Skip to main content Skip to navigation

3DG: A Framework for Using Generative AI for Handling Sparse Learner Performance Data From Intelligent Tutoring Systems

Project Overview

The document presents the 3DG framework, an innovative approach leveraging generative AI models, specifically Generative Adversarial Networks (GAN) and Generative Pre-trained Transformers (GPT), to tackle data sparsity challenges in Intelligent Tutoring Systems (ITS). By integrating tensor factorization with generative models, the framework significantly enhances data imputation and augmentation, thereby improving learner modeling and personalizing educational technologies. Its primary objective is to capture a wide array of learning patterns and create scalable simulations that refine the learner model effectively. The findings indicate that GAN outperforms GPT-4 in this educational context, showcasing its potential to revolutionize personalized learning experiences and optimize educational outcomes through advanced AI techniques. Overall, the 3DG framework stands as a promising advancement in the application of generative AI within education, aiming to foster individualized learning pathways and improve engagement and achievement among students.

Key Applications

3DG framework for data imputation and augmentation

Context: Intelligent Tutoring Systems (ITS) focusing on reading comprehension for adult learners

Implementation: Combines tensor factorization with GAN and GPT to create a 3-dimensional tensor and simulate learning performance data.

Outcomes: Enhanced learner modeling and personalized instruction through improved data quality and scalability.

Challenges: Data sparsity in learning performance data; limitations of GPT in numerical computations.

Implementation Barriers

Technical Barrier

Data sparsity leads to biased analysis and modeling of learner performance.

Proposed Solutions: Utilization of tensor factorization for data imputation and augmentation using generative models.

Computational Barrier

Insufficient capability of GPT for deep learning tasks involving numerical computations.

Proposed Solutions: Exploration of integrating GAN with GPT to enhance computational capabilities.

Project Team

Liang Zhang

Researcher

Jionghao Lin

Researcher

Conrad Borchers

Researcher

Meng Cao

Researcher

Xiangen Hu

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Liang Zhang, Jionghao Lin, Conrad Borchers, Meng Cao, Xiangen Hu

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies