AI Tutor Uses Reinforcement Learning for Sustainable Online Education

Online education offers unparalleled scalability and accessibility, yet it frequently struggles with low learner engagement and suboptimal long-term learning effectiveness. To address these critical issues, a new reinforcement learning-based model called AI Tutor has been developed, aiming to foster sustainable learning by optimizing both immediate and enduring educational outcomes. AI Tutor operates on two fronts: in the short term, it applies cognitive theory to guide learners through a balanced curriculum of acquiring new knowledge and reinforcing previously learned material. For the long term, the model actively tracks and predicts learner engagement, informing strategies designed to maintain motivation and significantly reduce dropout rates. This dual focus allows AI Tutor to provide highly personalized guidance that supports both effective knowledge acquisition and sustained participation. Empirical evaluations conducted on a massive dataset of 23 million learning records from 33,700 learners demonstrated AI Tutor's consistent superiority. It outperformed state-of-the-art baselines across key metrics including engagement, knowledge retention, and final learning outcomes. Further analysis of learning paths revealed the model's ability to adapt its strategies to diverse learner profiles, offering adaptive and human-centered support.