Teacher-in-the-Loop Learning Analytics for LLM-Enhanced Intelligent Tutoring Systems

The rapid emergence of Generative AI (GenAI), which gained momentum following the introduction of transformer architecture by Vaswani et al. (2017) , has expanded the scope of what learning analytics (LA) can achieve. Traditional approaches to LA have emphasized the importance of understanding learners and their surrounding contexts ( Ferguson, 2012 ). Yet in an increasingly AI-mediated world, there is a growing need to examine not only learners themselves, but the dynamic context of student-AI interactions ( Noy & Zhang, 2023 ). Insights about how students appropriate, resist, or adapt to GenAI-based agents are integral to their continued learning process, particularly as we see the GenAI become increasingly integrated into educational tools such as intelligent tutoring systems (ITS) ( Schmucker et al., 2023 ). Consequently, this opens an opportunity for LA to evolve beyond tracking traditional learning behaviors, toward analyzing how students seek support and construct meaning through interactions with GenAI-driven ITS.