How AI shapes student agency and educational stratification: a qualitative interpretive meta-synthesis

Existing research on artificial intelligence in education has largely addressed technical applications and learning outcomes while leaving sociological dimensions of AI-mediated stratification and student agency inadequately theorised. Drawing on Bourdieu's cultural capital framework, digital divide scholarship, and Yosso's Community Cultural Wealth model, this qualitative interpretive meta-synthesis examines how AI-mediated learning environments interact with established stratification mechanisms and transform student agency. Systematic searches across Scopus, Web of Science Core Collection, and ERIC identified five qualitative and mixed-methods studies encompassing 3,849 students across the United States, Norway, Israel, and China; thematic analysis followed Braun and Clarke's framework with four-analyst triangulation. Five mechanisms emerged through which AI technologies simultaneously reproduce traditional educational inequalities while generating alternative stratification forms: cultural capital mobilization through resistant, communal, and creative capital; differential engagement patterns across four distinct agency expressions; digital divide persistence and evolution; trust calibration in human-AI interaction; and educational equity implications with career dimensions. Students from underserved communities demonstrated sophisticated algorithmic bias recognition, yet gender-differentiated engagement patterns and socioeconomic disparities indicate emergent stratification with professional consequences. Lower AI trust paradoxically correlated with stronger educational outcomes, tentatively suggesting that healthy skepticism promotes more agentic learning relationships. These findings indicate that AI-mediated stratification operates through qualitative differences in how students position themselves relative to algorithmic systems and mobilize cultural resources, rather than through differential access alone, with implications for how educational institutions conceptualise equity-oriented AI integration.