AI-Generated Stories Outrank Human Writing in Quality, Readability Tests

Most people prefer AI-written stories - and can't tell when they are made by humans | New Scientist

AI-generated literature may have an unfair reputation for poor quality. Despite a cultural backlash to AI-generated content, people seem to prefer it - and if they're lied to and told that it was created by a human, their opinion rises even higher. Sydney Sears and Deena Weisberg at Villanova University in Pennsylvania recruited 1682 adults - selected to be representative of the US in terms of their gender, age, and race - and asked them to read a story. Half the participants read a story written by a human, and half read one written by AI. Each of these groups was again split in two, with half told that their story was written by a human and half that their story was written by AI, regardless of its true origin. The human-created stories were sourced from a literary journal, while the AI-generated ones were created using ChatGPT 4.0. Those who read an AI-generated story ranked its quality as 1.54, on average, when asked to score it on several qualities on a scale ranging from -3 to 3. Meanwhile, participants who read a human-written story gave it an average score of just 0.97. Participants who read an AI-generated story also ranked it as more absorbing, based on several qualities on a similar scale, with an average score of 1.42. Those who read a human-written story, meanwhile, gave an average score of just 1. In all cases, the story was rated slightly higher if the participant was told it was written by a human. The researchers also ran two further studies, where several hundred people were given an AI-generated story and a human-written story and asked to identify which was which. In the first test, only 39 per cent made a correct identification, while in the second, 52 per cent were correct. The researchers have no explanation for the difference, except that the second test took place several months after the first; society may be adapting to better understand AI output. In any case, they claim that it shows people are no better than chance at distinguishing human work from AI output. Claire Hardaker at Lancaster University, UK, says that humans are having to come to terms with the uncomfortable idea that we aren't any better than robots at writing short stories. She says that in her own research on testing people's ability to identify AI-generated speech, text and music, it is the music that angers them most of all when they are tricked. "I think there's a huge sensitivity [around AI generating art]," she says. "When you have AI that is continually improving at the speed that it's improving at, there is going to come a point where every single model can produce Shakespeare-like output. What does that mean for humans?" While AI is becoming increasingly capable with every passing month, the concept of AI-generated literature and art is creating a backlash. Earlier this year, a short story given the Commonwealth prize and published in Granta was accused by some of being AI-generated - something which the competition explicitly forbade. An investigation was inconclusive. Last year, artists criticised an auction of AI-generated art at Christie's because they saw the work as "mass theft", due to models being trained on existing images. Rodney Jones at the University of Reading, UK, says that part of the preference for AI-generated writing in this case may be down to researchers taking stories from a literary journal, rather than choosing a more accessible and popular genre. "People like expected things. People like easy-to-process things. Real human writing, real kind of human literary writing, is, frankly, not very popular, because it's hard work," says Jones. Although ChatGPT was asked to create new stories for the experiment that had traits similar to those of the human-written stories, Jones thinks it would have produced a simpler, easier-to-read version by default. Weisberg agrees that AI produces more "vanilla" content than humans do, and ascribes the perceived difference in quality to this. She plans to run experiments in the future on different types of writing to see if AI retains its lead. "I think AI will be able to accomplish some truly amazing feats of creativity, but I think AI creativity is just different from human creativity," says Weisberg. "I think it will be able to do something that we will recognise as having artistic value, but it might be measured on a different scale."

