ChatGPT in the classroom. Exploring its potential and limitations in a Functional Programming course
Project Overview
The document examines the application of ChatGPT, a generative AI tool, in educational settings, specifically within a Functional Programming course at POLITEHNICA University of Bucharest. It showcases the tool's ability to generate code, assist with homework, and conduct code reviews, demonstrating its potential to enhance learning. However, the authors also identify challenges, including issues with the accuracy and readability of AI-generated code, as well as concerns regarding academic integrity and the implications for teaching practices. Survey results reveal that while a significant number of students utilize generative AI, many express skepticism about its effectiveness in improving their understanding of the material. In light of these findings, the authors advocate for a thoughtful integration of AI into educational frameworks, emphasizing the importance of addressing ethical considerations while leveraging the benefits of technology in learning environments.
Key Applications
ChatGPT for generating code and code reviews
Context: Undergraduate students in a Functional Programming course
Implementation: ChatGPT was used to solve coding assignments and provide code reviews for student submissions.
Outcomes: ChatGPT provided correct answers in 68% of cases, improved to 86% with follow-up questions. It also successfully reviewed code with 77% accuracy.
Challenges: 43% of correct solutions were illegible or inefficient. Many students struggle to write effective tests for AI-generated code.
Implementation Barriers
Ethical and Academic Integrity
Concerns about plagiarism and reliance on AI-generated solutions that may hinder learning.
Proposed Solutions: Require students to explain and rewrite code under supervision, focusing on understanding and ethical practices.
Technical Limitations
ChatGPT's inability to provide reliable references and the accuracy of its solutions varies significantly.
Proposed Solutions: Implementing a tool for semi-automated code reviews and ensuring human oversight in evaluating AI-generated content.
Project Team
Dan-Matei Popovici
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Dan-Matei Popovici
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