Examining Social Reinforcement Learning Biases in Social Anxiety

Although social connectedness is critical to health and wellbeing, the 12% of Americans who experience social anxiety disorder in their lifetime often avoid social situations, which can lead to pervasive impairments. Aberrant social reinforcement learning, or differences in learning from positive and negative social feedback, may underlie many of the cognitive, emotional, and behavioral difficulties in social anxiety disorder. While reinforcement learning is well studied in other mental disorders and in the non-social domain, only a few studies have begun to probe aspects of social reinforcement learning. This dissertation serves as a more comprehensive, direct examination of this critical, understudied learning process. In it, I take a computational approach to investigate how biases in social reinforcement learning may contribute to social anxiety and whether these learning processes can be changed with a targeted, online intervention.
The three studies that comprise this dissertation are drawn from a larger data collection that included two laboratory sessions separated by five weeks, with half of the sample randomly assigned to an intervention between them. Studies 1 and 2 assess the extent to which socially anxiety is characterized by biased learning from social feedback in two domains relevant to social anxiety disorder: social interactions and social performance. Study 1 uses a social probabilistic learning task to assess how people use positive and negative social feedback to adjust their expectations of others. Study 2 examines the extent to which people use positive and negative social feedback to adjust their expectations of their own performance on a speech. Study 3 tests the degree to which the social reinforcement learning biases measured in Studies 1 and 2 are malleable through a brief online cognitive bias modification intervention. Each study is written as a standalone paper to facilitate submission for publication. This dissertation seeks to advance knowledge about social anxiety disorder by pinpointing specific biases in social reinforcement learning (Studies 1 and 2), which may improve our ability to develop targeted treatments (Study 3).