Social anxiety disorder (SAD) is associated with abnormal features in both emotion processing and social intention inference. However, there is a lack of research and clinical predictive models for the common mechanisms of emotion processing and social intention inference underlying social anxiety. The current project aims to use a combination of behavioral experiments, functional magnetic resonance imaging, and computational modeling to systematically examine the mechanisms of negative cognitive biases in SAD. Based on classic biological motion paradigms targeting emotion processing and social intention inference, combining the facial expressions, we aim to establish predictive models of SAD clinical symptoms based on multi-dimensional data. This project has the prospect of revealing the psychopathology underlying SAD, examining the association between behavioral and neuroimaging data underlying mental disorders, and promoting objective classification and prediction of mental disorders.