Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Mapping Social Trait Judgments on the Human Cerebral Cortex Using Naturalistic, Dynamic Stimuli

Po-Yuan Alan Hsiao¹, Guo Jiahui¹; ¹University of Texas at Dallas

Presenter: Po-Yuan Alan Hsiao

Humans rapidly form impressions of trustworthiness, dominance, and attractiveness from faces in everyday life. While prior work has identified brain regions involved in face processing, it remains unclear how social trait judgments are represented across the human cerebral cortex under naturalistic conditions. Here, we modeled neural responses to dynamic face videos using a multi-trait encoding framework. Leveraging hyperaligned fMRI time series, we tested how well social trait ratings predict cortical responses. The cross-validated variance partitioning analysis revealed that attractiveness mainly explained the variance in occipital and ventral temporal cortex, dominance showed stronger contributions along the superior temporal sulcus, and trustworthiness was associated with lateral and frontal regions. We then compared these results with models based on trait values predicted by a state-of-the-art face-trained deep neural network (DNN). Although DNN-derived traits achieved comparable predictive performance, they engaged a distinct set of cortical regions. Together, these findings demonstrate that social trait representations are distributed across the cortex and differ systematically between human- and DNN-derived judgments.

Topic Area: Computational Models of Vision & Visual Cortex