You Like It and We Can See It: Spatial and Temporal Rules to Explain the Sense of Liking

Neuroaesthetics is the science that studies how the human brain perceives and responds to beauty and art, such as paintings. By exploring neural processes, it reveals how visual elements, emotions, and cognitive interpretations interact to create aesthetic experiences. A promising approach to study this connection is the use of eye-tracking and pupillometry data to analyze ocular patterns during the observation of artworks. The aim of this work is to demonstrate that features extracted from eye-tracking and pupillometry data can be used to predict, through machine learning models, the subjective appreciation of an observed painting. Starting from a proprietary dataset of 3438 trials from 175 subjects exposed to art paintings in a real ecological context, in this paper we apply feature extraction and selection techniques with the aim of generating machine learning models that are able to predict the level of appreciation of an artwork by a human subject. As it turns out, our extracted model presents a high level of (balanced) average accuracy, greater than 0.75, which becomes 1.00 for specific pictures. In terms of number of subjects/trials, as well as in terms of applied techniques, there seems not to exist other comparable results in the current literature.