Alwin de Rooij
Artificial intelligence (AI) has become integral to many visual art practices. Audiences, however, may be biased against AI in visual art: knowing that AI was used in art creation can diminish aesthetic experience, independently of the objective qualities of an artwork. Whether this bias occurs structurally and across the systems involved in an aesthetic experience remains an open question. To address this, a meta-analysis was conducted. The results showed significant negative but small pooled effect sizes for data categorized under the sensory-motor (six studies, 16 effect sizes) and emotion-valuation (27 studies, 94 effect sizes) systems. For the knowledge-meaning system (26 studies, 49 effect sizes), a significant negative moderate pooled effect size was found. The magnitude of the bias was significantly moderated by age, suggesting a generational shift in how audiences think about using AI in art creation, and by art style, image source, and situatedness. Significant residual heterogeneity suggested that the bias against AI in visual art is unlikely to be a structural phenomenon. Furthering our understanding of bias against AI requires future research on invisible human and machine authorship and the potential of AI for creating rich aesthetic experiences. Herewith, this meta-analysis contributes novel insights into the bias against AI in visual art.