Revealing dichotomous prior biases in social anxiety through a social prism model

Abstract

Social anxiety is associated with atypical interpretations of social situations, but the computational origins of these biases remain unclear. This study introduces the Social Prism Model, a Bayesian cognitive framework that decomposes social scenes into distinct perceptual and inferential cues. Across eight social-valence judgment experiments involving 541 participants, hierarchical drift-diffusion modeling indicated that social-anxiety-related differences were driven primarily by cue-dependent prior expectations rather than altered sensory evidence accumulation. Individuals with higher social anxiety showed over-negative priors for facial and ambiguous cues, but over-positive priors for animacy, emotion, social interaction, and bodily cues. Bayesian simulations further showed that variation in prior expectations could reproduce biased social judgments, supporting a mechanistic account in which heterogeneous, cue-specific priors shape social cognition in social anxiety.