Beauty is in the efficient coding of the beholder

Sparsity in an artificial neural network predicts beauty: Towards a model of processing-based aesthetics

Generations of scientists have pursued the goal of defining beauty. While early scientists initially focused on objective criteria of beauty ('feature-based aesthetics'), philosophers and artists alike have since proposed that beauty arises from the interaction between the object and the individual who perceives it. The aesthetic theory of fluency formalizes this idea of interaction by proposing that beauty is determined by the efficiency of information processing in the perceiver's brain ('processing-based aesthetics'), and that efficient processing induces a positive aesthetic experience. The theory is supported by numerous psychological results, however, to date there is no quantitative predictive model to test it on a large scale. In this work, we propose to leverage the capacity of deep convolutional neural networks (DCNN) to model the processing of information in the brain by studying the link between beauty and neuronal sparsity, a measure of information processing efficiency. Whether analyzing pictures of faces, figurative or abstract art paintings, neuronal sparsity explains up to 28% of variance in beauty scores, and up to 47% when combined with a feature-based metric. However, we also found that sparsity is either positively or negatively correlated with beauty across the multiple layers of the DCNN. Our quantitative model stresses the importance of considering how information is processed, in addition to the content of that information, when predicting beauty, but also suggests an unexpectedly complex relationship between fluency and beauty.

Like Father, Like Child? Paternal Age at Birth and Offspring’s Facial Asymmetry and Distinctiveness

Paternal age at birth, a potential proxy of mutation load transmitted to the offspring, has previously been related to the offspring’s health, biological condition and reproductive potential. As higher facial asymmetry and distinctiveness serve as putative markers of the lower genetic quality of an individual, we hypothesize that more advanced paternal age at birth will be related to children’s higher levels of facial asymmetry and distinctiveness. To the best of our knowledge, this is the first study evaluating the link between paternal age at birth and facial asymmetry in offspring. Based on archived photographs of 159 children born within 47 Polish families, we have conducted facial geometric morphometric measurements and calculated the levels of facial asymmetry and distinctiveness. The relationship between paternal age at birth and the offspring’s facial features was explored with the use of Bayesian Linear Mixed-effects Models, controlling for sex, age and birth order of the offspring, and maternal age at child’s birth. No associations between paternal age at birth and facial asymmetry or distinctiveness in children were found. The lack of such a relationship might be a result of the potentially insufficient influence of newly accumulated paternal mutations affecting the offspring’s phenotype or higher importance of maternal (prenatal) and postnatal environments in shaping facial features.

Darwin, sexual selection, and the brain

One hundred fifty years ago Darwin published The Descent of Man, and Selection in Relation to Sex, in which he presented his theory of sexual selection with its emphasis on sexual beauty. However, it was not until 50 y ago that there was a renewed interest in Darwin’s theory in general, and specifically the potency of mate choice. Darwin suggested that in many cases female preferences for elaborately ornamented males derived from a female’s taste for the beautiful, the notion that females were attracted to sexual beauty for its own sake. Initially, female mate choice attracted the interest of behavioral ecologists focusing on the fitness advantages accrued through mate choice. Subsequent studies focused on sensory ecology and signal design, often showing how sensory end organs influenced the types of traits females found attractive. Eventually, investigations of neural circuits, neurogenetics, and neurochemistry uncovered a more complete scaffolding underlying sexual attraction. More recently, research inspired by human studies in psychophysics, behavioral economics, and neuroaesthetics have provided some notion of its higher-order mechanisms. In this paper, I review progress in our understanding of Darwin’s conjecture of “a taste for the beautiful” by considering research from these diverse fields that have conspired to provide unparalleled insight into the chooser’s mate choices.

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

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.

Aesthetic Perception of Line Patterns: Effect of Edge-Orientation Entropy and Curvilinear Shape

Curvilinearity is a perceptual feature that robustly predicts preference ratings for a variety of visual stimuli. The predictive effect of curved/angular shape overlaps, to a large degree, with regularities in second-order edge-orientation entropy, which captures how independent edge orientations are distributed across an image. For some complex line patterns, edge-orientation entropy is actually a better predictor for what human observers like than curved/angular shape. The present work was designed to disentangle the role of the two features in artificial patterns that consisted of either curved or angular line elements. We systematically varied these patterns across two more dimensions, edge-orientation entropy and the number of lines. Eighty-three participants rated the stimuli along three aesthetic dimensions (pleasing, harmonious, and complex). Results showed that curved/angular shape was a stronger predictor for ratings of pleasing and harmonious if the stimuli consisted of a few lines that were clearly discernible. By contrast, edge-orientation entropy was a stronger predictor for the ratings if the stimuli showed many lines, which merged into a texture. No such differences were obtained for complexity ratings. Our findings are in line with results from neurophysiological studies that the processing of shape and texture, respectively, is mediated by different cortical mechanisms.

