I, ChatGPT: linguistic properties and human experiences of human- versus AI-generated stories

作者：Markus Appel, Wojciech Małecki, Tanja Messingschlager, Julia R. Winkler · 发表于：Humanities and Social Sciences Communications · 年份：2025 · DOI：10.1057/s41599-025-06341-2 · 被引用次数：2 · 研究领域：Artificial Intelligence in Healthcare and Education、AI in Service Interactions、Mental Health via Writing

Abstract Generative AI is increasingly used to create texts, including fictional stories. Do stories generated by AI differ from stories created by humans regarding linguistic properties and recipients’ experiences? To answer these questions, we first asked ChatGPT and 100 non-professional human authors (i.e., students) to create stories based on similar prompts (Study 1). Linguistic analyses showed that ChatGPT stories included fewer personal pronouns and fewer descriptions of relativity than human stories, but more positive emotions. In Study 2 ( N = 380), naïve participants were randomly assigned to read an AI-generated or a human story from the pool of 200 stories from Study 1. No differences in novelty or entertainment experiences were found, but the readers of AI stories were less transported into the story world. Mediation analyses show that this difference can be attributed to ChatGPT’s use of fewer personal pronouns. Differences in the use of literary techniques between AI and humans are discussed.