Reproduce “CodeHelp: Using Large Language Models with Guardrails for Scalable Support in Programming Classes” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mark Liffiton, Brad Sheese, Jaromir Savelka, Paul Denny
CodeHelp is an LLM-powered teaching assistant offering scalable, on-demand support in programming classes, utilizing guardrails to prevent direct solution disclosure and foster independent learning. A pilot deployment demonstrated consistent student engagement and high perceived value, with 95% of students interested in future use and an operational cost of approximately $0.002 per query.
Key insights
- LLMs can provide effective and scalable support in programming education when integrated with explicit pedagogical guardrails to prevent student over-reliance and promote active learning.
- An always-available AI assistant reduces student anxiety about asking questions and offers a private, judgment-free channel for support, benefiting hesitant learners.
- Analyzing student queries provides instructors with unfiltered insights into common misconceptions and struggles, enabling more responsive curriculum adjustments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.