Students' Perspective on AI Code Completion: Benefits and Challenges
Project Overview
The document examines the role of generative AI, particularly code completion tools, in enhancing computer science education from the perspective of students. It identifies several key applications, such as boosting productivity, providing accurate syntax suggestions, and serving as an interactive coding tutor. While these tools present significant benefits, students also voice concerns regarding potential over-reliance, which may affect their learning and academic assessments, as well as issues related to the quality of generated code. Additionally, there is a clear expectation among students for these tools to offer better explainability and more personalized features to cater to individual learning needs. Overall, the findings indicate that while generative AI can facilitate learning and improve coding skills, it is essential to address the challenges it presents to ensure a balanced and effective educational experience.
Key Applications
AutoAurora - a Visual Studio Code extension for AI code completion
Context: Educational context focusing on undergraduate computer science students
Implementation: Developed an open-source extension using the StarCoder model, with an interview study conducted to gather student feedback
Outcomes: Enhanced productivity and efficiency, improved learning of programming syntax, and provided coding assistance like a tutor
Challenges: Over-reliance on AI leading to superficial understanding of programming concepts, academic assessment issues, and concerns about the quality of AI-generated code
Implementation Barriers
Cognitive and Assessment Barrier
Over-reliance on AI code completion tools may hinder the development of problem-solving skills and make it difficult to assess student's true coding abilities.
Proposed Solutions: Implement strategies like documentation, oral examinations, and randomized assessments to ensure students understand underlying concepts and gauge their understanding.
Quality Barrier
AI code completion tools may not always generate high-quality, accurate code.
Proposed Solutions: Encourage students to critically evaluate AI suggestions and use AI tools as assistance rather than replacements for coding knowledge.
Project Team
Wannita Takerngsaksiri
Researcher
Cleshan Warusavitarne
Researcher
Christian Yaacoub
Researcher
Matthew Hee Keng Hou
Researcher
Chakkrit Tantithamthavorn
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Wannita Takerngsaksiri, Cleshan Warusavitarne, Christian Yaacoub, Matthew Hee Keng Hou, Chakkrit Tantithamthavorn
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