Abstract
Contemporary aesthetics, influenced by the interactionist philosophical movement, posits that beauty results from the interaction between an observed object and the observer. Perceptual fluency theory, stemming from cognitive sciences, suggests that an object's attractiveness relies on fast, precise, and cognitively economical information processing. However, quantifying fluency and extending its application beyond humans is challenging. This PhD thesis explores this phenomenon in another primate species, the mandrill (Mandrillus sphinx), introduced in Chapter 1. In Chapters 2 and 3, a focus is done on one dimension of perceptual fluency, the prototypicality. The fluency theory proposes that prototypes are preferred due to their fluent processing. More specifically, femininity, considered as a particular form of prototypicality, is quantified using convolutional neural networks (CNNs) as models of visual perception. The femininity metric developed explains up to 26% of perceived femininity and 18% of perceived attractiveness in humans. In Chapter 2, the impact of adult female mandrills' femininity on their socio-sexual behaviors with their groupmates is examined. Results indicate that less feminine females attract more attention and sexual interest from their conspecifics, challenging the application of the fluency theory in this context. In Chapter 3, I study the causes and consequences of facial femininity variation among juvenile male mandrills. Results indicate that the most feminine males are born from young mothers and have fewer older sisters than more masculine males. However, this increased femininity does not change behaviors of these juvenile males. Chapter 4, a methodological study, broadens the perspective by developing a measure of prototypicality applicable to humans and generalizable to other species. A new metric, the "statistical typicality," is generated using CNNs and compared to another fluency measure, the sparsity. Statistical typicality only explains, however, 11% of the variance in attractiveness in humans, while sparsity accounts for 27% of this variance. Overall, this PhD thesis seeks to establish interactions among beauty, fluency, perception modeling using artificial intelligence tools, and social behaviors.