only in the loss frame, based on a learning signal (negative prediction error) that was generated if an ingroup individual reduced their earnings (i.e., in the Loss frame). East Asian participants showed the same learning mechanism, but only if they scored low on ingroup identification. Together, these results indicate that learning from negatively perceived ingroup interactions is a mechanism that can decrease ingroup bias in different cultures, modulated by individual ingroup identification.