Abdelghani, Kaiser & Murayama (2026) — cs.CY 📄 Full text (arXiv)
Abdelghani, Kaiser & Murayama (2026) trace how middle and high school students' interactions with AI math tutors evolve over time, identifying a trajectory from superficial prompting ('tell me the answer') to what they term 'epistemic proactivity' — the active, self-directed pursuit of conceptual understanding through AI dialogue. This developmental framework is a significant contribution to ai-literacy research, as it suggests that productive AI use is not a binary skill but a capacity that matures through scaffolded practice. The temporal analysis reveals that students who receive metacognitive prompting show faster transitions to epistemic proactivity, directly linking to metacognition and self-regulated-learning theories. The study has implications for stem-education and k-12 classrooms, where AI tools are increasingly used as math tutors. It challenges the assumption that students intuitively know how to learn with AI, arguing instead that deliberate scaffolding is essential — a finding that resonates across student-experience research in AI-mediated learning environments.
Related Pages
- representation-robustness-llm-math-problem-solving — Representation Robustness under Executable Reasoning Constra