Social Anxiety and Dynamic Social Reinforcement Learning in a Volatile Environment

Human behavior is driven by seeking rewards and avoiding punishments, so difficulties learning about rewards and punishments can lead to maladaptive behavior. In fact, aberrant monetary reinforcement learning has been implicated in depression, schizophrenia, and other disorders, and researchers are beginning to find evidence for altered social reinforcement learning in social anxiety disorder. But learning is not a static process. Because the social environment is uncertain and unstable, the present study examines how social anxiety affects probabilistic social reward and punishment learning and dynamic updating of learned probabilities in a volatile environment. N=222 online participants completed a novel version of Cyberball, a computerized ball-catching game, and questionnaires. Mixed-effects regression analysis was used to analyze throw patterns as a function of social anxiety symptoms. To assess the relative importance ascribed to new information in response to volatility, dynamic learning rates were calculated by applying Q-learning algorithms to overlapping sets of throws. The new Cyberball task appears to be a valid measure of social reward learning; throws to the computerized avatars varied according to their probabilities of social reward and punishment, and performance improved over time as participants became more certain of the probability of social reward. Moreover, higher social anxiety predicted fewer throws to the avatar that had been the punisher in the previous block, suggesting that social anxiety may be characterized by difficulty updating learned social probabilities such that socially anxious individuals miss the chance to learn that a once-punishing situation no longer poses a threat.