Mind Over Model? Students' Evaluation and Use of ChatGPT ‐Generated Versus Human‐Generated Texts

ABSTRACT This experimental study investigated how high‐school students judged the credibility of ChatGPT‐generated versus human‐generated texts on two different topics, as well as how they justified their text credibility judgments and used the texts in a post‐reading integrative writing task. Results showed that, across both topics, students judged the texts to be less credible when they were presented as generated by ChatGPT than when they were presented as generated by a human, and they also justified their credibility judgments more by referring to how the texts were generated when they were presented as generated by ChatGPT. However, on the post‐reading writing task, students reading ChatGPT‐generated texts displayed a more integrated understanding of the texts on the two topics than did students reading human‐generated texts. These findings may have not only theoretical but also practical implications, suggesting that although students may put less trust in model‐generated texts due to the way they are created, it may be possible to harness their critical stance toward such texts in the service of deeper and more integrated understanding of the issues discussed in the texts.