ENTERPRISE AI ANALYSIS

A Generalized Apprenticeship Learning Framework for Capturing Evolving Student Pedagogical Strategies

This paper introduces THEMES, a novel Apprenticeship Learning (AL) framework designed to capture dynamically evolving student pedagogical strategies within Intelligent Tutoring Systems (ITSs). Unlike traditional Reinforcement Learning (RL) or Deep Reinforcement Learning (DRL) methods that struggle with sample inefficiency and reward function design, THEMES infers expert students' underlying reward functions from limited demonstrations. Its key innovation lies in time-aware hierarchical sub-trajectory partitioning, which models how student intentions and learning goals change over time. By combining this with an Expectation-Maximization(EM)-Energy-based Distribution Matching (EDM) approach, THEMES induces effective pedagogical policies even with a small number of expert demonstrations. Evaluated against six state-of-the-art baselines, THEMES consistently outperforms them in predicting student pedagogical decisions, demonstrating superior AUC (0.899) and Jaccard (0.653) scores using only 18 trajectories. This framework presents a powerful, sample-efficient alternative for developing adaptive ITSs that can respond to the dynamic nature of human learning.