Designing Structured Conversational Support for Tuberculosis Treatment Adherence and Patient Coping

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Health & Wellbeing — Public Health & Epidemiology, Mental Health & Psychological Support, Medical Devices & Health Technology · Depth: Intermediate, quick

Summary

A Spanish-language chatbot was developed and iteratively refined to provide structured conversational support for Tuberculosis (TB) treatment adherence and patient coping. This system addresses unmet needs for timely, personalized, and emotionally supportive communication outside clinical settings. It integrates three core functions: TB information grounded in curated resources, coping-oriented support inspired by Dialectical Behavior Therapy (DBT), and deterministic safety-critical crisis handling. Implemented with a routed architecture and shared conversational state, the development process identified common failure modes in unstructured systems, such as weak grounding and inconsistent safety. Findings suggest that reliable conversational behavior in clinical support relies on structured interaction design and explicit control over routing, memory, and safety, rather than solely on model capability.

Key takeaway

For AI Engineers designing conversational agents for sensitive clinical support, prioritize structured interaction design over relying solely on large language model capabilities. You should implement explicit routing, state tracking, and task-specific prompting to ensure reliable behavior, strong grounding, and consistent safety. This approach helps mitigate common failure modes like poor multi-turn continuity and inconsistent safety responses in critical healthcare applications.

Key insights

Reliable clinical conversational AI requires structured interaction design and explicit control over routing, memory, and safety.

Principles

• Clinical chatbot reliability depends on structured design, not just model capability.
• Unstructured conversational systems often fail in grounding, continuity, and safety.
• Separating core functions enhances conversational system robustness.

Method

A Spanish-language TB treatment-support chatbot was iteratively refined through expert evaluation, implementing a routed architecture with shared conversational state for information, DBT-inspired coping, and deterministic crisis handling.

In practice

• Design distinct modules for information, emotional support, and crisis intervention.
• Implement explicit routing and state tracking for multi-turn continuity.
• Ground information support in curated, reliable resources.

Topics

• Tuberculosis Treatment
• Chatbots
• Conversational AI
• Digital Health
• Patient Adherence
• Dialectical Behavior Therapy
• Clinical Support Systems

Best for: AI Architect, AI Scientist, NLP Engineer, AI Engineer, Research Scientist

Related on AIssential

• Exploring the feasibility of conversational diagnostic AI in a real-world clinical study — Supervised conversational AI for pre-visit history taking is feasible, safe, and well-received by patients and clinicians.
• Developing and evaluating a chatbot to support maternal health care — Trustworthy medical chatbots in noisy, multilingual settings require defense-in-depth design and multi-method evaluation.
• Talking to Your Data: Exploring Embodied Conversation as an Interface for Personal Health Reflection — Embodied conversational agents can facilitate objective personal health data reflection, moving beyond passive dashboard viewing.
• TherapyGym: Evaluating and Aligning Clinical Fidelity and Safety in Therapy Chatbots — TherapyGym evaluates and aligns therapy chatbots using clinical fidelity and safety metrics, improving performance through RL with expert-validated feedback.
• Rethinking Patient Education as Multi-turn Multi-modal Interaction — Patient education requires multi-turn, multimodal interaction with robust visual grounding and safety considerations.
• Beyond Chatbots: Designing Conversational Banking Experiences That Scale — Third-party AI assistants are disintermediating banks by becoming the primary interface for consumer financial queries.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.