Date
Publisher
arXiv
This study investigates how LLMs, specifically GPT-3.5 and GPT-4, can develop tailored questions for Grade 9 math, aligning with active learning principles. By utilizing an iterative method, these models adjust questions based on difficulty and content, responding to feedback from a simulated 'student' model. A novel aspect of the research involved using GPT-4 as a 'teacher' to create complex questions, with GPT-3.5 as the 'student' responding to these challenges. This setup mirrors active learning, promoting deeper engagement. The findings demonstrate GPT-4's superior ability to generate precise, challenging questions and notable improvements in GPT-3.5's ability to handle more complex problems after receiving instruction from GPT-4. These results underscore the potential of LLMs to mimic and enhance active learning scenarios, offering a promising path for AI in customized education. This research contributes to understanding how AI can support personalized learning experiences, highlighting the need for further exploration in various educational contexts
Study at a glance
- Question: How accurately can large language models generate customized math test questions and teach a simulated student model?
- Setting and data: Grade 9 mathematics curriculum; 90 questions per subject across Numbers and Financial Mathematics, plus 50 algebra word problems.
- Method: Model evaluation and simulated experiment comparing teacher and student model interactions under varying instructional conditions.
Why it matters
Personalized test design requires adaptive learning tools that align with active learning principles by tailoring content to individual student needs and targeting areas where students struggle.
Key findings
- GPT-4 generated correct questions more accurately than GPT-3.5 and fine-tuned GPT-3.5, achieving 90% in Numbers and 93% in Financial Mathematics compared to 50% and 70% to 76%.
- Teaching GPT-3.5 with example problems and answers increased its accuracy on algebra word problems across all difficulty levels, reaching up to 51.43% on level 4 problems.
- Providing detailed explanations alongside example answers did not produce additional gains in GPT-3.5's problem-solving accuracy, with scores remaining between 34.29% and 37.14%.
Summary drafted with AI from the full paper and may contain errors. See the full paper for details.
What is the application?
Who age?
Why use AI?
Study design
