HealthBots: Expert-in-the-loop Chatbots

Patients and their caregivers often experience anxiety before undergoing medical treatment. They require timely, trustworthy, detailed, and accurate information about their condition and care. To address this need, we designed and developed a family of HealthBots powered by state-of-the-art generative AI models grounded in clinician-provided knowledge bases. These systems assist patients and caregivers in obtaining reliable answers to their medical queries throughout the treatment journey.

A key aspect of our approach is the active involvement of healthcare professionals and patient coordinators through an expert-in-the-loop framework. Experts review and provide feedback on the bot’s responses, helping ensure safety and reliability while continuously improving answer quality over time. The bots are designed to be multilingual and multimodal, supporting both voice and text interactions to make them accessible to users with diverse language preferences and literacy levels.

We have developed and deployed several HealthBots in real-world healthcare settings, including:

• CataractBot, for patients undergoing cataract surgery, in collaboration with Sankara Eye Hospital.
• OncoBot, for patients undergoing cancer treatment, in collaboration with Kasturba Medical College, Manipal.
• DRBot, for patients undergoing diabetic retinopathy treatment, in collaboration with Sankara Eye Hospital.

Building on this work, we also developed and deployed ASHABot, in collaboration with Khushi Baby, to support community health workers with their questions related to maternal and child health. ASHABot extends the HealthBot paradigm beyond patients and caregivers, demonstrating how AI assistants can support frontline healthcare workers in delivering care and accessing reliable health information.