Leveraging Pedagogical Theories to Understand Student Learning Process with Graph-based Reasonable Knowledge Tracing

Jiajun Cui,Hong Qian,Bo Jiang,Wei Zhang

2024-06-07

Abstract:Knowledge tracing (KT) is a crucial task in intelligent education, focusing on predicting students' performance on given questions to trace their evolving knowledge. The advancement of deep learning in this field has led to deep-learning knowledge tracing (DLKT) models that prioritize high predictive accuracy. However, many existing DLKT methods overlook the fundamental goal of tracking students' dynamical knowledge mastery. These models do not explicitly model knowledge mastery tracing processes or yield unreasonable results that educators find difficulty to comprehend and apply in real teaching scenarios. In response, our research conducts a preliminary analysis of mainstream KT approaches to highlight and explain such unreasonableness. We introduce GRKT, a graph-based reasonable knowledge tracing method to address these issues. By leveraging graph neural networks, our approach delves into the mutual influences of knowledge concepts, offering a more accurate representation of how the knowledge mastery evolves throughout the learning process. Additionally, we propose a fine-grained and psychological three-stage modeling process as knowledge retrieval, memory strengthening, and knowledge learning/forgetting, to conduct a more reasonable knowledge tracing process. Comprehensive experiments demonstrate that GRKT outperforms eleven baselines across three datasets, not only enhancing predictive accuracy but also generating more reasonable knowledge tracing results. This makes our model a promising advancement for practical implementation in educational settings. The source code is available at <a class="link-external link-https" href="https://github.com/JJCui96/GRKT" rel="external noopener nofollow">this https URL</a>.

Artificial Intelligence,Computers and Society,Machine Learning

What problem does this paper attempt to address?

The paper aims to address the issue of irrationality in current Deep Learning Knowledge Tracing (DLKT) methods for dynamic knowledge mastery tracking. Specifically, many existing DLKT models focus too much on predicting students' answer performance while neglecting the core goal of tracking students' knowledge mastery status. This leads to these models being unable to reasonably represent the process of knowledge mastery changes, making it difficult for educators to understand and apply their results.

To solve these problems, the authors propose a Graph Neural Network (GNN)-based Rational Knowledge Tracing method (GRKT). By utilizing GNN, this method can deeply explore the mutual influence between knowledge concepts, thereby more accurately representing how knowledge mastery evolves throughout the learning process. Additionally, the authors propose a fine-grained and psychologically consistent three-stage modeling approach, including knowledge retrieval, memory reinforcement, and knowledge learning/forgetting stages, to achieve a more reasonable knowledge tracing process. Experimental results show that GRKT outperforms 11 baseline models on 3 datasets, not only improving prediction accuracy but also generating more reasonable knowledge tracing results, making the model promising for practical application in educational scenarios.