PD 2024 · Using Gibbs Sampling with People to characterize perceptual and aesthetic evaluations in multidimensional visual stimulus space

Eline Van Geert & Nori Jacoby
Perception Day 2024 · November 8, 2024 · Utrecht, The Netherlands

Aesthetic appreciation is intrinsically multidimensional: our aesthetic experiences are influenced by a variety of stimulus dimensions (e.g., colors, shapes, sizes). However, the majority of empirical aesthetics studies used either parametrically controlled unidimensional or non-parametrically controlled multidimensional stimuli, which prevented insight into the relative contribution of each stimulus dimension or any potential interactions between them to perceptual and aesthetic evaluations. To fill this gap, we combined two recent developments: the Order & Complexity Toolbox for Aesthetics (Van Geert, Bossens, & Wagemans, 2023), which generates multidimensional parametrically controlled stimuli, and Gibbs Sampling with People (Harrison et al., 2020), which efficiently characterizes subjective evaluations in multidimensional stimulus space. In five studies, we used this innovative technique to estimate multidimensional probability distributions for both perceptual (order and complexity) and aesthetic evaluations (pleasure and interest) in various visual multidimensional parametric stimulus spaces. In contrast to unidimensional or non-parametric approaches, our multidimensional parametric approach enabled us (a) to provide richer results and interpretations, (b) to assess the relative contributions of different stimulus aspects, and (c) to determine whether interactions between stimulus dimensions were important. This investigation taught us, for example, that when studying appreciation, both absolute and difference values must be considered, and that the different feature dimensions studied (i.e., colors, shapes, sizes) acted relatively independently on appreciation.