Machine learning estimation of users' implicit and explicit aesthetic judgments of web-pages

The aesthetic appearance of websites can influence the perception of their usability, reliability, and trustworthiness. A majority of studies investigating the relationship between aesthetic features of web pages and their user perception consider only a limited number of web pages' visual features and focus exclusively on explicit aesthetic judgments. In this work, we aim to overcome the limitations of previous works by employing multiple visual features, as well as implicit aesthetic appreciation measures estimated by individuals' neurophysiological activity. Furthermore we aim to study the ability of machine learning models to predict the aesthetic judgments of webpages. We also investigate the differences between the prediction accuracy of explicit and implicit judgments of web pages. Our approach, based on the analysis of physiological signals, uses machine learning and neural network models to estimate users' implicit aesthetic pleasure. In our experiments, a group of young adults (N = 59, 33 females, Mean age = 21.52 years) assessed the aesthetic appeal of 100 web pages and 50 emotional pictures while we recorded their physiological activity. Our results demonstrate that machine learning models have a higher accuracy at predicting users' explicit judgments, as compared to implicit judgments.