Study reveals AI stories beat human writing in tests

Readers consistently rated short stories written by artificial intelligence higher than those penned by published authors, and most could not tell which was which, according to new research from Villanova University in the United States.

Fictional stories created using artificial intelligence ranked six per cent more superior in quality and eight per cent more engaging than those manually written by humans.

This research work, published by Cambridge University Press, involved participants between the ages of 18 and 81 years to determine their ability to distinguish between human-written fiction and fictional text created using ChatGPT.

The researchers used three stories written by published authors against three AI-generated stories on the same theme.

In the first experiment, 1,682 participants were presented with one story and then told, either accurately or falsely, who wrote it before rating its quality and engagement levels. Two additional experiments using 424 and 481 participants, respectively, involved identifying the author of the stories.

Dr Deena Weisberg, senior author and professor at the Department of Psychological and Brain Sciences of Villanova, stated that the stories that ranked highly were those that were produced using AI but which the reader thought were written by humans.

According to Weisberg, there is a bias towards human-authored stories based on the belief that creative stories require emotional intelligence.

Those that considered themselves conversant with AI tools were better at detecting AI-written content. Literary skills did not make any difference.

Weisberg suggested AI's clarity and directness may explain why readers found it easier to engage with, while human writing tended to be more subtle and demanded more cognitive effort to process.