A Model-Based Investigating of the Biological Origin of Human Social Perception of Faces

AbstractHumans readily form social impressions of faces at a glance, whether assessing trustworthiness, attractiveness, or dominance. However, little is understood about how such computations are carried out neurally. Here, we leverage a computational model of human face perception to quantify and characterize the extent to which macaque monkey face patch neurons encode information relevant for social trait perception. Specifically, we use a social trait prediction model to estimate the social trait ratings for face stimuli viewed by monkeys during a neural recording experiment. We find that, while the monkey face patch neurons are linearly tuned to facial features different from those used by humans to make social judgments, the subspace spanned by the face patch neurons and the subspace spanned by the facial features supporting human social perception are highly overlapping. This result implies that the information present in the monkey face patch neurons are largely sufficient, after linear decoding, to support human social perception, thus shedding light on the biological origin of human social processing of faces.


Return to previous page