Artwork has long served as a method of creative expression for humans. Techniques and styles evolved to adapt to cultures and societies, and with the advent of digital technology, contemporary art adopted new meanings for artists.
Central to this technology, artificial intelligence (AI) harnesses machine learning algorithms to create novel images from existing ones. Just as digital cameras revolutionized portrait painting, artificial intelligence sparks a paradigm shift in the world of artwork. However, not all support this shift. While many find AI-generated artwork pleasing and beautiful, others criticize its authenticity and claim it lacks intrinsic value.
To better understand the debate between AI-generated and human-made art, ThinkAlike Laboratories, LLC designed an experiment to test people’s brain reactions to AI-generated and human-created artwork1. In this experiment, participants viewed two sets of images: one of AI-generated artwork and another of human-created artwork. However, we were not always honest about whether the artwork was AI-generated or human-created.
In half of the trials, the participants were correctly told that AI-generated art was AI-generated and that human-created art was human-created. For the other half of the trials, participants were incorrectly told that the AI-generated art was human-created and that human-created art was AI-generated. We then asked participants whether they liked, disliked, or were indifferent to the artwork they saw.
Participants clearly preferred AI-generated art over human-created pieces (Image 1). Remarkably, the top twelve most-liked artworks were AI-generated, with 81% of participants expressing a preference for the top-rated piece. In contrast, the ten least-liked artworks were human-created, with only 11% of participants saying they liked the lowest-rated piece. These results suggest that humans are positively biased towards AI-generated art when they do not know whether it is AI-generated.
We also used electroencephalography (EEG) to investigate the activity of a neural signature known as P2, which indicates whether a person perceives a stimulus as more or less beautiful2. A lower P2 means that the stimulus is perceived more positively. We found a lower P2 amplitude in response to AI-generated than human-created art, which suggests that participants tended to perceive AI-generated art as more positive and beautiful than human-created art.
However, we found surprising results with a different neural signature, the frontal late positive potential (fLPP). The fLPP is found on both the brain's left and right hemispheres. Past research found that a greater amplitude difference between the two sides of the brain (right hemisphere minus left hemisphere) tends to indicate a more negative stimulus3. When we measured participants’ fLPP, we found a greater amplitude difference when participants were told that artwork was AI-generated than when they were told it was human-created. This indicates that, although the participants preferred AI-generated artwork overall, they were biased against AI-generated artwork when told that a machine created a particular piece.
The results of this experiment reveal two important points. First, people can prefer AI-generated artwork over human-created artwork when unaware of whether a human made it. Second, people tend to be biased against AI-generated artwork when told it was created by artificial intelligence, even when they may prefer it.
Moreover, these results reveal how neuroscience can be used to reveal people’s preferences. While self-reporting surveys can reveal people’s preferences for a particular product, their responses can be prone to biases due to the nature of self-reporting. By contrast, a neural study can reveal people’s preferences more objectively by measuring pre-conscious responses to stimuli. This allows us and other researchers to better understand how people respond to different products.
These findings open the door to a fresh perspective on artificial intelligence. While people often express skepticism or even resistance to AI in areas like art and writing, their neural responses reveal a genuine preference for AI-generated work. This anti-AI bias may stem from concerns about job displacement or the undervaluation of human creativity. Yet, despite these concerns, people frequently enjoy and even favor AI-generated creations. Ultimately, while debates around artificial intelligence continue, these results suggest that AI is becoming an enduring part of the creative landscape—one that challenges and reshapes traditional notions of art and creativity.
Note: ThinkAlike Laboratories, LLC received permission from the Association of Consumer Research to include the images in our blog post. These images are used with full authorization from the Association, ensuring compliance with their guidelines for publication and distribution.
Torrence, R. D., Egedus Hernandez, V., Dougherty, K., Barnett, S. B. (in Press). AI of the Beholder: The Brain's Paradoxical Liking of—Yet Bias Against—AI Artwork. Association of Consumer Research Conference Proceedings, Paris, France (2024).
Cona, Giorgia, Matthias Kliegel, and Patrizia S. Bisiacchi. 2015. “Differential Effects of Emotional Cues on Components of Prospective Memory: An ERP Study.” Frontiers in Human Neuroscience 9 (JAN). https://doi.org/10.3389/fnhum.2015.00010.
Cunningham, William A., Stacey D. Espinet, Colin G. Deyoung, and Philip David Zelazo. 2005. “Attitudes to the Right- and Left: Frontal ERP Asymmetries Associated with Stimulus Valence and Processing Goals.” NeuroImage 28 (4): 827–34. https://doi.org/10.1016/j.neuroimage.2005.04.044.