Project Description

As per pioneering educational psychologist Prof. Benjamin S. Bloom, the most effective form of education is one-on-one tutoring. Despite this, millions of students globally lack access to high-quality one-on-one tutoring due to shortages of expert and qualified tutors. While Generative Artificial Intelligence (GenAI), powered by Large Language Models (LLMs), holds promise, unguided use often fails to support meaningful learning. To address these gaps, the proposed Ph.D. project, EduPilot, combines LLMs with established pedagogical principles to empower tutors toward human-AI collaborative tutoring, without replacing them. The multidisciplinary project pursues three core objectives. First, it develops a robust evaluation framework to measure and track the quality of tutor responses in long tutoring conversations. This framework goes beyond simple correctness to assess how well guidance aligns with a student's knowledge, misconceptions, language, and cultural context. Second, it builds an adaptive conversational AI tutor capable of generating pedagogically sound, context-aware responses. By continuously monitoring a student's evolving understanding, the adaptive tutor can suggest interventions that are neither too easy nor too difficult, supporting optimal learning. Third, it integrates these capabilities into a real-time tutor co-pilot, which ranks multiple AI-generated options and presents the best choices to human tutors, allowing them to adopt, modify, or compose responses as needed. Expected outcomes include a validated evaluation toolkit, an adaptive tutoring model, and a Tutor Co-Pilot interface, all released as open-source tools for educators and researchers. These methods will be validated using real-world classroom data and randomized controlled trials to ensure genuine learning gains. In practice, EduPilot enhances teaching quality, supports both novice and expert tutors, and adapts to diverse student needs across languages and cultures.

In summary, EduPilot envisions a future where AI amplifies the impact of human tutors in one-on-one tutoring, making personalized education accessible, effective, and inclusive for every learner.