LangTutor is an agentic AI tutor that decides what to teach next after each learner response, combining pedagogical methods such as scaffolding, hint-based guidance, retrieval practice, spaced repetition, and targeted corrective feedback. It also personalizes instruction to the learner’s background, preferences, goals, and interests to improve engagement. This matters because unrestricted AI assistance can undermine learning: a recent study found that students using a plain GPT-4 tutor performed worse once the AI was removed, while a hint-based version avoided that harm. LangTutor therefore places the model inside a structured tutoring system that adapts both the learning sequence and feedback while withholding answers when appropriate. Using German language learning as one test case, we ask whether this combination of adaptive pedagogy and personalization improves learning outcomes and engagement over generic AI tutoring.
Team: Team: Sumaiya Karim Katha, Ezharuddin Jubaer, Tanvir Mahmud, Mir Sazzat Hossain, Amin Ahsan Ali, PhD
