Researchers from Keio University in Tokyo have found that people exhibit a bias toward artworks created by neural networks. At the same time, many viewers struggle to distinguish pieces made by living artists from those produced by artificial intelligence. The study appeared in the scientific journal Perception.
The experiment involved 34 students who did not have formal training in evaluating art. They were shown 40 images in total, with half produced by real artists and the other half generated by the Disco Diffusion neural network.
In the first phase, participants viewed a sequence of both human-made and AI-generated artworks displayed on a screen for 20 seconds each, with one second between images. Afterward, they rated each artwork on criteria such as aesthetic appeal and perceived technical skill. The initial results revealed that the amount of time spent viewing each image did not depend on whether it was created by a human or by AI.
In a subsequent step, the participants were asked to identify which works originated from the neural network. Their accuracy stood at 68 percent for paintings by living artists and 43 percent for AI-generated illustrations. These figures indicated a meaningful confusion between human and machine authorship.
Eye-tracking data offered additional insight: when participants believed an image to be AI-produced, they tended to examine it for a shorter duration than pieces attributed to human creativity. This pattern suggests a subconscious tendency to scrutinize human-made art more closely.
Authors summarized the findings by noting an implicit bias toward AI art. Although participants could not reliably tell whether a painting was AI-generated and rated human- and AI-created works similarly in terms visible quality, they consistently assigned higher value to pieces categorized as human-made. The study’s participants spent more time contemplating those works, reinforcing the notion of a latent preference for human authorship.
One interpretation is that negative attitudes toward AI-generated art may persist at an unconscious level. Even as artificial intelligence can perform many creative tasks traditionally associated with humans, many viewers still regard true artistic creativity as a uniquely human trait.
These results align with a broader discussion about the role of human intention, expression, and originality in art, even as technology advances rapidly in the creative domain. They underscore the ongoing tension between human and machine contributions in visual culture and point to the need for ongoing dialogue about what constitutes true originality in art. Researchers emphasize that perception is shaped not only by technical execution but also by beliefs about authorship and the social meanings attached to art in contemporary society. .