Skip to main content Skip to navigation

Scaling CS1 Support with Compiler-Integrated Conversational AI

Project Overview

The document discusses the implementation of DCC Sidekick, a conversational AI tool embedded within a C/C++ compiler that supports students in debugging programming errors. Specifically designed for an introductory programming course (CS1) at a large Australian university, the tool generates comprehensive error explanations based on the context provided by the compiler, thereby enriching the educational experience. The findings indicate that DCC Sidekick enhances student learning, particularly during peak times when students are most likely to seek assistance. The paper highlights key applications of generative AI in education, focusing on its scalability and effectiveness in facilitating understanding and problem-solving in programming. Overall, the integration of this AI tool demonstrates significant pedagogical implications, showcasing a promising approach to improving student engagement and success in computer science education.

Key Applications

DCC Sidekick

Context: Introductory programming course (CS1) at a university, targeting novice programmers.

Implementation: Integrated into an existing C/C++ compiler, allowing students to access AI-generated error explanations through a web interface.

Outcomes: Strong adoption with 959 users and 11,222 sessions within seven weeks, facilitating over 17,982 error explanations. Significant engagement during business and non-business hours.

Challenges: Potential over-reliance on AI assistance, issues with academic integrity, and managing the balance between providing help and fostering independent problem-solving skills.

Implementation Barriers

Technical

Large language models (LLMs) can produce hallucinations or biased answers, which may mislead students.

Proposed Solutions: Incorporation of guardrails and prompting strategies to filter and refine AI-generated responses.

Pedagogical

Concerns about students developing dependency on AI tools rather than learning debugging skills independently.

Proposed Solutions: Encouraging gradual engagement with AI assistance while maintaining a focus on skill development and critical thinking.

Project Team

Jake Renzella

Researcher

Alexandra Vassar

Researcher

Lorenzo Lee Solano

Researcher

Andrew Taylor

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Jake Renzella, Alexandra Vassar, Lorenzo Lee Solano, Andrew Taylor

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

Let us know you agree to cookies