Mid-level visual factors to predict and explain human aesthetic preferences for images: A multi-methods psychological approach

The main goal of this PhD is to contribute the psychological building blocks for a model to predict and explain human aesthetic preferences for images (natural scenes as well as artworks). The first phase of the project consists of three components, each with their dedicated methods: (1) a large-scale online study with large samples of images of everyday scenes and paintings and large samples of observers to obtain preference data, (2) focus groups of visual artists, professional photographers, and art school teachers to describe the visual features that make an image good or bad, and (3) a series of psychophysical studies with well-controlled stimuli allowing parametric variations of specific Gestalt-level image characteristics (e.g., strength of perceptual grouping and figure-ground organization, degree of order, clutter and ambiguity). While component (1) is needed to train an initial machine-learning model, components (2) and (3) are needed to enrich the initial machine-learning model with more specialized, more relevant mid-level features (which will be extracted by computer-vision algorithms). In the second phase of the project, the psychophysical approach will be extended to also assess the role of individual differences in low-level visual abilities (e.g., acuity, color vision), in mid-level perceptual styles (e.g., local-global processing) and in relevant personality traits (e.g., aesthetic sensitivity, need for closure) in relation to the aesthetic appreciation of images on a wider range of dimensions (e.g., beauty, pleasure, interest). In the third phase of the project, the computational models for universal, group-level and individual preferences will be tested and validated further by extensions to new stimulus sets (e.g., artificial images, artistic photographs), new aspects determining aesthetics (e.g., order, complexity), new measurements of perceptual and aesthetic processing (e.g., eye-movements), and new outcome measures (e.g., more refined aesthetic emotions and appraisals such as awe, wonder, thrills, the sublime).