Data Augmentation for Sparse Multidimensional Learning Performance Data Using Generative AI

2024-09-24

arXiv.org

AI Summary

In intelligent tutoring systems, highly sparse student response data (80–90% missingness) severely impairs knowledge tracing accuracy and constrains theoretical modeling. Method: This paper proposes a dynamic knowledge state representation based on a three-way tensor (learner × question × attempt), integrating tensor decomposition with generative AI for synergistic data augmentation. We introduce a novel GAN/GPT dual-path generation framework—the first to couple tensor decomposition with generative modeling—and propose a joint optimization mechanism that simultaneously enforces knowledge tracing task constraints, missing-value imputation, and clustering-aware pattern generation. Results: Evaluated on the Adult Reading Comprehension dataset, tensor decomposition significantly improves knowledge tracing accuracy. GAN-generated responses exhibit greater stability and lower bias than GPT-generated ones, and their statistical consistency is preserved as sample size increases.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Matrix & Tensor Methods
• Knowledge Representation and Reasoning: Knowledge Acquisition

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Learning performance data describe correct and incorrect answers or problem-solving attempts in adaptive learning, such as in intelligent tutoring systems (ITSs). Learning performance data tend to be highly sparse (80%(sim)90% missing observations) in most real-world applications due to adaptive item selection. This data sparsity presents challenges to using learner models to effectively predict future performance explore new hypotheses about learning. This article proposes a systematic framework for augmenting learner data to address data sparsity in learning performance data. First, learning performance is represented as a three-dimensional tensor of learners' questions, answers, and attempts, capturing longitudinal knowledge states during learning. Second, a tensor factorization method is used to impute missing values in sparse tensors of collected learner data, thereby grounding the imputation on knowledge tracing tasks that predict missing performance values based on real observations. Third, a module for generating patterns of learning is used. This study contrasts two forms of generative Artificial Intelligence (AI), including Generative Adversarial Networks (GANs) and Generate Pre-Trained Transformers (GPT) to generate data associated with different clusters of learner data. We tested this approach on an adult literacy dataset from AutoTutor lessons developed for Adult Reading Comprehension (ARC). We found that: (1) tensor factorization improved the performance in tracing and predicting knowledge mastery compared with other knowledge tracing techniques without data augmentation, showing higher relative fidelity for this imputation method, and (2) the GAN-based simulation showed greater overall stability and less statistical bias based on a divergence evaluation with varying simulation sample sizes compared to GPT.

Problem

• Intelligent Tutoring Systems
• Sparse Data
• Predictive Analytics

Innovation

• Generative AI
• GANs vs GPT
• Sparse Data Imputation