Beauty is in the efficient coding of the beholder

Abstract

Sexual ornaments are often assumed to be indicators of mate quality. Yet it remains poorly known how certain ornaments are chosen before any coevolutionary race makes them indicative. Perceptual biases have been proposed to play this role, but known biases are mostly restricted to a specific taxon, which precludes evaluating their general importance in sexual selection. Here we identify a potentially universal perceptual bias in mate choice. We used an algorithm that models the sparseness of the activity of simple cells in the primary visual cortex (or V1) of humans when coding images of female faces. Sparseness was found positively correlated with attractiveness as rated by men and explained up to 17% of variance in attractiveness. Because V1 is adapted to process signals from natural scenes, in general, not faces specifically, our results indicate that attractiveness for female faces is influenced by a visual bias. Sparseness and more generally efficient neural coding are ubiquitous, occurring in various animals and sensory modalities, suggesting that the influence of efficient coding on mate choice can be widespread in animals.

Biological bases of beauty revisited: The effect of symmetry, averageness, and sexual dimorphism on female facial attractiveness

“…more factors than what current theories have described, even if those theoretical factors do offer some predictive ability. Other factors like facial adiposity and body mass index [110,111] or representational sparseness [112] may contribute more to predictive power and explaining more variance [79].…”

Comparing theory-driven and data-driven attractiveness models using images of real women’s faces.

“…These measures are derived from algorithms that estimate the sparseness of neurons in the visual cortex required to represent a given face image and are thought to predict attractiveness because they index image-coding efficiency (Renoult et al, 2016). In other words, the sparseness of the activity of simple cells in V1 can be estimated for individual face images and is positively correlated with attractiveness (Renoult et al, 2016). In light of the above, we compared the performance of top-down models of facial attractiveness that included measured asymmetry, averageness, sexual dimorphism, BMI, and representational sparseness as predictors, to a bottomup model using shape and color principal components derived from a principal component analysis of our face stimuli.…”

“…This algorithm uses a feature dictionary based on Olshausen and Field's (1997) work on the properties of receptive fields in the visual cortex. Also following Renoult et al (2016), we defined sparseness of the encoding as the kurtosis of the estimated feature coefficients. Our MATLAB code for calculating sparseness is publicly available at https://osf.io/jurcq/.…”

Processing bias: extending sensory drive to include efficacy and efficiency in information processing

“…Neurons selective to locally oriented line segments (as found, for example, in the primary visual cortex of mammals or in the tecto-isthmic area in fishes) can be computationally modeled using simple Gabor filters [107], or by training a set of basis functions (each one modeling one neuron) to encode images of visual stimuli as sparsely as possible [33]. Then, efficiency is modeled by estimating the sparseness of the neuronal responses to a stimulus image [26,64,71,108]. Here, sparseness is measured as the proportion of neurons activated (i.e., with a non-zero response), or the kurtosis of the response distribution [109].…”