Social rejection hurts, but it can also be informative: Through experiences of acceptance and rejection, people identify partners interested in connecting with them and choose which ties to cement or to sever. What is it that people actually learn from rejection? In social interactions, people can learn from two kinds of information. First, people generally learn from rewarding outcomes, which may include concrete opportunities for interaction. Second, people track the "relational value" others ascribe to them-an internal model of how much others value them. Here, we used computational neuroimaging to dissociate these forms of learning. Participants repeatedly tried to match with others in a social game. Feedback revealed whether they successfully matched (a rewarding outcome) and how much the other person wanted to play with them (relational value). A Bayesian cognitive model revealed that participants chose partners who provided rewarding outcomes and partners who valued them. Whereas learning from outcomes was linked to brain regions involved in reward-based reinforcement, learning about relational value was linked to brain regions previously associated with social rejection. These findings identify precise computations underlying brain responses to rejection and support a neurocomputational model of social affiliation in which people build an internal model of relational value and learn from rewarding outcomes.
Keywords: computational modeling; fMRI; reinforcement learning; social rejection.