Answers are grounded in this paper and cite their sources.
The rapid integration of large language models (LLMs) into higher education has intensified debate over how AI-assisted learning may be associated with students’ cognitive development. Although prior studies have examined the benefits and risks of AI use, less attention has been paid to the distinction between frequent AI usage as a learning behavior and large language model dependence (LLMD) as a habitual reliance on LLMs for cognitive and academic tasks. This study therefore distinguishes frequent AI usage as a learning behavior from LLMD as a habitual reliance on LLMs when completing cognitive and academic tasks. Drawing on the Job Demands-Resources (JD-R) model and Conservation of Resources theory, this study examined how AI usage is associated with university students’ critical thinking through LLMD, metacognitive awareness, and AI literacy. Survey data were collected from 878 Chinese university students and analyzed using partial least squares structural equation modeling. The results showed that AI usage was positively associated with LLMD, which represented the strongest direct association in the model, while LLMD was negatively associated with metacognitive awareness, AI literacy, and critical thinking. The mediation analysis further indicated a significant negative total indirect association between AI usage and critical thinking through LLMD and lower levels of metacognitive awareness and AI literacy. These findings contribute to the literature by clarifying the distinction between AI usage and LLMD and by identifying metacognitive awareness and AI literacy as cognitive resources associated with critical thinking in AI-assisted learning contexts. Practically, the study suggests that universities should move beyond simple restriction-oriented responses and pay greater attention to the quality of students’ AI engagement, including metacognitive monitoring, AI literacy, and awareness of dependence-related patterns.
No takes yet. Share an insight, caveat, or question.
Ouyang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: