CoLearn adapts tutoring questions to each learner's needs in real time
CoLearn: An Agentic Tutor that Learns its Learner in a Human--AI Co-Learning Loop
Computation and Language
Summary
Many tutoring tools give the same questions to everyone and only notice right or wrong answers. The authors created CoLearn, a tutor that remembers what you know and what mistakes you tend to make. It uses this memory to ask you better, personalized questions that help you learn more effectively. Tests showed people liked these personalized questions better and the system's guesses about what you know get more accurate over time.
What this means in practice
- •For online course platforms: Create personalized practice questions for learners based on ongoing mastery estimation and misconceptions.$Commercial implications: Enables development of AI-powered adaptive tutoring systems that improve learner engagement and outcomes for commercial e-learning products.
- •For language learning app developers: Improve language practice exercises by targeting topics where the user struggles and adapting questions dynamically.
Authors
Kailai He, Zhihao Wu, Linhai Zhang, Runcong Zhao, Yulan He, Jiazheng Li
Abstract
Good tutoring adapts to the individual: it tracks what a learner knows, notices why they go wrong, and asks the next question that will help most. Most deployed tutoring tools instead serve fixed item banks and treat a wrong answer as a single bit of signal. We present CoLearn, an interactive, agentic tutor that supports an iterative tutoring loop: the learner practises, and the system builds an evidence-grounded memory of the learner's mastery and misconceptions. This memory is updated as evidence accumulates and is used to generate the next personalised question. CoLearn has three components: (i) a persistent learner-state memory that updates per-topic mastery with a soft-evidence variant of Bayesian Knowledge Tracing, where a large language model acts as a continuous observation function; (ii) adaptive question generation that targets the learner's weakest topic and recurring misconceptions; and (iii) an evidence view that makes personalisation visible and testable through live progress visualisation and blind A/B comparison. In blind A/B evaluation, questions conditioned on this memory are preferred over non-personalised ones 68-69% of the time, and in persona simulations with hidden ground-truth mastery the agent's belief converges toward the learner's true mastery.