Reflective writing assignments in the era of GenAI: student behavior and attitudes suggest utility, not futility

Reflective writing is widely used in health sciences education, but overreliance on generative artificial intelligence (GenAI) could undermine the reflective writing process. To explore this concern, students in three undergraduate courses with reflective writing assignments and policies permitting GenAI use were asked to retrospectively and anonymously self-report their GenAI-related behaviors and attitudes. Only 33% of respondents (n = 310) reported ever using GenAI on a reflective writing assignment. Among GenAI users, 81% reported that usage was motivated by learning, efficiency, and/or (to a significantly lesser extent) grades. Eighty-six percent of users reported benefits to learning, efficiency, and/or grades, but 10% reported that learning was hindered by using GenAI. Most GenAI users (83%) believed their usage of GenAI was ethical, and only 4% regretted their use. Notably, 19% of users and 38% of nonusers wished they had used GenAI more. Overall, only four assignments (representing 1.3% of respondents and 0.3% of submissions) were reportedly "mostly written by GenAI." Instead, most students reported using GenAI selectively and in ways that were supportive rather than substitutive of their own reflective process. This finding inspires optimism that reflective writing assignments have retained their pedagogical value in the early GenAI era and that most students are well intentioned in their usage of GenAI. Heterogeneity in self-reported student behavior, motivations, and perceptions of GenAI's benefits and harms highlights the need for further research into factors influencing GenAI adoption and usage. Understanding and responding to this diversity will be crucial for developing inclusive and equitable strategies to help maximize GenAI's benefits while minimizing its harms.NEW & NOTEWORTHY We examined students' use of GenAI tools to complete reflective writing assignments in health and exercise science courses where these tools were permitted. Findings do not support common concerns about student overuse and misuse of GenAI tools. Instead, we provide evidence that students are using GenAI tools selectively and in ways they believe to be ethical and supportive of their learning. Tremendous variability in student behavior and attitudes warrants further consideration.