The Agency Gap in AI-Assisted Higher Education: When Learner Control Shapes Reflective Practice and Critical Thinking

Abstract: The proliferation of generative artificial intelligence (GenAI) in higher education has intensified questions about student agency, cognitive effort, and the conditions under which AI-supported learning fosters deep engagement rather than passive consumption. This study examines the relationships between perceived human-AI agency—operationalized as students' reported initiative, monitoring of AI outputs, and decision-making control—and reflective engagement, self-reported critical thinking, and academic self-concept. Drawing on survey data from 309 university students across UK-based (n=145) and China-based (n=164) higher education contexts, partial least squares structural equation modeling (PLS-SEM) revealed that perceived human-AI agency positively predicted reflective engagement, which in turn predicted self-reported critical thinking in both samples. However, context-specific patterns emerged: reflection was associated with academic self-concept only in the UK sample, while AI literacy moderated the agency-reflection relationship exclusively in the China sample. These findings theorize an "agency gap," suggesting that students who retain perceived control when using AI also report stronger reflective and critical orientations. The study underscores the need for context-sensitive pedagogical strategies that prioritize learner agency, metacognitive scaffolding, and responsible AI integration.

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.