Student Data Paradox and Curious Case of Single Student-Tutor Model: Regressive Side Effects of Training LLMs for Personalized Learning
Shashank Sonkar, Naiming Liu, Richard G. Baraniuk
TL;DR
The paper investigates the Student Data Paradox, showing that training LLMs on student–tutor data to capture learner behavior can undermine the model's own factual knowledge and reasoning. It formalizes a unified-model perspective, arguing that separating student and tutor representations leads to a proliferation of models rather than a single coherent system. To address the paradox, it introduces hallucination tokens embedded in training data to mark incorrect content, and demonstrates substantial but incomplete recovery across benchmarks like ARC, TruthfulQA, HaluEval, and MemoTrap using CLASS-derived dialogues. The findings highlight a crucial balance between accurately modeling student cognition and preserving educational accuracy, indicating a need for further methods to reconcile these competing goals in personalized learning with LLMs.
Abstract
The pursuit of personalized education has led to the integration of Large Language Models (LLMs) in developing intelligent tutoring systems. To better understand and adapt to individual student needs, including their misconceptions, LLMs need to be trained on extensive datasets of student-tutor dialogues. Our research uncovers a fundamental challenge in this approach: the ``Student Data Paradox.'' This paradox emerges when LLMs, trained on student data to understand learner behavior, inadvertently compromise their own factual knowledge and reasoning abilities. We investigate this paradox by training state-of-the-art language models on student-tutor dialogue datasets and evaluating their performance across multiple benchmarks. These benchmarks assess various aspects of language model capabilities, including reasoning, truthfulness, and common sense understanding. Our findings reveal significant declines in the models' performance across these diverse benchmarks, indicating a broad impact on their capabilities when trained to model student behavior. Our research makes two primary contributions: (1) empirical demonstration of the Student Data Paradox through quantitative analysis of model performance, and (2) introduction of ``hallucination tokens'' as a mitigation strategy. These tokens, while improving performance, highlight the persistent challenge of balancing accurate student behavior modeling with maintaining the LLM's integrity as an educational tool. This study emphasizes the need for innovative solutions to reconcile the conflicting goals of faithfully understanding diverse student cognition while preserving the model's ability to provide accurate information and guidance.