LEARNING TO FEAR SOCIAL INTERACTIONS: DYSREGULATED NEURAL MECHANISMS OF SOCIAL LEARNING IN ADOLESCENT SOCIAL ANXIETY

Abstract

Social anxiety (SA) disorder is prevalent, chronic, and impairs quality of life. Typical onset occurs in early adolescence, when social relationships become more salient and complex. Difficulty learning from these newly complex, important interactions may potentiate SA. SA is associated with suboptimal adaptive learning rates in non-social and uncertain contexts. However, the impact of social contexts on learning during social interactions with peers in SA remains unknown. Enhanced SA symptoms while anticipating social feedback is associated with dysregulated engagement of neural circuits implicated in salience and reward processing, which are critical hubs for learning. Despite this overlap, the neural mechanisms that support learning from social feedback remain relatively unexplored in SA. To study the influence of social context and feedback on learning in SA, we paired computational modeling with a novel social interaction fMRI task to determine the extent to which peer value as well as the valence and predictability of peer feedback modulate the neural bases of social learning about peers and their relation to adolescent SA. Fifty-nine youth (age 10-15yrs) with a range of SA engaged in real-time social interactions with purported peers while undergoing an fMRI scan. Youth with clinically relevant SA learn to fear social interactions by emphasizing unexpected negative feedback, and discounting unexpected positive feedback, in predictably nice and unpredictable social contexts to rapidly adjust learning. More severe SA was associated with decreased engagement of salience, reward and cognitive control regions in predictably nice social contexts, and only engaged cognitive control regions in unpredictably mean social contexts. Results elucidate which aspects of the social context contribute to learning to fear social interactions on a neural and behavioral level that potentiate SA, which can inform specific intervention targets.