Dissociable computational mechanisms of generalized and social anxiety in social interaction

Maladaptive social behavior can be considered both a cause and effect of pathological anxiety. These difficulties may arise from both the cognitive demands of social interaction, such as managing uncertainty, and the distinctively social demands of interactions, such as managing social friction. Previous work has linked generalized anxiety (GA) to reduced learning from aversive events, particularly under uncertainty, and social anxiety (SA) to fear of negative evaluation. However, these distinct hypothesized effects have not been previously dissociated within a single experimental framework. In the current study, we used a novel task, the Asymmetric Social Exchange (ASE) game, to study how people reason under uncertainty about the hidden intentions of others and weigh the real personal costs and benefits of accepting interpersonal inequality, a major source of social friction. In a large sample of online participants (N = 723), we tested the hypothesis of a double dissociation between anxiety subtypes in behavior. Behavioral and computational modeling supported this: GA, not SA, was specifically related to perturbed learning under uncertainty through aversive overgeneralization. By contrast, SA, not GA, was related to increased sensitivity to social friction, increasing acceptance of unfair, disadvantageous outcomes. Our study combined a novel social decision-making task with computational modeling to identify dissociable cognitive mechanisms underlying GA and SA phenotypes.