Amber X. Chen

Presentation Title

Seeing Faces Visually and Conceptually: Evidence from Single-neuron Data

Abstract

We examined how conceptual knowledge of celebrities, derived from a decade of U.S. web news using large language models, together with visual features from a deep neural network predicts neuronal responses to faces in the amygdala and hippocampus. Through representational similarity analysis (RSA), we found that these news-derived conceptual representations contributed to neuronal responses and social trait perceptions like competence and practicalness, particularly in familiar faces. These findings underscore the significant impact of external media on the neuronal and perceptual processing of faces.