Often situated as diametrically opposed, it seems that creativity and artificial intelligence (AI) are far more connected than one might think.

Assistant Professor of Economics Martin Abel and Senior Lecturer in Russian, East European and Eurasian Studies Reed Johnson have been collaborating on a project to investigate the links between AI and creative writing.

“[Johnson] and I were interested in how people perceive AI, especially for creative work. There’s plenty of research showing that people have this sort of intrinsic bias against AI-created work, and we were interested in how much it actually translates to behavior beyond what people report,” Abel said.

In the fall of 2024, Abel and Johnson began conceiving an experiment to test people’s assessment of AI writing and their willingness to pay for AI-generated versus human-created work. They gathered a sample of 700 people across the United States and told half of the participants that a writing sample was human-generated and the other half that it was AI-generated. Both were, in fact, generated by AI.

Before the experiment was conducted, Abel asked ChatGPT 3.5 to write a creative writing sample. He showed it to Johnson, who was surprised and impressed.

“I thought, ‘Wow, that’s actually a lot better than I thought it was going to be,’” Johnson said. “If he had told me that it was a human, I certainly would have believed it … This was more than one generation ago for AI, so I was struck by that. I was very happy to get involved in this research in order to think about what distinguishes human writing from AI writing.”

This early sample was the same one they later used in their experiment.

“[Abel] and I connected over a shared fascination with these large language models [LLMs] and what they could do, and what they couldn’t do,” Johnson said. “[Abel] has done a lot more research around this, and I was just interested, particularly as someone who has a strong interest in creative writing, as a scholar and practitioner.”

What the pair has found so far seems to confirm what other researchers have uncovered.

“We replicate what other people have found: that people have this bias against AI,” Abel said. “We find that it’s perceived as less authentic, less complex, less creative, more predictable.”

But once participants reach halfway through the story, there’s a twist in the experiment.

“Halfway through the story, we stop it, and we tell them, ‘In order to read the end of the story, you have to pay a price. How much are you willing to pay?’” Abel said.

The participants received money at the beginning of the experiment, so they could spend some of the compensation to finish the story.

“Something that really surprised us was that we basically found no difference whatsoever in the willingness to pay to read the end of the story, despite these big differences in subjective assessment,” Abel said.

This study is the first to show that this existing AI bias does not translate into willingness to pay. However, the duo was curious about how AI might function within the real-world market for fiction and literary works.

“Even if people believe that AI is inferior work, the market drives a lot of what gets produced, what gets read, what labor is compensated,” Johnson said. “We found that troubling because despite people’s apparent dislike of AI-generated works, it didn’t show up in terms of whether they’re willing to pay for them or not.”

This begs the question of what actually constitutes “creative writing” as we know it today.

“The current technological limitations are such that the things LLMs tend to produce, in terms of fiction, are derivative by nature,” Johnson said. “So I think the question is: Is human writing substantially different in that respect?”

Johnson acknowledged the inherent derivative nature of both human and AI writing, but he is still certain that current AI writing is not exactly on par with human writing.

“One thing that’s lacking in the current technology we have is a sense of a world larger than the textual training data that it’s been trained on,” Johnson said. “What it doesn’t have is experience. It’s very hard for an LLM to tell us something that will force us to see the world we live in anew if it’s only recycling bits of training data.”

Abel and Johnson’s research is still in its early stages, but they are interested in seeing more studies comparing AI and human-generated creative writing.

“My sense is that LLMs now match lay-writers, if not surpass them,” Johnson said. “But what I’m interested in investigating, in a further iteration of our research, is whether they can catch up to expert writers. My sense is that they won’t and probably won’t anytime soon.”

However, Ada Potter ’26, a visual arts major, does not see the connection between LLMs and creativity, specifically within her discipline.

“I think a lot of the studio work that I’ve seen and done at Bowdoin has seemed very antithetical to AI. All of the classes I’ve taken have been very materially driven. We are encouraged to dive into a material, whether it’s charcoal or oil paint or plaster or linocut or cardboard or ceramic, and to think just as much with our hands as we would with our minds or eyes,” she said. “I think that is how I create best—by experimentation and iteration.”

Potter says there has not been a lot of discussion about AI in her studio arts classes. She says she finds herself enamored with the physicality of art—how one needs to use their body to create it—and is uncertain about how that could be possible with AI.

“I love creating art because your body is thinking the whole time. Your hands are in tune with your eyes and your mind and your breath and your posture,” she said. “I am not familiar with many modes of AI creativity, but I know that it often prioritizes only the thinking in your brain, which can be such a loss. Or, it minimizes the thinking and problem-solving in your brain, streamlining the process of creation, which is even more of a loss.”

But  Potter  did see some positive potential when considering how AI could help people understand and define what art is.

“I think that one thing that AI has made us all think about is how to define art, which is very useful. I think that for so long, in sort of an ineffable way, art has been seen as integral to our humanity, and humanity seen as integral to our art,” she said. “But now, AI is making us all talk about what specifically creation loses, if anything, when it is separated from human hands.”