Creativity and aesthetic evaluation of AI-generated artworks: bridging problems and methods from psychology to AI

This paper contributes to the debate on creativity, art, and artificial intelligence (AI) by integrating insights from cognitive psychology and empirical aesthetics into the field of AI, with the goal of inspiring novel empirical research. We focus on two main topics. First, we examine the indices used in psychology to operationalize creativity in closed-ended and open-ended tasks, with the aim not only of demonstrating the multidimensionality involved in defining creativity, but also of stimulating reflection on the benefits that might arise from developing a similar standard set of indices to test AI scoring models for assessing creativity (of both human and AI-generated responses). Second, we focus on the situation in which the creative products generated by AI are works of art, and on their aesthetic evaluation by non-expert human observers. Bridging the literature developed in psychology of art and empirical aesthetics with the literature on AI, a number of questions emerge, regarding the bias about the “expected style” of AI-generated art, and possible variables that play a role in aversion to AI-generated art. They all suggest possible future empirical research directions.