Educational Inequalities and Generative AI: A Focus on Language Diversity

ongoing

ChatEQUITY investigates the impact of generative AI (GenAI) on educational inequalities amid growing societal linguistic diversity. The central hypothesis is that although multilingual GenAI can overcome linguistic barriers, mere access to it is insufficient because competent use is essential. Moreover, since GenAI language models are trained on and steered by human language—which reflects social class, ethnic, and gender differences—GenAI might reproduce biases and educational disparities. Research on this is lacking due to underlying limitations in the prevailing concepts, methods and data. ChatEQUITY addresses these with three objectives: 1. To introduce prompting (l)iteracy (PIL) as a novel concept for competent human-GenAI interactions and develop a pioneering assessment instrument (PIL-AI) for its measurement. 2. To identify shifts from traditional digital divides to GenAI divides—social disparities in GenAI access, use and skills —with innovative data from 3,000 students and 700 teachers across 60 high schools. 3. To evaluate the impact of GenAI use and PIL-education on educational inequalities by examining their effects on teaching and learning with three refined RCTs. ChatEQUITY enables researchers to move beyond merely observing the effects of GenAI, toward understanding why and how GenAI affects life outcomes. Because PIL-AI uniquely combines an authentic task-based approach with scalability, this tool marks a paradigm shift in the assessment of competencies that currently depends on self-reported measures. The open data on GenAI divides establish the groundwork for research into the interplay of sociological, pedagogical, linguistic, and technological factors. Finally, by refining traditional binary RCTs (moving beyond simple comparisons between tech and non-tech users), we recenter education in educational technology research. In essence, ChatEQUITY aligns with UNESCO’s call to promote equity through GenAI capacity-building in education and research.