Computational Processes Underlying Learning From Social Evaluative Feedback

Learning in a stochastic and constantly changing social environment is a difficult task. Social learning is crucial for individuals to adapt their behaviors by drawing insights from experiences, enabling them to effectively monitor and participate in such environments. A growing body of research underscores that learning from social feedback is biased in individuals with social anxiety. Reinforcement learning modeling can provide insight into how people learn from the outcomes of their actions, and use these outcomes to guide their future decisions. In the present study, computational modeling based on the influential Rescorla-Wagner model was adopted to explore how individuals adjust their predictions about prospective social evaluation by learning from social-evaluative feedback and its relationship with fear of negative evaluation (FNE). In the present study, 148 healthy adults participated in the SELF-Profile paradigm in which they were required to predict social evaluative feedback from peers by trial by trial. This feedback could be either social acceptance or social rejection as well as either congruent or incongruent with participants’ predictions. The computational modeling findings suggest that individuals adjusted their expectations more based on anticipated social feedback outcomes than unexpected ones. Furthermore, regression analysis results reveal no significant correlation between FNE and learning from social evaluative outcomes, irrespective of the accuracy of participants' predictions. This suggests that the individual's level of FNE does not seem to influence their learning from social evaluative outcomes. In sum, the present study emphasizes that individuals learn primarily by incorporating expected social feedback outcomes into their predictions.