This is the 4th piece in a multi-part series on the psychology of AI-derived art and music. To begin with the first piece, start here

The pleasure we derive from art is complex and layered. The enjoyment we may feel for any given song, or piece of music, isn’t merely a sensory process. Instead, our capacity to feel is deeply tied to our underlying belief about the art and where it came from.

As we’ve seen, much of this is due to the psychology of essentialism: the tendency for us to find and interpret “the soul” within the objects and artworks we interact with. This helps explain our leanings towards works of art borne from inspiring origins. 

Essentialism also helps explain why we have such an aversion to art created through artificial intelligence. It's impossible to imagine a person with a pre-potent hatred for AI, and yet who loves listening to AI-derived music. Or who regularly finds deep meaning in AI-created art. As far as origin stories go, being spit out of an LLM is as uninspiring as it comes. 

But why does this anti-AI bias exist for art in the first place? Is it merely a symptom of the much broader “anti-big-tech” sentiment? Or does it reflect a deeper psychological dynamic; an anxiety over our own sense of human value and uniqueness?  Ultimately, the biggest barrier to AI Art isn’t its ability to create beautiful output, or even the ability to create works of art that people find deeply meaningful. The biggest barrier is that AI art has a major branding problem. 

Let’s dive in

The Psychology of Anti-AI Bias in Art

To find a piece of art meaningful is for it to take you beyond the immediate sensory experience. As we’ve seen, a big contribution here comes from an appreciation of the creative process. As Dennis Dutton distills in his 2011 book, The Art Instinct, “The value of a piece of art is rooted in assumptions about the human performance underlying its creation”

Rose Guingrich, a PhD student at Princeton University studying Human-AI interactions, echoes and extends this sentiment. “Above and beyond the art itself, we have a natural tendency to see the act of creation itself as being intrinsically valuable. And Artificial Intelligence can be seen as removing this process”

There’s a special connection between creativity and experience which we implicitly recognize. As Guingrich continues, “From a psychological perspective, creative ideas can come from a combination and recombination of ideas in your head that can come together in novel and unanticipated ways. In this way, you can’t separate the creative process, and the creative output from everything else you’ve ever experienced.” 

From the creator’s perspective, inspiration is mysterious and often spontaneous. New ideas simply pop into your head out of the blue, and we often have no clue where an idea comes from, or which set of previous experiences and encounters may have had a role. In this way, there’s little distinction between the creative process, and everything else that we do. 

Just as it is for LLMs, human creativity is inseparable from its training data: in this case, the entire corpus of one's experiences.

The Soul of AI-Created Music

As we’ve seen with the psychology of music, one way to think about the specialness of art is in terms of its deeper essence - its soul. Up until recently, this deeper layer was the exclusive domain of the human artist. Can AI create art with a soul? 

If you’re like most people, your answer here is most likely no. You’re in good company. This is the stance of musical mastermind Rick Rubin. The superproducer of an impressive array of artists that spans Jay-Z, The Red Hot Chili Peppers, and Adele, Rubin has spent decades with some of the most talented musicians in the world. 

When asked by Tyler Cowen about the prospect of AI creating great music, he replied, “I don’t think so, but I don’t know enough about it. I feel like what makes art good is the point of view. It’s not the actual content itself, but it’s the humanity in it. It’s the soul in it that makes it good. I don’t know that AI-generated art will have a soul. The soul is that element that makes something.”

Increasingly, AI isn’t being used as a tool to execute human creativity, but as the creative engine itself. It’s at this juncture that it trespasses into a special domain that feels special and uniquely human. If human experience ceases to be the engine of artistic output, it’s a threat to both art and human experience. 

Living Colour guitarist Vernon Reid articulates this sentiment after hearing SUNO AI’s most popular creation, a “blues” song entitled “Soul of The Machine”. As he told Rolling Stone, “The long-running dystopian ideal of separating difficult, messy, undesirable, and despised humanity from its creative output is at hand,”. In particular, he’s heavily critical of AI creating a song that is so divorced from the essential human conditions which created Blue music in the first place, a genre rooted in “an African American idiom, deeply tied to historical human trauma, and enslavement.”

The slow creep of AI into musical composition also reflects one of our deepest, oldest fears about technology. Beyond the fears of AI taking away our jobs, there’s a much deeper, existential fear about what AI may threaten to take away: human uniqueness. 

Assessing the Influence of The AI Brand

This anti-AI bias runs deep and wide. But it turns out, it has very little to do with the art itself. Instead, it has everything to do with AI as a brand. 

In 2023, researchers at Duke conducted a series of experiments that examined this directly. They presented participants with a series of artworks, and asked them to simply respond to how beautiful they felt it was. Each piece was accompanied by a simple label, indicating whether it was AI-created, or human-created. In reality, however, all of the pieces were created through AI. 

This design allowed the researchers to examine the specific contribution of the AI brand. Would the same pieces look as beautiful if people were lead to believe they were created with AI? Their findings were telling: Controlling for the art itself, the pieces labeled “human created” were rated as significantly more beautiful. The inverse was also true: A work of art suddenly appears less beautiful, and less meaningful, the moment we learn that it is created by AI. Like morality, AI has a serious branding problem.

We like human art more, but only when we know it's created by a human. But this then, begs the all-important question: Can we distinguish AI Art from Human Art? Increasingly, the answer is no. 

