Skip to main content Skip to navigation

AI Teaches the Art of Elegant Coding: Timely, Fair, and Helpful Style Feedback in a Global Course

Project Overview

The document examines the integration of a Real-Time Style Feedback tool (RTSF) in a large-scale online CS1 course at Stanford University, leveraging large language models (LLMs) to deliver immediate, personalized feedback on coding style. The study reveals that students who received real-time feedback exhibited significantly higher engagement and improvement in their coding skills compared to peers who received feedback after a delay. This highlights the effectiveness of the RTSF tool in enhancing learning outcomes, demonstrating that timely and tailored feedback is crucial for fostering student development in programming. Overall, the findings underscore the potential of generative AI applications in education to optimize the learning experience and improve student performance.

Key Applications

Real-Time Style Feedback tool (RTSF)

Context: CS1 online course with over 8,000 global students

Implementation: Integrated into the students' IDE, providing real-time feedback after functionality tests.

Outcomes: Students receiving real-time feedback were five times more likely to engage with it and made significant style edits, with 79% incorporating the feedback into their code.

Challenges: Quality of LLM feedback, potential demographic bias, and ensuring safety and reliability of the tool.

Implementation Barriers

Quality Control

Ensuring the quality and reliability of feedback from LLMs can be challenging, especially in identifying misleading variable names and providing useful suggestions.

Proposed Solutions: Implementing rigorous checks and control mechanisms for the LLM output, including prompt engineering and structured JSON responses.

Demographic Bias

Potential biases in LLM-generated feedback that could negatively affect specific groups of students.

Proposed Solutions: Safeguards such as anonymizing student data, limiting interaction rounds, and ensuring feedback is processed through a validation system.

Project Team

Juliette Woodrow

Researcher

Ali Malik

Researcher

Chris Piech

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Juliette Woodrow, Ali Malik, Chris Piech

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