Max Planck Institute for Empirical Aesthetics, ArtLab Foyer
The Westend Lectures on Brain and Cognition with Alexander Todorov:
Modeling Social Perception of Faces
People form instantaneous impressions from facial appearance, agree on these impressions, and often act on these impressions. But what determines the content of impressions? In the last 10 years, we have introduced data-driven computational methods that allow us to visualize the configurations of facial features that lead to specific social inferences (e.g., trustworthiness). Using these methods, we can model evaluation of faces on any dimension and identify the perceptual basis of this evaluation. The methods generalize to face responses other than explicit social judgments. We can visualize, for example, the face configurations that emerge faster in consciousness. However, perceptual determinants are insufficient to account for the full content of impressions. One’s learning history of faces is another important determinant of impressions. These types of learning include statistical learning of the distribution (and typicality) of faces and idiosyncratic learning based on familiar faces. The determinants of impressions account for their compelling force without the need to posit that they are accurate. In fact, the evidence for accuracy of first impressions is extremely weak.