Distinguishing AI Art from Human Art

As the sophistication of AI has increased, so has our difficulty discerning its output. We have no idea what’s real and what’s AI. For example, an interactive tool at The New York Times, entitled “Test Yourself: Which Faces Were Made by A.I.?” allows readers to test their ability to distinguish pictures of actual faces or AI-generated faces. Participants are at chance. The same is true of artistic work. Research finds that, in most mediums, people can’t tell the difference between AI art and human art.  

We have a decidedly “pro-human” bias when it comes to art. We venerate the “human element” in creative works. But if we can’t even tell if a piece of art is created by a human, how can we claim to appreciate its human origins? It turns out though, that this search for human uniqueness is even more convoluted: We don’t even know what to look for. The elements that feel telling of a human creator are some of the easiest for an AI to emulate. 

Consider the following experiment. A team of researchers gave participants a range of different media to sort through, from AirBnB listings to dating profiles. Some of these were real and created by people while others were created by generative AI tools. Their task was to try and tell the difference. 

Like most people, participants did no better than chance - they simply could not discern between real listings and generative AI listings. This wasn’t at all surprising. What was most interesting was seeing what people looked for when trying to make this discernment.

Participants rated certain features of the listings as appearing more “human-like”. For example, AirBnB listings that told personal stories, or spoke in the first person, were more likely to be rated as human-generated. These features, however, are just as easy for AI to generate as anything else. They feel human, but they’re not. 

The same is increasingly true for art. The features we look for to try and identify the human origins of a piece of work aren’t as reliable as we think. What this suggests is that, when we look at a piece of media and we feel another human staring back at us, we should question the validity of these feelings. It’s becoming nearly impossible to discern human involvement in the creative process by looking at the creative output. 

This is a major issue when it comes to the deeper appreciation of the art. As we’ve seen, we naturally connect with what feels to be the “soul” of the art - the reflection of the humanity of which the art is an expression. This takes art beyond the immediate and the sensory, and into the deeper soul of the piece. But just like this sensory aspect of the art, AI, increasingly, has no trouble mimicking the soul part, as well. This all-important human element - the uncanny feeling we get when we look at something and feel humanity staring back, can easily be fooled. 

The Role of the AI Brand in AI Art

We like and appreciate human art, but we can’t tell it apart from AI-derived works. Moreover, we don’t even know what to look for when we try to make this discernment. The only way to know what’s what is if someone tells us which is which. The sheer knowledge of an artwork’s human origins - perhaps above all else, is the deciding factor for how much we can appreciate it.   

We’re now beginning to see this play out in the real-world art community. In 2022, outrage ensued after an A.I.-generated work, “Théâtre D’opéra Spatial,” won first prize in a Colorado State art fair. The piece was submitted by Jason M. Allen, who created the piece using MidJourney. Research finds that this anti-AI bias runs both deep and broad, transcending the art world altogether. Swap paintings for emails, and you get very similar results: When people are told that an AI tool was used for drafting an email, their reports of trust in the email writer decrease dramatically. 

All of this is strangely reminiscent of the Classic Pepsi-Challenge marketing experiments. Human creation is to the enjoyment of art what Coca-Cola is to the enjoyment of soda.

For those who need a refresher, these experiments date back to the 1970s when Pepsi's marketing team observed that, in blind taste tests, people seemed to favor Pepsi over Coke. Pepsi probed this further is a series of experimental trials, and the findings were consistently surprising.

When participants were aware of the brands they were sipping, a whopping 80 percent leaned towards Coke. However, when the brands were concealed, the tables turned dramatically: Blind taste testers showed a clear preference for Pepsi, tipping the scales at 53 percent compared to Coke's 47 percent. The results were clear: Despite its second place standing in the market, Pepsi emerged as the undisputed flavor champion.

A similar victor may soon emerge in the creative world. Human art may become, sadly, the Coca-Cola of the creative world: an inferior product, held up by a superior brand. 

Photo by Quino Al on Unsplash

So where does that leave us, and how should we think about this new wave of art? We examine these questions with the final piece in this series. Be the first to be notified, by signing up here (for free)

References for The Psychology AI Art Perception: The Influence of The Brand

Ahn, W. K., Taylor, E. G., Kato, D., Marsh, J. K., & Bloom, P. (2013). Causal essentialism in kinds. Quarterly journal of experimental psychology, 66(6), 1113-1130.

Bellaiche, L., Shahi, R., Turpin, M. H., Ragnhildstveit, A., Sprockett, S., Barr, N., ... & Seli, P. (2023). Humans versus AI: whether and why we prefer human-created compared to AI-created artwork. Cognitive Research: Principles and Implications, 8(1), 42.

Chiarella, S., Torromino, G., Gagliardi, D., Rossi, D., Babiloni, F., & Cartocci, G. (2022). Investigating the negative bias towards artificial intelligence: Effects of prior assignment of AI-authorship on the aesthetic appreciation of abstract paintings. Computers in Human Behavior, 137, 107406.

Guingrich, R. (March 2024) Interview, personal correspondence via zoom

Johnson, M., & Ghuman, P. (2020). Blindsight: The (mostly) hidden ways marketing reshapes our brains. BenBella Books.

Liu, Y., Mittal, A., Yang, D., & Bruckman, A. (2022). Will AI console me when I lose my pet? Understanding perceptions of AI-mediated email writing. In CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/3491102.3517731.

Turpin, M., Walker, A., Kara-Yakoubian, M., Gabert, N., Fugelsang, J., & Stolz, J. (2019). Bullshit makes the art grow profounder. Judgment and Decision Making, 14, 658–670.

Wilson, E. A. (2011). Affect and artificial intelligence. University of Washington Press.