Survey Results Show People Prefer More Human Involvement in AI-driven Art

Think of your favorite piece of art—a painting, a song, a novel, a movie or even a video game—and try to remember why it made such a strong impression on you. Was it the color, the cadence of notes, the way the writer made you feel understood, the deep emotion of the actors?

Now imagine that artificial intelligence created it.

The question might seem flippant, yet this is the future toward which we are racing. Over the past few years, AI developers have improved the technology’s ability to create art across nearly every field: not just writing, digital art, photos and videos but also three-dimensional models, dance choreography and architectural designs. With AI so rapidly learning to produce art forms previously considered the exclusive domain of human ingenuity, we thought it important to understand how people view this transformation.

On supporting science journalism

If you’re enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.

We each approached this question from different backgrounds. One of us (Béchard) is a journalist and fiction writer who has been publishing novels for 20 years, and the other (Kreiman) is a professor at Harvard Medical School who researches the intersection between biological and artificial systems. Earlier this year we conducted a survey on AI art using Prolific, an online platform that pays people to participate in research. The only restriction we placed was that the respondents reside in the U.S. We enrolled 150 people. What we found has not yet been published or peer-reviewed.

The results were striking. The majority of people who participated disliked the idea of AI-generated art and held the view that human art has an emotional depth that machines either can’t or shouldn’t reproduce. Yet they were open to AI-generated art so long as there was an artist involved, heavily guiding and prompting their chosen platform. At a moment when we face a deluge of AI-generated content, we believe that AI companies should heed these data and focus on what people value—rather than create systems that generate large volumes of art, they should design better tools to give people the power to convey their unique artistic visions. In doing this, their platforms could make creative expression more accessible and egalitarian in a world where creating art comes with barriers for many people. In this way, AI could emerge as another path for human expression.

We first asked people to name their favorite work of art. Answers included J. D. Salinger’s The Catcher in the Rye, J.R.R. Tolkien’s The Fellowship of the Ring, Vincent van Gogh’s The Starry Night and Gustav Klimt’s The Kiss, as well as the musical Hamilton, the TV shows Gilmore Girls, Twin Peaks and Game of Thrones, and music by the Beatles, Metallica and Cat Stevens. Then we asked how they would feel upon learning that their favorite work of art was generated by AI with no humans involved, not even real actors in films. Responses ranged from outrage to enthusiasm. More than 62 percent said they would like the art less, 32 percent said that their feelings wouldn’t change, and nearly 5 percent said that they would like it more.

In comments, a devotee of Good Will Hunting expressed horror, calling the movie “a masterpiece of the human experience” that “no AI could ever come close to replicating.” Many bluntly declared that AI art isn’t “real,” that it is manipulative, inauthentic and, of course, artificial. Then there were the pragmatists, who shrugged and argued that if the artwork touches your soul, why fuss about the creator—or lack thereof? One respondent even welcomed the idea of AI-generated TV, envisioning the dream scenario for an insatiable binge-watcher: endless episodes churned out by algorithms.

The reactions, however, went deeper. When asked if there is a difference between the emotional value of human and AI art, 81 percent said yes, reminding us that we seek the human experience in art. We want to know if an artist drew from personal experience or imagination. We pause while reading a book to learn about its author, and we follow the lives of singers and actors. Art, after all, is a way humans communicate. To wit: art “is the most universal and freest form of communication,” philosopher John Dewey said in his book Art as Experience; “You think your pain and your heartbreak are unprecedented in the history of the world, but then you read,” civil rights activist James Baldwin told Life magazine in a 1963 interview.

But if great art is forged from human hope, desire, disappointment and sorrow, what does creativity mean in a world with increasingly powerful AI? Are people who use AI, versus paint, film or even a word processor, artists? We asked this. Thirteen percent said yes, another 13 percent were unsure, and 31 percent said no, whereas 42 percent selected “yes, but only if they are providing significant guidance to the AI; otherwise, no.” These responses shed light on why so many people, in person and online, complain about AI-generated “slop.” What participants are often pointing out is the impersonal, almost meaningless nature of AI: the online posts hyping an idea in generic language, the onslaught of images that say little about the person posting them beyond their ability to write a short prompt and hit send. But many of the respondents in our survey appeared more open to the idea of people using AI as a tool—an electronic quill or paintbrush—to help bring their artistic vision into the world.

This is important.

Art is not accessible to everyone. Many people with ideas for films or music or stories may never have the resources to create them—the lyricist who wants to put music to words, the scriptwriter who craves to see their lines spoken on a screen. Art supplies and studio space are expensive, and people can be limited by geography, finances, physical disability or industry gatekeepers. Darren Aronofsky, award-winning director of Black Swan, founded the AI film studio Primordial Soup to lower barriers for emerging storytellers and to develop scripts that remain unproduced because of high costs and technical limitations. Increasingly, though, consumer AI systems are becoming so powerful that people can create films with them on their phone or desktop.

In the few years that commercial AI, such ChatGPT and Midjourney, have been available, certain ideas about AI art may have been normalized. When asked which kinds of art AI could acceptably create, more than one third of respondents listed digital art, which was closely followed by poetry and fiction. It’s no coincidence that early commercial AI systems could rapidly generate these types of content, and it’s likely this is what our participants are most familiar with. The observation that mass production decreases value (both perceived and actual) isn’t new. In the 1899 book The Theory of the Leisure Class, economist and sociologist Thorstein Veblen wrote, “The marks of hand labor come to be honorific, and the goods which exhibit these marks take rank as of higher grade than the corresponding machine product.” We also tend to appreciate tasks that are challenging—to place more value on a film that required years to create or even simply on handmade clothes. “Effort is used as a heuristic for quality,” wrote social psychologist Justin Kruger and his colleagues in a 2004 study. We believe that digital art, poetry or fiction, rather than having lost their value, require a clearer connection to the author’s story for people to be willing to trust them.

By contrast, the types of AI art that respondents rated as least acceptable were podcasts, TV shows and movies—art forms that AI still struggles to convincingly emulate. In recent years podcasts have soared in popularity and clearly embody the urge to communicate. Recent surveys have even found that many people prefer watching podcasts so that they can see the speakers’ facial expressions and gestures. Research on the new creative economy shows that the strongest driver for podcasting success is a clear personality with which the audience can connect. Whether these trends are in response to the recent deluge of impersonal media, we can’t say, but they tell us that even if artists choose to embrace AI, they should consider doing so as a means to more clearly convey the uniqueness of their vision. Still, creators must keep in mind copyright violations when human art is used to train systems and whether the AI tools they use will generate content such as deepfakes that inflict harm on people. People should invest in platforms that build safeguards and set ethical standards, such as the video-generation platform Moonvalley, which trains its model using only licensed data.

In the near future, as AI conquers more artistic domains and creates entire films or video podcasts indistinguishable from those made by and with humans, what people find acceptable may shift. But we do know that people still value art as a means of communication and connection, and taking the pulse of what society thinks about AI art can help align AI developers not only with artists but with the millions of people who find meaning and connection in art.

This is an opinion and analysis article, and the views expressed by the author or authors are not necessarily those of Scientific American.