.png){kind=icon}
.png){kind=icon}
Return
AI-driven educational reform and persistent inequality: unpacking classroom-level impacts in China's evolving school system
.png){kind=icon}
PRE
AI
S
A
B
DOI：10.1108/AEDS-01-2026-0006.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
PurposeThis study investigates how artificial intelligence (AI)-driven education reforms in China shape instructional quality, teacher workload and student learning experiences across urban, rural and migrant-serving schools, assessing whether these reforms mitigate or exacerbate existing inequalities.Design/methodology/approachA mixed-methods design integrates multilevel modeling, structural equation modeling and a causal difference-in-differences approach. Administrative census data were combined with platform-generated logs and weighted survey responses from over 18,000 teachers and 12,500 students. The staggered provincial rollout of AI platforms enabled quasi-experimental identification, supported by event-study pre-trend tests and robustness checks using modern staggered-adoption estimators. Latent constructs were validated through reliability and measurement invariance tests, and qualitative insights were incorporated through a systematic coding protocol.FindingsAI integration improves student engagement and learning outcomes on average, yet these gains are unevenly distributed. Urban, well-resourced schools show significant improvements, whereas rural and migrant-serving schools experience limited benefits. Causal estimates reveal that school-level resources moderate the strength of AI's impact. Mediation analyses show that improvements in instructional quality and engagement primarily drive gains in advantaged contexts. Infrastructure gaps and insufficient teacher preparation restrict these pathways in disadvantaged schools, thereby widening pre-existing disparities.Originality/valueThis study offers a rare classroom-level, mechanism-focused assessment of China's AI reforms, linking quasi-experimental evidence with latent-variable modeling to illuminate both the size and structure of unequal AI effects. It challenges deterministic narratives of technological progress and underscores the centrality of institutional capacity in shaping equitable educational innovation.
Keywords:
AI in education
Educational inequality
Digital divide
China education reform
Classroom-level impact
Technology integration
Structural equation modeling
Difference-in-differences
Journal
A
IF:
2
Papers:
73
Citations:
0
Organization
Cited Papers