Modelling the face-to-face effect: Sensory population dynamics and active vision can contribute to perception of social context

Abstract

We introduce a novel dynamical model for visual attention based on stimulus induced population dynamics in an oscillatory medium, and apply this model to active perception of social content in still images. Making use of images from the newly emerging face-to-face paradigm in social developmental psychology, we show that this model generates patterns of eye movements that exhibit increased frequency of gaze shifts between actors in the social condition, as do infants at 16 months of age. The number of gaze shifts can inform useful levels of classification for the social content in the images, demonstrating a potential role for the dynamics of active perception in social cognition. This adaptive performance does not require any long term changes in structure or information storage. Our results further suggest a potential functional role in selective attention for the spiral wave activity recently observed in primary visual neo-cortex.