Reproduce “CoachGPT: A Scaffolding-based Academic Writing Assistant” 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

Fumian Chen, Sotheara Veng, Joshua Wilson, Xiaoming Li, Hui Fang

CoachGPT, developed by the University of Delaware, applies Large Language Models within a pedagogical scaffolding framework to guide students through the academic writing process. The system aims to teach writing skills rather than just generating text, providing personalized, real-time feedback across 11 stages from pre-writing to revision.

Key insights

  • Integrating Large Language Models within an educational scaffolding framework can effectively shift AI writing assistance from mere text generation or correction to active skill instruction.
  • Careful prompt engineering is crucial for ensuring LLM responses are pedagogically sound, preventing misuse (e.g., direct content generation), and fostering a supportive learning environment.
  • An end-to-end, iterative guidance system can significantly benefit novice writers and second-language learners by making the academic writing process more manageable and accessible.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.