LunaAI: A Polite and Fair Healthcare Guidance Chatbot
Researchers built a healthcare chatbot prototype that scored 4.7/5 for politeness and 4.9/5 for fairness in user testing.
A research team has published findings on LunaAI, a healthcare chatbot prototype specifically engineered for ethical communication in sensitive medical contexts. The system, detailed in a 26-page arXiv paper, addresses critical gaps in emotional intelligence and fairness that often undermine patient trust in existing digital health solutions.
Technically, LunaAI was implemented using the Google Gemini API as its underlying language model, with the team developing a mobile-first Progressive Web App built on React, Vite, and Firebase. The researchers employed a user-centered design methodology, creating conversational scenarios that handle both routine interactions and hostile user inputs. For evaluation, they used established frameworks including the Godspeed Questionnaire and conducted comparative analysis against baseline outputs from uncustomized large language models.
The results showed measurable improvements in key interaction qualities, with LunaAI achieving average user ratings of 4.7 out of 5 for politeness and 4.9 out of 5 for fairness. These scores represent significant advancements over standard LLM outputs, demonstrating that intentional ethical design can substantially enhance user experience in healthcare applications. The findings highlight that technical implementation alone is insufficient for sensitive domains—deliberate conversational design is equally critical for building patient trust and reducing user anxiety.
- Built with Google Gemini API and deployed as mobile-first PWA using React/Vite/Firebase stack
- Achieved 4.7/5 politeness and 4.9/5 fairness ratings in user testing, outperforming uncustomized LLM baselines
- Designed specifically to handle hostile user interactions through ethical conversational scenarios
Why It Matters
Demonstrates that intentional ethical design, not just model capability, is critical for building patient trust in sensitive healthcare AI applications.