Kauro, a graph-based chatbot for high-fidelity information transmission conversations

Area of Science:

• Biomedical informatics
• Clinical research technology
• Artificial intelligence in healthcare

Background:

• Informed consent is critical but complex, labor-intensive, and inconsistent across sites.
• Current digital tools for informed consent lack reproducibility and auditability.
• Generative AI chatbots introduce risks like hallucination and regulatory non-compliance.

Purpose of the Study:

• To introduce Kauro, an open-source, graph-based chatbot for secure and reliable biomedical communication.
• To address the limitations of existing platforms in informed consent processes.
• To enhance reproducibility, auditability, and regulatory oversight in clinical research and care.

Main Methods:

• Developed Kauro, an open-source chatbot utilizing graph-based structures.
• Encoded scripted conversations as version-controlled JavaScript Object Notation (JSON) files.
• Implemented deterministic traversal, comprehensive audit logging, and modular client-server architecture.

Main Results:

• Kauro ensures deterministic conversation paths, enhancing predictability and safety.
• The platform provides complete audit logs, facilitating IRB-verifiable oversight.
• Modular design ensures portability and scalability across different institutions.

Conclusions:

• Kauro offers a scalable solution for safe, precise, and trustworthy biomedical communication.
• It reframes the deployment of AI in healthcare by prioritizing reproducibility and auditability.
• The platform is generalizable to any domain requiring secure and verifiable conversational data.