Revisiting Generalization Theory in the Age of Educational AI - Implications for Empirical Educational Research

The integration of Artificial Intelligence (AI) in education has rapidly advanced in recent years, offering personalized learning experiences, real-time feedback, and predictive analytics. However, research on educational inequities reinforced by the application of AI-driven tools raise important questions about reliability and adaptability across diverse educational settings. This paper revisits Generalizability Theory (G-Theory), originally developed by Cronbach and colleagues, and examines its relevance in the age of AI-driven education. It is argued that it is timely to use G-Studies and Decision (D-) Studies that could facilitate the effectiveness of AI tools across different contexts and ameloriate education inequities. Key use cases, such as AI-based assessments and predictive analytics are discussed, and limitations as well as implications for study designs highlighted. The paper concludes with suggestions for future research.