Quantifying Visual Aesthetics Based on Processing Fluency Theory: Four Algorithmic Measures for Antecedents of Aesthetic Preferences

Humans are inherently evaluative beings whose affective system is predisposed to continuously and efficiently produce spontaneous preferences regarding surrounding stimuli. These preferences guide us through the physical world by indicating which stimuli to approach and which to avoid. Vision is the key sense to provide the necessary input for this superficial evaluative process that focuses on the exterior appeal of stimuli. To allow a better understanding of the link between visual input and evaluative response, the present article aims to identify visual antecedents of immediate aesthetic preferences. In particular, we propose a set of 4 algorithmic measures that automatically extract the low-level visual stimulus properties that are predictive of aesthetic preferences according to processing fluency theory: visual simplicity, visual symmetry, visual contrast, and visual self-similarity. In 3 studies, we demonstrate the impact, the relative importance, and the validity of these 4 visual characteristics for understanding aesthetic preferences. We examine the link between the 4 measures and aesthetic liking using a large pool of digital abstract artworks in Study 1. In Study 2, we experimentally replicate the findings of the first study to confirm their robustness. To generalize the results, Study 3 examines preferences for landscape photographs using secondary data of an online photo community. Furthermore, we built a database of 620 abstract digital artworks that can be used to benchmark future developments in measuring antecedents of visual aesthetics and that can serve as a pool of stimuli for experimental research on aesthetics.