HistoChat

The High-Stakes Context

Traditional history education often struggles to foster emotional connection and deep engagement among middle school students. HistoChat targets the educational domain by acting as a conversational learning platform where students engage in direct dialogue with AI-driven historical personas.

The Architectural Orchestration

Architected a robust full-stack LLM pipeline utilizing React, Node.js, and WebSocket for real-time interaction. Orchestrated specialized GPT prompt engineering to ensure the historical personas maintained strict factual accuracy and age-appropriate educational boundaries.

The Deployment & Scale

Deployed the fully functioning system "in the wild" across multiple middle school classrooms for a 6-week continuous deployment study.

The Empirical Evaluation

Synthesized comprehensive system logs (MongoDB) with pre/post-study surveys. The mixed-methods evaluation proved a statistically significant improvement in student engagement, historical comprehension, and emotional connection to the subject matter.

The Outcome

• CSCW 2025 Full Paper accepted (1st author)
• Proposed formal design guidelines for AI educational persona systems