PHOENIX (AZFamily) \u2014 When it comes to creative writing, humans may have some competition.<\/p>\n
A new study suggests people cannot tell when a short story was written by artificial intelligence, and that they often prefer what the machine produces.<\/p>\n
Researchers at Villanova University asked 1,682 people to read one of six short stories. Three came from published human authors. Three were generated by ChatGPT. Readers were told who wrote the story they read, but half the time, the researchers lied about the label.<\/p>\n
The results revealed two competing effects. Stories that claimed to be human-written drew higher marks regardless of who actually wrote them. AI-generated stories were rated higher quality and more absorbing than the human ones, and the highest-rated stories of all were AI stories falsely labeled as human.<\/p>\n
In follow-up experiments, readers guessed the author correctly about 40% and 52% of the time \u2014 no better than chance. The findings were published in the journal Judgment and Decision Making.<\/p>\n
\u2018Humans are still team human\u2019<\/p>\n
On the latest episode of Generation AI, Ed Finn, founding director of the Center for Science and the Imagination at Arizona State University, said the first result is worth sitting with.<\/p>\n
\u201cLet\u2019s celebrate the fact that humans are still team human,\u201d said Finn, a science fiction scholar and former journalist for Time and Slate. \u201cOne of the reasons that we engage with art is to connect with other people. And we\u2019re looking for other people.\u201d<\/p>\n
That people also project humanity onto a machine, he said, is both interesting and ironic \u2014 evidence of how strongly readers are wired to seek an emotional connection.<\/p>\n
The second finding that humans preferred AI-written stories did not surprise him. Chatbots have been passing as human in limited settings for years, he said, and the gap has narrowed steadily.<\/p>\n
The authors of the study theorized that readers may have preferred the AI stories because AI-generated writing tends to be more direct and easier to understand, qualities people generally find appealing. The AI versions in the study typically stated their themes outright, rather than leaving readers to infer meaning from what the characters said and did, the authors noted.<\/p>\n
What it means for working writers<\/p>\n
Finn said the study raises a harder question about economics. Most writers already cannot make a living from writing alone, and illustration and design work is disappearing as clients accept what he called the good-enough version of AI content.<\/p>\n
\u201cThe floor is dropping, and it is incredibly easy for anyone to push a button and generate a short story or a screenplay or a news report,\u201d he said. \u201cAnd that doesn\u2019t necessarily mean that those products are going to be more compelling or better.\u201d<\/p>\n
He compared the situation to live music, another industry reckoning with generated content. Audiences fall in love with musicians and the community around them, he said, not just the recording. He said creative fields may need to build similar structures of value around human authorship.<\/p>\n
Asked whether AI threatens the apprenticeship that turns bad writers into good ones, Finn said it depends on how schools deploy the tools. He pointed to a Socratic model in which the student writes the first draft and the AI responds like an editor asking questions.<\/p>\n
\u201cThere\u2019s always this slippery slope, because it is so tempting to just push the button,\u201d he said.<\/p>\n
A feedback loop, and a warning for news<\/p>\n
He warned of a feedback loop as well. The more machine-generated content circulates, the more of it gets fed back into new models during training, potentially separating newer AI systems further from human creativity.<\/p>\n
\u201cWe\u2019re sort of eating the seed corn right now,\u201d Finn said.<\/p>\n
The same dynamic could reach journalism, he said. He sketched out a future in which the news is whatever an AI thinks a viewer wants to hear, eroding a shared sense of reality and the friction of encountering disagreement.<\/p>\n
\u201cWe\u2019re in this three-wishes genie situation, where you\u2019re going to get exactly what you wish for,\u201d Finn said. \u201cAnd it turns out that\u2019s not really a good thing.\u201d<\/p>\n
Generation AI airs Wednesday nights on AZ Family+.<\/p>\n
See a spelling or grammatical error in our story? Please click here to report it<\/a>.<\/p>\n