Sexual signaling pattern correlates with habitat pattern in visually ornamented fishes

Sexual signal design is an evolutionary puzzle that has been partially solved by the hypothesis of sensory drive. Framed in signal detection theory, sensory drive posits that the attractiveness of a signal depends on its detectability, measured as contrast with the background. Yet, cognitive scientists have shown that humans prefer images that match the spatial statistics of natural scenes. The explanation is framed in information theory, whereby attractiveness is determined by the efficiency of information processing. Here, we apply this framework to animals, using Fourier analysis to compare the spatial statistics of body patterning in ten species of darters (Etheostoma spp.) with those of their respective habitats. We find a significant correlation between the spatial statistics of darter patterns and those of their habitats for males, but not for females. Our results support a sensory drive hypothesis that recognizes efficient information processing as a driving force in signal evolution.

Global Image Properties Predict Ratings of Affective Pictures

Affective pictures are widely used in studies of human emotions. The objects or scenes shown in affective pictures play a pivotal role in eliciting particular emotions. However, affective processing can also be mediated by low-level perceptual features, such as local brightness contrast, color or the spatial frequency profile. In the present study, we asked whether image properties that reflect global image structure and image composition affect the rating of affective pictures. We focused on 13 global image properties that were previously associated with the esthetic evaluation of visual stimuli, and determined their predictive power for the ratings of five affective picture datasets (IAPS, GAPED, NAPS, DIRTI, and OASIS). First, we used an SVM-RBF classifier to predict high and low ratings for valence and arousal, respectively, and achieved a classification accuracy of 58–76% in this binary decision task. Second, a multiple linear regression analysis revealed that the individual image properties account for between 6 and 20% of the variance in the subjective ratings for valence and arousal. The predictive power of the image properties varies for the different datasets and type of ratings. Ratings tend to share similar sets of predictors if they correlate positively with each other. In conclusion, we obtained evidence from non-linear and linear analyses that affective pictures evoke emotions not only by what they show, but they also differ by how they show it. Whether the human visual system actually uses these perceptive cues for emotional processing remains to be investigated.

Endless forms of sexual selection

In recent years, the field of sexual selection has exploded, with advances in theoretical and empirical research complementing each other in exciting ways. This perspective piece is the product of a “stock-taking” workshop on sexual selection and sexual conflict. Our aim is to identify and deliberate on outstanding questions and to stimulate discussion rather than provide a comprehensive overview of the entire field. These questions are organized into four thematic sections we deem essential to the field. First we focus on the evolution of mate choice and mating systems. Variation in mate quality can generate both competition and choice in the opposite sex, with implications for the evolution of mating systems. Limitations on mate choice may dictate the importance of direct vs. indirect benefits in mating decisions and consequently, mating systems, especially with regard to polyandry. Second, we focus on how sender and receiver mechanisms shape signal design. Mediation of honest signal content likely depends on integration of temporally variable social and physiological costs that are challenging to measure. We view the neuroethology of sensory and cognitive receiver biases as the main key to signal form and the ‘aesthetic sense’ proposed by Darwin. Since a receiver bias is sufficient to both initiate and drive ornament or armament exaggeration, without a genetically correlated or even coevolving receiver, this may be the appropriate ‘null model’ of sexual selection. Thirdly, we focus on the genetic architecture of sexually selected traits. Despite advances in modern molecular techniques, the number and identity of genes underlying performance, display and secondary sexual traits remains largely unknown. In-depth investigations into the genetic basis of sexual dimorphism in the context of long-term field studies will reveal constraints and trajectories of sexually selected trait evolution. Finally, we focus on sexual selection and conflict as drivers of speciation. Population divergence and speciation are often influenced by an interplay between sexual and natural selection. The extent to which sexual selection promotes or counteracts population divergence may vary depending on the genetic architecture of traits as well as the covariance between mating competition and local adaptation. Additionally, post-copulatory processes, such as selection against heterospecific sperm, may influence the importance of sexual selection in speciation. We propose that efforts to resolve these four themes can catalyze conceptual progress in the field of sexual selection, and we offer potential avenues of research to advance this progress.

