When Robots Paint: How Will We Decide if AI Art Is Art?

Editor’s note: This article is part of a series called “1,000 words,” where we explore a current event through an image and explain the underlying social and behavioral science in 1,000 words or fewer—“a picture is worth a thousand words.” Above, in a photo from Sotheby’s, humanoid robot artist Ai-Da stands in front of its portrait of Alan Turing, which recently sold for $1.08 million. As AI becomes increasingly enmeshed in the human creative process, how will we think about the art it creates?  — Evan Nesterak, Editor-in-Chief

* * *

A portrait of mathematician and codebreaker Alan Turing recently sold for $1.08 million at auction, shattering original estimates. One might expect a flurry of emotions from the artist, perhaps a few tears. But she seemed to take the news in stride. 

The artist in question has a black bob and high cheekbones. She is unpretentious, her small frame clad in black denim overalls. She is also a robot.

Ai-Da (named after Ada Lovelace, considered the world’s first computer programmer) is the brainchild of gallerist Aidan Meller. His team built Ai-Da to provoke reflection on the evolving role of technology in human creativity. “The greatest artists in history grappled with their period of time,” Meller said, “and both celebrated and questioned society’s shifts.” 

In 2024, Meller prompted Ai-Da to create a piece for the AI for Good Global Summit. The portrait of Turing was the robot’s idea. Its elaborate creative process involved using its robot arm to paint 15 pieces of different parts of Turing’s face and one of the Bombe machine, which Turing built to decipher encrypted Nazi communications during World War II. It selected three, plus the painting of the Bombe, to compile into a single portrait. 

Meller hopes the experiment “challenges traditional notions of art and artists.” 

Questions of what is art and who is an artist elude straightforward answers—artists, philosophers, and scientists have been offering their interpretations for millennia. But researchers have made progress on a different question: What qualities matter to people in deciding whether to deem something “art?” 

It’s another question with many answers, but several emerge as particularly important: intention, effort, originality, communication, and essence. These qualities can help us understand the ways that AI-generated art is aligned, and the ways in which it is at odds, with what we feel art should be. 

What qualities matter to people in deciding whether to deem something “art?”

First is intention. In short, if you didn’t mean to make art, you didn’t make art. We develop a radar for intention from an early age—three-year-olds will call the same splotch of paint a “spill” if done by accident, a “painting” if done on purpose. If you prompt an AI to generate an image of a rainforest, but the AI is responsible for the sunbeams piercing the canopy, the dewdrops studding the leaves, and the pair of tree frogs peering out from underneath a rock, did you really intend to create that rainforest? 

Next: the more effort a work takes to produce, the more we like it. This is a qualm with simpler forms of AI-generated art, as in the early days of photography. All you did was push a button! But there may be a loophole in the preference for high-effort work in originality. Philosopher Raphaël Millière describes an artist who puts a canvas in the forest and lets the elements “do their work.” Though leaving a canvas outside and seeing what happens is about as low-effort it gets, it’s original. If everyone started leaving canvases in their backyards the way everyone is now generating images with tools like DALL-E, maybe we’d be less inclined to call those canvases “art.”

Others worry that AI will compromise the communicative quality of art. What happens to the dialogue between artist and audience when the artist isn’t human?

Related but more nebulous is the “essence” imbued by a creator in their work. Despite its intangibility, psychologist Ellen Winner and her colleagues found a clever way to investigate essence by exploring how we think about copies of art. 

An original work is always going to be valued more than a copy. Copies come with baggage—they’re worth less on the market, and the possibility of forgery provokes a moral repugnance. But what if you remove money and morality from the equation?

The researchers showed participants an original artwork, then asked them to evaluate one of three identical copies. They told participants that it was created by either the artist, the artist’s assistant, or a forger. As expected, the forged copy was devalued the most. Crucially, the copy by the assistant was devalued more than the copy by the artist, despite being otherwise indistinguishable. Both were copies, both were signed by the artist, neither sparked moral quandaries, and they would sell for the same amount. 

“Our argument is that what’s left is simply knowing that it was touched by the hand and the mind of the artist,” Winner said. “This is a mystical, magical belief…. I think it comes down to thinking that you’re reading the traces of the artist’s mind when you look at the work. And we’re not interested in the traces of a copier’s mind, because the mind is less interesting.”

This presents a formidable hurdle for AI-generated art. It’s impossible to read traces of an artist’s mind when the artist in question doesn’t have one.

Our preference for qualities like effort and essence illuminate why we often resist technologies that we feel make art easy and impersonal. 

There is no single condition that excludes AI-generated art from being art for the same reason there is no single condition that excludes human-generated art from being art. But our preference for qualities like effort and essence illuminate why we often resist technologies that we feel make art easy and impersonal. 

This dilemma isn’t new. Synthesizers were banned by the U.K. Musician’s Union in the 1980s; Charlie Chaplin said he’d “give the talkies three years, that’s all.” Painter J. M. W. Turner glimpsed an early photograph in 1839 and declared, “This is the end of Art. I am glad I have had my day.”

But today, synthesizers are uncontroversial. Talkies are just movies; silent films require the extra qualifier. Museums feature painters and photographers side-by-side. Art is alive and well, if different than before. The question remains how AI’s growing role in the creative process will change what we create and how we perceive it. 

Researchers recently found that evaluating human-created art alongside AI-generated art made people more appreciative of human creativity. Study coauthor Sheena Iyengar reflected: “AI can be a powerful tool enabling human–AI collaborations to emerge as a distinct art form that elevates, rather than erases, the unique role of human talent.”