Human bias in evaluating the aesthetic quality of AI artworks and its relation to moral intention

Abstract

This study investigated the influence of perceived moral intentions on the aesthetic evaluation of artificial intelligence (AI) and human artworks. Participants were asked to rate the attractiveness of AI and human pictorial artworks, whose moral intention was manipulated (low morality, neutral, high morality). The creator of each artwork (AI or a human artist) was disclosed to half the participants (label group) but not to the rest (no-label group). The no-label group participated in an authorship estimation task after the aesthetic evaluation task. The results showed that artworks with higher perceived morality were rated as more attractive in both groups. Furthermore, the authorship estimation results with the no-label group suggested that works with higher moral contexts were presumed to be human creations, whereas those with lower moral contexts were more likely to be associated with AI works; this suggests that AI may be considered as an entity with insufficient moral autonomy.