Evolution of sexual cooperation from sexual conflict

In many species that form pair bonds, males display to their mate after pair formation. These displays elevate the female's investment into the brood. This is a form of cooperation because without the display, female investment is reduced to levels that are suboptimal for both sexes. The presence of such displays is paradoxical as in their absence the male should be able to invest extra resources directly into offspring, to the benefit of both sexes. We consider that the origin of these displays lies in the exploitation of preexisting perceptual biases which increase female investment beyond that which is optimal for her, initially resulting in a sexual conflict. We use a combined population genetic and quantitative genetic model to show how this conflict becomes resolved into sexual cooperation. A cooperative outcome is most likely when perceptual biases are under selection pressures in other contexts (e.g., detection of predators, prey, or conspecifics), but this is not required. Cooperation between pair members can regularly evolve even when this provides no net advantage to the pair and when the display itself reduces a male's contributions to raising the brood. The findings account for many interactions between the sexes that have been difficult to explain in the context of sexual selection.

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

Communication signals often comprise an array of colours, lines, spots, notes or odours that are arranged in complex patterns, melodies or blends. Receiver perception is assumed to influence preference and thus the evolution of signal design, but evolutionary biologists still struggle to understand how perception, preference and signal design are mechanistically linked. In parallel, the field of empirical aesthetics aims to understand why people like some designs more than others. The model of processing bias discussed here is rooted in empirical aesthetics, which posits that preferences are influenced by the emotional system as it monitors the dynamics of information processing and that attractive signals have effective designs that maximize information transmission, efficient designs that allow information processing at low metabolic cost, or both. We refer to the causal link between preference and the emotionally rewarding experience of effective and efficient information processing as the processing bias, and we apply it to the evolutionary model of sensory drive. A sensory drive model that incorporates processing bias hypothesizes a causal chain of relationships between the environment, perception, pleasure, preference and ultimately the evolution of signal design, both simple and complex.

Biological Bases of Beauty Revisited: The Effect of Symmetry, Averageness, and Sexual Dimorphism on Female Facial Attractiveness

The factors influencing human female facial attractiveness—symmetry, averageness, and sexual dimorphism—have been extensively studied. However, recent studies, using improved methodologies, have called into question their evolutionary utility and links with life history. The current studies use a range of approaches to quantify how important these factors actually are in perceiving attractiveness, through the use of novel statistical analyses and by addressing methodological weaknesses in the literature. Study One examines how manipulations of symmetry, averageness, femininity, and masculinity affect attractiveness using a two-alternative forced choice task, finding that increased masculinity and also femininity decrease attractiveness, compared to unmanipulated faces. Symmetry and averageness yielded a small and large effect, respectively. Study Two utilises a naturalistic ratings paradigm, finding similar effects of averageness and masculinity as Study One but no effects of symmetry and femininity on attractiveness. Study Three applies geometric face measurements of the factors and a random forest machine learning algorithm to predict perceived attractiveness, finding that shape averageness, dimorphism, and skin texture symmetry are useful features capable of relatively accurate predictions, while shape symmetry is uninformative. However, the factors do not explain as much variance in attractiveness as the literature suggests. The implications for future research on attractiveness are discussed.

Neural correlates of novelty and appropriateness processing in externally induced constraint relaxation

Novelty and appropriateness are considered the two fundamental features of creative thinking, including insight problem solving, which can be performed through chunk decomposition and constraint relaxation. Based on a previous study that separated the neural bases of novelty and appropriateness in chunk decomposition, in this study, we used event-related functional magnetic resonance imaging (fMRI) to further dissociate these mechanisms in constraint relaxation. Participants were guided to mentally represent the method of problem solving according to the externally provided solutions that were elaborately prepared in advance and systematically varied in their novelty and appropriateness for the given problem situation. The results showed that novelty processing was completed by the temporoparietal junction (TPJ) and regions in the executive system (dorsolateral prefrontal cortex [DLPFC]), whereas appropriateness processing was completed by the TPJ and regions in the episodic memory (hippocampus), emotion (amygdala), and reward systems (orbitofrontal cortex [OFC]). These results likely indicate that appropriateness processing can result in a more memorable and richer experience than novelty processing in constraint relaxation. The shared and distinct neural mechanisms of the features of novelty and appropriateness in constraint relaxation are discussed, enriching the representation of the change theory of insight.

