Abstract
People can efficiently extract summary statistics from a set of objects-a process known as ensemble perception-including the average emotion of a facial crowd. Individuals with high social anxiety, however, tend to perceive facial crowds as more negative, reflecting a systematic perceptual bias. Because the ability to interpret group emotions is important to adaptive social functioning, this study examined whether a feedback-based training paradigm could reduce ensemble perception bias, and whether its effects varied by social anxiety. A total of 120 Korean university students were randomly assigned to either a training (n = 60) or control (n = 60) condition. Participants first completed baseline questionnaires assessing trait and state social anxiety and depression, then performed an ensemble perception task. In the training condition, participants rated the mean emotional intensity of a facial crowd on a continuous scale and received visual feedback displaying both their rating and the actual mean intensity. The control group performed the same task without feedback. Ensemble perception bias and absolute error were assessed before and after training. State social anxiety was reassessed after the task. Overall, feedback training significantly reduced bias and marginally reduced error but did not affect state social anxiety. Although formal moderation by trait social anxiety as a continuous variable was non-significant, exploratory subgroup analyses revealed that participants with high social anxiety showed notable reductions in bias, whereas no such changes were observed in the low social anxiety group. These effects remained after controlling for depressive symptoms. These findings suggest that individuals with high social anxiety may be particularly responsive to corrective feedback, enabling recalibration of their perceptual tendencies. Accordingly, feedback-based training may represent a promising approach for reducing perceptual bias in social perception among socially anxious individuals.