AI-generated stories rated better quality than human-written ones, study finds

Researcher says simpler writing by AI is easier to read and digest, but this does not mean human authors are obsolete From Homer's Odyssey to Agatha Christie's crime fiction, humans have produced plenty of tales to grace the bookshelf. But research suggests artificial intelligence may now be providing some competition. In a study published in the journal Judgment and Decision Making, 1,682 adults were asked to read one of six short stories, three of which were written by humans and three by ChatGPT. Each AI story had a similar theme to one of the human-authored works. The team told participants whether their story was written by a human or AI, but this information was not always correct. The researchers then asked participants to rate how absorbing and engaging they found the story, and its quality. Participants who read an AI-generated story rated it as more absorbing and of higher quality than those who read a story written by a human. However, participants gave higher ratings to stories they had been told were written by people. Dr Deena Skolnick Weisberg, a senior author of the research from Villanova University in Pennsylvania, said: "AI systems can already generate short stories that are seen as being at least as good as - if not better than - human-written stories. We should update our views of AI's abilities accordingly." However, Weisberg, who is herself a creative writer, said that did not mean writing should be left to AI, noting that novels produced by tech were probably going to be different from those written by humans. "We may need to make room for AI-generated novels, and for AI/human co-written novels, but that doesn't mean that there's no longer space for us to appreciate the process of human creativity," she said. "Humans write as a form of creative self-expression or to challenge ourselves, or to make sense of our experiences. The fact that AI can generate human-like stories doesn't change any of that." The results also suggest attitudes to AI matter: participants with more positive attitudes toward AI gave higher ratings to stories they were told were produced by ChatGPT than those with a less positive outlook. The researchers carried out two further experiments in which 905 adults were given two of the six short stories - both on the same theme and only one of which was penned by a person. They were then asked to guess which was written by a human and which by AI. In one experiment, about 40% guessed correctly, a result that is worse than chance, while in the other experiment 52% guessed correctly. "There is no general evidence that participants were able to distinguish human-written from AI-generated stories," the authors write. The researchers found that while participants' self-reported expertise with fiction was not linked to how accurately they guessed, greater expertise with AI systems was. "AI writing has a particular style that we can learn to recognise with practice - just as we can learn to recognise different human authors' unique styles," Weisberg said. Weisberg suggested that the preference for stories labelled as being written by humans was linked to a desire for authenticity, but that people actually preferred AI-generated stories because they were easier to read and digest. However, Luke Kennard, a professor of film and creative writing at the University of Birmingham, who was not involved in the work, emphasised that AI was highly contentious. "Because we're human, we tend to personify AI and imagine it as some kind of glowing robot with a benign and neutral face, when it's really just predictive code based on a massive stolen database of actual writing, physically based in vast, expensive and destructive datacentres with catastrophic repercussions for the surrounding communities," he said. Kennard also stressed that AI was not a "tool" similar to spellcheck or a thesaurus, but an existential threat. "Of course it's capable of spitting out a coherent and even enjoyable poem with the right prompting and finessing. Given that it's been trained on thousands of years of human work, it would be embarrassing if it wasn't," he said. "I suppose the question we should be asking is: is it worth it? And the answer is absolutely, categorically, it is not."

Is AI Telling Better Stories Than Us? Why Human Writers May Be Less Appealing

Not only are people unable to tell the difference between stories written by humans or artificial intelligence, they actually prefer it when the stories are created by AI. And that's especially true when they are told that the stories had human authors, according to a new study from Villanova University. The study consisted of multiple experiments, all of which indicated a preference for AI. In the first, 1,682 participants were told to rate both the quality of six short stories they were given and how "engaging" they were. They were told, either correctly or incorrectly, who or what had authored each one. Not only were AI-generated stories found to be higher-quality (by 6%) and more engaging (by 8%), they were rated "even more highly," researchers concluded, when participants believed that they were written by humans (by an additional 3%). Deena Weisberg, the senior author of the study and an associate professor from Villanova's Department of Psychological & Brain Sciences, said in a press release that the finding "reveals a bias towards narratives written by real people." That may be because "we assume creative writing requires uniquely human qualities, such as emotional understanding and lived experience," she said, explaining that it shows how "public assumptions about AI's capabilities are increasingly out of date." Furthermore, people were only able to identify AI-generated content roughly 40% and 52% of the time in two subsequent experiments, each with over 400 participants. People who said they were AI-literate had more success, while people who claimed to have a background in literature were less accurate in their guesses. "Familiarity with AI systems appeared to help people recognize the patterns typical of AI-generated writing, such as em dashes and sentence structures such as 'it's not just X, it's Y,'" Weisberg said. "That suggests that improving AI literacy would be one way to help people to navigate the new AI-enabled world that we're living in." Cameron Jones, assistant professor of psychology at Stony Brook University tells TIME that the study's findings are not necessarily surprising, considering the history of research going back years that shows people's diminishing ability to tell between artificial intelligence and humans. He also points to the trajectory of his own studies, which look at whether people can differentiate between them in conversation. In those scenarios, he says, "The person actually gets to quiz the model and ask follow-up questions." Even then, they had trouble distinguishing between human and generative content partners. Jones says that this is because the large language models driving the content responses are trained to appeal to users. "We basically get the model to generate little bits of text, and then we get people to read them, and they give a thumbs up if they like what the model's saying and they give a thumbs down if they don't like what the model's saying," he says. It's no surprise that people prefer the results, he explains, when "we're optimizing these models' outputs against our preferences." It can oversimplify output, but it leads to writing with higher overall appeal -- whereas humans might write something truly exceptional that only caters to select tastes. "Real humans are weird and idiosyncratic, and they're very different from one another," Jones says. "But models can kind of learn to be this kind of milquetoast everyman who appeals to everybody." While the findings align with rising exposure to artificial intelligence, along with algorithmic reward systems for broadly approachable content on platforms like TikTok, Weisberg does not think that the current digital landscape is to blame for the results. She says, "Technologies amplify existing tendencies, rather than creating them." Plus, she adds, the human-authored content might have simply challenged readers more. "AI writing tends to be clearer, more direct and easier to process," she said. And it's reasonable for participants to find themselves more deeply engaged with content that offers greater predictability and less friction. In other words, Weisberg says, "Difficult or subtle material requires more brainpower."