Computational and Experimental Approaches to Visual Aesthetics

Aesthetics has been the subject of long-standing debates by philosophers and psychologists alike. In psychology, it is generally agreed that aesthetic experience results from an interaction between perception, cognition, and emotion. By experimental means, this triad has been studied in the field of experimental aesthetics, which aims to gain a better understanding of how aesthetic experience relates to fundamental principles of human visual perception and brain processes. Recently, researchers in computer vision have also gained interest in the topic, giving rise to the field of computational aesthetics. With computing hardware and methodology developing at a high pace, the modeling of perceptually relevant aspect of aesthetic stimuli has a huge potential. In this review, we present an overview of recent developments in computational aesthetics and how they relate to experimental studies. In the first part, we cover topics such as the prediction of ratings, style and artist identification as well as computational methods in art history, such as the detection of influences among artists or forgeries. We also describe currently used computational algorithms, such as classifiers and deep neural networks. In the second part, we summarize results from the field of experimental aesthetics and cover several isolated image properties that are believed to have a effect on the aesthetic appeal of visual stimuli. Their relation to each other and to findings from computational aesthetics are discussed. Moreover, we compare the strategies in the two fields of research and suggest that both fields would greatly profit from a joined research effort. We hope to encourage researchers from both disciplines to work more closely together in order to understand visual aesthetics from an integrated point of view.

Statistical Image Properties in Large Subsets of Traditional Art, Bad Art, and Abstract Art

Several statistical image properties have been associated with large subsets of traditional visual artworks. Here, we investigate some of these properties in three categories of art that differ in artistic claim and prestige: (1) Traditional art of different cultural origin from established museums and art collections (oil paintings and graphic art of Western provenance, Islamic book illustration and Chinese paintings), (2) Bad Art from two museums that collect contemporary artworks of lesser importance (© Museum Of Bad Art [MOBA], Somerville, and Official Bad Art Museum of Art [OBAMA], Seattle), and (3) twentieth century abstract art of Western provenance from two prestigious museums (Tate Gallery and Kunstsammlung Nordrhein-Westfalen). We measured the following four statistical image properties: the fractal dimension (a measure relating to subjective complexity); self-similarity (a measure of how much the sections of an image resemble the image as a whole), 1st-order entropy of edge orientations (a measure of how uniformly different orientations are represented in an image); and 2nd-order entropy of edge orientations (a measure of how independent edge orientations are across an image). As shown previously, traditional artworks of different styles share similar values for these measures. The values for Bad Art and twentieth century abstract art show a considerable overlap with those of traditional art, but we also identified numerous examples of Bad Art and abstract art that deviate from traditional art. By measuring statistical image properties, we quantify such differences in image composition for the first time.

Sensory Drive, Color, and Color Vision

Colors often appear to differ in arbitrary ways among related species. However, a fraction of color diversity may be explained because some signals are more easily perceived in one environment rather than another. Models show that not only signals but also the perception of signals should regularly evolve in response to different environments, whether these primarily involve detection of conspecifics or detection of predators and prey. Thus, a deeper understanding of how perception of color correlates with environmental attributes should help generate more predictive models of color divergence. Here, I briefly review our understanding of color vision in vertebrates. Then I focus on opsin spectral tuning and opsin expression, two traits involved in color perception that have become amenable to study. I ask how opsin tuning is correlated with ecological differences, notably the light environment, and how this potentially affects perception of conspecific colors. Although opsin tuning appears to evolve slowly, opsin expression levels are more evolutionarily labile but have been difficult to connect to color perception. The challenge going forward will be to identify how physiological differences involved in color vision, such as opsin expression levels, translate into perceptual differences, the selection pressures that have driven those differences, and ultimately how this may drive evolution of conspecific colors.

Global Sensory Qualities and Aesthetic Experience in Music

A well-known tradition in the study of visual aesthetics holds that the experience of visual beauty is grounded in global computational or statistical properties of the stimulus, for example, scale-invariant Fourier spectrum or self-similarity. Some approaches rely on neural mechanisms, such as efficient computation, processing fluency, or the responsiveness of the cells in the primary visual cortex. These proposals are united by the fact that the contributing factors are hypothesized to be global (i.e., they concern the percept as a whole), formal or non-conceptual (i.e., they concern form instead of content), computational and/or statistical, and based on relatively low-level sensory properties. Here we consider that the study of aesthetic responses to music could benefit from the same approach. Thus, along with local features such as pitch, tuning, consonance/dissonance, harmony, timbre, or beat, also global sonic properties could be viewed as contributing toward creating an aesthetic musical experience. Several such properties are discussed and their neural implementation is reviewed in the light of recent advances in neuroaesthetics.