From Recall to Reasoning: Automated Question Generation for Deeper Math Learning through Large Language Models
Project Overview
The document discusses the emerging role of generative AI (GenAI) in enhancing educational content, particularly in advanced mathematics. It highlights the capability of GenAI to automate question generation, thereby aiding educators in creating relevant and tailored math problems with minimal input. However, the research identifies that the quality of these outputs can be significantly improved when contextual information is provided. Despite these advancements, challenges persist, especially concerning the cognitive depth of the questions generated and the effective integration of GenAI tools within educational environments. To address these issues, the study aims to develop a framework that aligns the question generation process with established educational taxonomies, ultimately enhancing the relevance and depth of the questions produced. The findings suggest that while GenAI has the potential to transform educational practices, careful consideration of its application and integration is essential to fully realize its benefits in learning contexts.
Key Applications
Automated Question Generation using Large Language Models
Context: Educational context focused on advanced mathematics learning, targeting educators and students in math courses.
Implementation: The study employed two phases: initially exploring GenAI's capabilities for question generation and later developing a context-aware framework integrating retrieval-augmented generation with Webb's Depth of Knowledge.
Outcomes: The framework improved the relevance and cognitive depth of generated questions, particularly for higher-order thinking skills. Educators reported that AI can automate parts of their workload and enhance student learning experiences.
Challenges: GenAI sometimes produces irrelevant or overly simplistic questions and can struggle with higher cognitive levels, leading to inaccuracies in generated content.
Implementation Barriers
Technical Barrier
GenAI struggles with the cognitive depth of questions, particularly at mid-level complexities, leading to inconsistencies in question quality.
Proposed Solutions: Adopting frameworks like Webb's Depth of Knowledge to inform question generation and integrating retrieval-augmented generation to provide contextual grounding.
Implementation Barrier
Educators express reservations about the relevance and quality of AI-generated content, leading to underutilization of GenAI tools.
Proposed Solutions: Providing training and resources for educators to effectively integrate GenAI tools into their workflows.
Project Team
Yongan Yu
Researcher
Alexandre Krantz
Researcher
Nikki G. Lobczowski
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Yongan Yu, Alexandre Krantz, Nikki G. Lobczowski
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