From likely to likable: The role of statistical typicality in human social assessment of faces

Abstract

Humans readily form social impressions, such as attractiveness and trustworthiness, from a stranger’s facial features. Understanding the provenance of these impressions has clear scientific importance and societal implications. Motivated by the efficient coding hypothesis of brain representation, as well as Claude Shannon’s theoretical result that maximally efficient representational systems assign shorter codes to statistically more typical data (quantified as log likelihood), we suggest that social “liking” of faces increases with statistical typicality. Combining human behavioral data and computational modeling, we show that perceived attractiveness, trustworthiness, dominance, and valence of a face image linearly increase with its statistical typicality (log likelihood). We also show that statistical typicality can at least partially explain the role of symmetry in attractiveness perception. Additionally, by assuming that the brain focuses on a task-relevant subset of facial features and assessing log likelihood of a face using those features, our model can explain the “ugliness-in-averageness” effect found in social psychology, whereby otherwise attractive, intercategory faces diminish in attractiveness during a categorization task.

“Liking” as an early and editable draft of long-run affective value

“…The worth of this change turns out to be closely related to the efficiency with which the input can currently be processed (assuming that the plasticity referred to above is modest). This notion of efficiency is the basis of a second, and traditionally competing, popular class of theories for aesthetic value [ 78 – 80 ]. Thus a potential-based shaping theory of “liking” unifies these two concepts of aesthetic value.…”

'Liking' as a First Draft of the Affective Future

“…The first component then realizes one popular form of aesthetic value associated with learning progress or prediction error (Schmidhuber, 2010) -rewarding the change of the sensory system associated with the input. If changes to the sensory system associated with any individual input are modest, the second component amounts to rewarding stimuli that are likely under the generative model -a factor closely related to the fluency with which the input is processed, which is a second, and traditionally competing, popular theory for aesthetic value (Reber et al, 2004;Ryali et al, 2020;Van de Cruys and Wagemans, 2011). Thus a potential-based shaping theory of 'liking' unifies these two concepts of aesthetic value.…”

Preferences need inferences: Learning, valuation, and curiosity in aesthetic experience

“…Indeed, an unexpected inability to perform one's habits (prediction error) can arouse intensely negative emotions. But in isolation, habits do not normally come with intensely positive appreciations, similar to how the most frequent or predictable stimuli are usually not liked much, except in special cases like faces (Ryali et al, 2020) or wallpapers. The preference for regular wallpapers and average faces may have more to do with the fact that we just want those to be the clean canvases for the actually interesting matter: Social expressions or interactions.…”