Modeling Social Cue Integration in Aesthetic Evaluations

Abstract

How do people form aesthetic evaluations in light of information about others’ evaluations? How can we understand this process at the computational level? Here, we explore these questions using a multi-generational online experiment in which participants rate artworks given information about previous participants’ ratings. We compare this data to task performance by a Bayesian model of the task, adapted from a model shown to accurately describe multi-sensory cue combination. Our study shows that evaluations tend to correlate with social information, providing evidence for social influence in aesthetic evaluations. Furthermore, we find that to appropriately generalize the Bayesian model used in multi-sensory cue combination to social and individual cue combination, we must consider a bias towards individual evaluations in addition to the noisiness of each signal. Future directions for refining and applying this model are discussed.