Investigating the negative bias towards artificial intelligence: effects of prior assignment of AI-authorship on the aesthetic appreciation of abstract paintings

Art, regarded as one of the last bulwarks of human prerogatives, is avalid model for investigating the relationship between humans andArtificial Intelligence (AI). Recent studies investigated the response tohuman-made vs. AI-made artworks, reporting evidence of either anegative bias towards the latter or no difference. Here, we investigatedwhether prior knowledge of authorship can influence the aestheticappreciation of two abstract paintings by manipulating the preassignment of human- vs. AI-authorship. In the ecological setting of anart fair, participants were asked to explicitly rate their aesthetic appreciation, while psychophysiological measure - electrodermal activity(EDA) and heart rate (HR) - were recorded during the observation of thetwo paintings. Presentation order was balanced among participants andartworks. Results show that when the human-declared painting wasshown as first, aesthetic judgement on the AI-declared painting werelower, while with the opposite presentation order judgements wereequal. Furthermore, although no modulation of HR was found, EDAactivation was always higher during the second presentation. In linewith literature, the results showed that looking at abstract artworksreduces the negative bias towards AI. However, the negative bias stillemerges when AI-artworks are implicitly compared to human-artworks.Implications are discussed.