Question Personalization in an Intelligent Tutoring System
Project Overview
The document examines the application of generative AI within intelligent tutoring systems (ITS), highlighting its role in personalizing educational experiences through tailored question phrasing based on individual student proficiency levels. The findings indicate that such personalized adaptations can lead to significantly improved learning outcomes, especially for beginner students. An A/B test conducted on the Korbit platform evidenced that students exposed to personalized question variants outperformed their peers who received standard questions, underscoring the effectiveness of this approach. Overall, the use of generative AI in education demonstrates a promising potential to enhance student engagement and learning efficacy by providing customized support that aligns with each learner's unique capabilities.
Key Applications
Korbit - Intelligent Tutoring System
Context: The system targets students at various levels of subject proficiency, providing personalized question variants to enhance learning outcomes.
Implementation: Students interact with the Korbit platform, where questions are adapted to their proficiency levels using a logistic regression model to predict success. An A/B test was conducted over two months with 400 students.
Outcomes: Students receiving personalized questions showed improved solution acceptance rates, reduced ultimate failure rates, and lower skip rates, indicating higher engagement and better learning gains.
Challenges: The main challenge was ensuring that the variants maintained the integrity and educational value of the original questions, as well as scalability in automating question generation.
Implementation Barriers
Technical Barrier
The challenge of automating the creation of question variants to ensure scalability and maintain educational quality.
Proposed Solutions: Future work aims to develop more sophisticated approaches to variant assignment and automate the question variant generation process.
Project Team
Sabina Elkins
Researcher
Robert Belfer
Researcher
Ekaterina Kochmar
Researcher
Iulian Serban
Researcher
Jackie C. K. Cheung
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Sabina Elkins, Robert Belfer, Ekaterina Kochmar, Iulian Serban, Jackie C. K. Cheung
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