Generative artificial intelligence produces differential academic outcomes across socioeconomic, linguistic, and disability contexts
Generative artificial intelligence tools in education offer significant potential to advance equitable learning, yet their impacts may vary across socially and structurally diverse student groups. While such technologies can democratize access to academic support, they may also reproduce existing inequalities if access, skills, and benefits are unevenly distributed. This study investigates how ChatGPT influences academic outcomes across key equity dimensions, including socioeconomic status (SES), language background, and disability status. A multigroup field experiment was conducted with 240 university students from eight countries. Participants were stratified by SES, language proficiency, and disability status. Each participant completed academic writing tasks under two conditions—with and without ChatGPT access—in a counterbalanced design. Outcomes assessed included writing quality, time on task, self-efficacy, and learning retention. Results demonstrate significant differential effects. Low-SES students exhibited greater improvements in writing quality compared to high-SES students, with digital literacy mediating a substantial portion of the gains. Non-native developing speakers achieved the highest performance improvements but showed lower error detection and reduced retention. Students with disabilities reported higher self-efficacy gains but required more time to complete tasks. Intersectional analysis reveals that students with multiple disadvantages experienced the largest immediate benefits, alongside increased risks to long-term learning outcomes. Generative AI has uneven and context-dependent impacts on educational equity. While it can enhance short-term academic performance, it may also reinforce structural disparities without supportive interventions. Sustainable and inclusive AI integration requires addressing gaps in digital literacy, evaluative capacity, and accessibility to ensure equitable and meaningful learning outcomes.
Authors
- Mayadhar Sethy (ORCID: https://orcid.org/0009-0002-7413-3597)
Institutions
- Centre For Development Studies (IN)
Publication Details
- Journal
- Discover Sustainability
- Published
- 2026-09-06
- DOI
- https://doi.org/10.1007/s43621-026-04645-0
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 0.00