Seeing Social Interactions with Leyla Isik

The podcast explores social vision, specifically how humans recognize and understand social interactions from visual inputs. Leyla Isik presents evidence suggesting that the human visual system represents social interactions, highlighting the role of the superior temporal sulcus (STS) in processing dynamic social information. The discussion covers experiments using videos of interacting individuals and geometric shapes to demonstrate the brain's ability to recognize social cues. The podcast further investigates how computer vision models perform on social interaction tasks, revealing that standard models often fail, but models incorporating relational and dynamic information show promise. Isik also touches on ongoing research into multimodal integration of visual and auditory cues in understanding social interactions.