Art made by artificial intelligence: The effect of authorship on aesthetic judgments.

Abstract

Human–robot interaction requires that competent robot partners have a multiplicity of human characteristics. Can we accept that these competencies extend to the artistic domain, where humans have always expressed their uniqueness as a species? This study investigated whether aesthetic judgments evoked by abstract artworks vary depending on whether the author was believed to be a human or a robot. Adult participants were asked to give beauty (BJ) and liking (LJ) judgments, where BJ reflected artistic competence and LJ a more idiosyncratic, affective facet associated with the stimuli. Aesthetic judgments were made in a blind-baseline condition, devoid of authorship information, and a primed condition, where authorship information (human or robot) was provided. A significant variation was found in LJ and BJ between the blind and primed conditions. The human-authored paintings received a higher liking rating in the primed than the blind and robot conditions; opposite, the robot-authored paintings received a lower beauty rating in the primed than the blind condition. These results suggest a resistance to accepting artificial intelligence in the production of art and highlight the emotional component associated with human art-making. Furthermore, aesthetic judgments were correlated with the attribution of mental states to a human and a robot to evaluate what mental characteristics are most related to aesthetic judgments. Both BJ and LJ of robot-authored art were significantly associated with the mental ability of creativity, thus pinpointing this skill as a marker of human art-making.

Algorithmic Bias in Image-Generating Artificial Intelligence: Prevalence and User Perceptions

“…At the same time, scientists have begun to examine how image-generating AI is perceived by recipients. Initial research focused on how AI source information alters the aesthetic appreciation of artworks (Di Dio et al, 2023;Messingschlager & Appel, 2023;Oksanen et al, 2023). We connected this relatively young line of research to scientific work on algorithmic bias, which has been a concern for many in the past years (e.g., Angwin et al, 2016;Courtland, 2018).…”