Human-AI Co-Creation of Worked Examples for Programming Classes
Project Overview
This document explores the application of generative AI, particularly large language models (LLMs), in enhancing programming education through the generation of worked examples. It introduces an authoring system that promotes collaboration between humans and AI in creating code explanations, effectively addressing the challenge faced by instructors who struggle to provide comprehensive line-by-line clarifications due to time constraints. The study assesses the efficacy of AI-generated explanations in comparison to those created by expert educators, highlighting that while AI can serve as a valuable tool for generating initial content, the involvement of human oversight is crucial to ensure the quality and accuracy of the educational material. Overall, the findings underscore the potential of generative AI in supporting programming instruction while emphasizing the importance of maintaining a balance between automated assistance and human expertise.
Key Applications
Worked Example Authoring Tool (WEAT)
Context: Programming classes, specifically for Java language instruction; target audience includes programming instructors and students.
Implementation: Instructors provide code examples and problem statements; the AI generates explanations for code lines, which instructors can edit or accept.
Outcomes: Reduced time in creating interactive worked examples, higher engagement in coding tasks, and improved quality of explanations as perceived by both instructors and students.
Challenges: AI-generated explanations may still contain inaccuracies and lack the contextual understanding that human instructors provide.
Implementation Barriers
Quality Control
AI-generated explanations may not always meet the quality expectations of instructors and students, requiring human intervention for editing and approval.
Proposed Solutions: Instructors can review and edit AI-generated content to ensure clarity and correctness before presenting it to students.
User Interface
Instructors faced issues with the interface, such as dialogs blocking the main editing window and lack of features to merge or reorder explanation fragments.
Proposed Solutions: Future iterations of the tool should address these interface usability issues, enhancing the user experience for instructors.
Project Team
Mohammad Hassany
Researcher
Peter Brusilovsky
Researcher
Jiaze Ke
Researcher
Kamil Akhuseyinoglu
Researcher
Arun Balajiee Lekshmi Narayanan
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Mohammad Hassany, Peter Brusilovsky, Jiaze Ke, Kamil Akhuseyinoglu, Arun Balajiee Lekshmi Narayanan
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