Dual neurocomputational mechanisms underlying intergroup bias in costly punishment

Group membership shapes social behavior, often giving rise to both preferential treatment of ingroups and differential responses to outgroups. However, the neurocomputational mechanisms through which group membership shapes costly punishment remain incompletely understood. Here, we addressed this gap by examining the intergroup bias in both reactive punishment (retaliation against unfairness) and proactive punishment (strategic sanctions against fairness), integrating a second-party punishment paradigm with computational modeling and brain imaging. Participants sanctioned outgroups more than ingroups for both unfair and fair splits, indicating robust intergroup bias across reactive and proactive punishment. Model-based analyses revealed that, in reactive punishment, intergroup bias was associated with attenuated aversion to disadvantageous inequity toward ingroups. This shift in inequity sensitivity was tracked by stronger medial prefrontal cortex responses to disadvantageous offers from ingroups, consistent with processes related to affiliative modulation of valuation. In contrast, in proactive punishment, intergroup bias was associated with increased preference for relative advantage over outgroups. This bias was accompanied by greater engagement of the dorsal anterior cingulate cortex and enhanced coupling with the ventral striatum for fair offers from outgroups, consistent with increased involvement of control-related processes in value-guided decision-making. Together, these findings reveal dissociable computational and neural signatures underlying intergroup bias in reactive and proactive punishment.