Social Visual Representations in Humans and Machines

Explore the fascinating world of social visual representations in humans and machines in this 27-minute lecture by Leyla Isik from Johns Hopkins University. Delve into neural face representations, disentangled representation learning, and the perception of social interactions in natural vision. Examine the encoding model performance, voxel preference, and variance partitioning in social vision research. Gain insights into the unique variance explained by sensory features, social features, and social interactions, as well as the role of the posterior superior temporal sulcus (pSTS) in perceiving social interactions.

Syllabus

Intro
Social vision
(Shared?) social visual representations
Neural face representations
Disentangled representation learning
Disentangled face representations
Social interaction selectivity?
Perceiving social interactions in the PSTS
Social interaction perception in natural vision
Encoding model performance
Voxel preference
Second preferred feature
Variance partitioning
Unique variance explained by sensory features
Unique variance explained by social features
Unique variance explained by social interactions