Study Finds AI-Generated Stories Are Preferred, Especially if They Are Thought to Be Written by Humans

JAKARTA - Human writing is not necessarily more liked by readers. A study found that ChatGPT-made stories got higher scores, especially when readers thought the story was written by a human.

CNet quoted Saturday, August 8, reported that the study led by Cambridge University Press and Villanova University researchers involved 1,682 participants aged 18 to 81 years.

Participants are asked to read a number of fictional stories, assess their quality, and try to distinguish between human works and artificial intelligence or AI.

The research was carried out in three experiments.

In the first experiment, participants were given a story. Some were told the story was created by AI, while others were informed that the story was written by a human. They were then asked to assess its quality and appeal.

In the second and third experiments, participants were divided into two groups, each of which numbered 424 and 481 people.

Each participant read one human-written story and one ChatGPT-generated story without being told which was AI-generated. They were then asked to guess which story was written by ChatGPT.

As a result, the stories made by ChatGPT received the highest rating. The value was even higher when participants were deliberately given the wrong information that the story was written by a human.

Many participants also had difficulty distinguishing between AI and human writing. More than 60 percent of participants in the second experiment and more than 48 percent in the third experiment failed to correctly identify AI-generated content.

Deena Weisberg, senior author of the study and professor of psychology and brain science at Villanova University, said the results showed a continued tendency to place a higher value on writing believed to be human-made.

"We assume that creative writing requires human-specific qualities, such as emotional understanding and life experience. This makes people underestimate the ability of AI," said Weisberg.

According to Weisberg, public perceptions of AI capabilities are increasingly lagging behind the development of the technology.

Similar findings have emerged in a 2023 study led by Renee Richardson Gosline of MIT.

The study found that AI-generated content and AI that makes final decisions using human results as a reference were judged to be of higher quality than the work of human experts and humans who make final decisions with the help of AI results.

Weisberg suspects that AI writing is more acceptable because it tends to be direct, clear, and simple.

In contrast, human writing is usually more subtle, complex, and gives readers more room to interpret.

In the tested stories, the AI-generated text generally used simpler sentences, shorter paragraphs, and fewer anecdotes, metaphors, and metaphors.

According to CNet, the results of the study do not mean that readers have bad taste or that AI has abilities that human writers do not have.

AI writing tends to use story patterns, sentence structure, language, and styles that many people have already liked. The AI system itself is trained using various human writings that have been enjoyed and consumed for decades.

Weisberg said the findings also could not be fully explained by issues such as declining attention spans or the influence of platforms such as TikTok.

"This technology reinforces existing tendencies rather than creating new tendencies," he said.