Trust in Generative AI and Critical Thinking Among College Students: A Structural Equation Modeling Analysis Based on Perceived Usefulness, Cognitive Offloading, and AI Self-Efficacy

Abstract

This study examines how college students’ trust in generative artificial intelligence (GenAI) is associated with critical thinking engagement in GenAI-supported learning through multiple psychological mechanisms. Drawing on Distributed Cognition Theory (DCT), we propose and test a structural model including five constructs: trust in AI, perceived usefulness, cognitive offloading, AI self-efficacy, and critical thinking. Drawing on survey data from a convenience sample of college students who engaged in learning activities using GenAI tools (e.g., ChatGPT), we assessed hypothesized relationships using structural equation modeling (SEM). The results show that trust in AI is positively associated with perceived usefulness and AI self-efficacy, and both constructs are positively associated with critical thinking. In contrast, cognitive offloading is negatively associated with critical thinking, indicating a potential constraining pathway through which stronger reliance on GenAI may be associated with lower engagement in higher-order thinking. Overall, the findings suggest that trust in GenAI is linked to critical thinking through both supportive and constraining pathways. This study extends DCT in GenAI-supported learning by showing that trust in AI is not uniformly beneficial. The findings highlight the importance of educational approaches that cultivate students’ efficacy beliefs and strategic use of GenAI while discouraging overreliance that reduces cognitive engagement.