Bridging Learning Analytics and Generative AI through Retrieval-Augmented Generation: A Conceptual Framework for Self-Regulated Learning Support

Abstract

Learning Analytics (LA) extracts meaningful behavioral patterns from digital learning environments; however, the interpretation and application of its outputs remain heavily dependent on data science expertise. While Generative AI (GenAI) offers real-time dialogue and content generation capabilities, it is constrained by insufficient domain knowledge and a lack of individualized learner context. This paper proposes a conceptual framework integrating LA, Retrieval-Augmented Generation (RAG), and GenAI, with the core innovation of leveraging learners' learning trajectories as a RAG knowledge source, thereby enabling GenAI to generate personalized recommendations grounded in authentic learning contexts. Grounded in Zimmerman's three-phase Self-Regulated Learning model, this paper delineates how the integrated framework provides differentiated support across the Forethought, Performance, and Self-Reflection phases, forming a dynamic learning support loop. Future work are further discussed.