Study: Readers Prefer AI Stories When Believed Human

A recent study published by Cambridge University Press indicates that readers ultimately prefer AI-generated stories, particularly when they are led to believe the work is human-authored. The research involved 1,682 adult participants who were presented with six short stories. Three of these stories were written by human authors, while the other three were generated by ChatGPT. Each AI-generated story was designed to have a similar topic and style to one of the human-written samples. Participants were informed whether the story they read was written by a human or by artificial intelligence, though this information was not always accurate. Across the board, participants rated the AI-written stories as more engaging and of higher quality compared to those written by humans. When participants were mistakenly told that the AI-generated stories were written by humans, they assigned even higher ratings, suggesting a pre-existing bias against AI-generated content that is independent of its actual quality. Deena Skolnick Weisberg, an author of the study and a researcher from Villanova University's department of psychological and brain sciences, stated in a press release that public perceptions of AI's capabilities are often outdated. Weisberg explained that the assumption that creative writing necessitates unique human attributes like emotional understanding and lived experience leads to an underestimation of AI's current abilities. The study highlights a disconnect between the perceived limitations of AI and its actual performance in creative writing tasks. The AI-generated writing is trained on vast amounts of human work, enabling it to mimic and even surpass human creativity in certain aspects. This finding challenges the prevailing notion that AI-generated art inherently lacks a 'human touch' and suggests that reader preference is more influenced by their expectations and biases than by the objective quality of the work. The implications of this study extend to the literary industry, where the use of AI remains a sensitive topic, often met with skepticism and backlash, as evidenced by past controversies surrounding AI-assisted publications and awards. The research provides empirical evidence that the negative reactions to AI-generated creative content may stem more from the knowledge of its origin rather than its inherent artistic merit. This suggests a need for a re-evaluation of how AI-generated creative works are perceived and judged in the future, potentially leading to a more open acceptance of AI as a tool in creative fields.