[Master Thesis/ Semester Thesis] Adaptive Interaction Strategies in Healthcare Chatbots

27.01.2026, Studentische Hilfskräfte, Praktikantenstellen, Studienarbeiten

Background: Conversational agents are increasingly used for health-related information and as symptom checkers. In these high-stakes contexts, how a chatbot communicates (e.g., clarity, transparency, credibility, emotional support) strongly shapes user trust, acceptance, and safe use. A key open question is when a chatbot should adapt its communication: different user states (e.g., anxiety, confusion, skepticism) may require different response characteristics. Evidence on which user signals should trigger adaptation is still limited, especially at the level of concrete, implementable chatbot characteristics. Objective: This thesis aims to identify and validate user-state signals that should trigger adaptive chatbot responses in healthcare. Through an online user study, different user states and preferences for chatbot response characteristics (e.g., explainability/clarity, transparency/credibility, empathy/support). The outcome will be a trigger-to-response guideline that can inform the design of adaptive medical diagnosis chatbots. Tasks: - Literature Review (Chatbot design, dialog management, conversational agents in healthcare) - Study design and execution - Derivation of a "signal → preferred characteristics” mapping and design recommendations for adaptive chatbots [Preferred] Prerequisites: - Interest in Human-AI Interaction, chatbots - Experience with survey tools, quantitative user research methods - Familiarity with chatbot interfaces or AI tools

Background

Conversational agents are increasingly used for health-related information and as symptom checkers. In these high-stakes contexts, how a chatbot communicates (e.g., clarity, transparency, credibility, emotional support) strongly shapes user trust, acceptance, and safe use. A key open question is when a chatbot should adapt its communication: different user states (e.g., anxiety, confusion, skepticism) may require different response characteristics. Evidence on which user signals should trigger adaptation is still limited, especially at the level of concrete, implementable chatbot characteristics.

Objective

This thesis aims to identify and validate user-state signals that should trigger adaptive chatbot responses in healthcare. Through an online user study, it examines different user states and preferences for chatbot response characteristics (e.g., explainability/clarity, transparency/credibility, empathy/support). The outcome will be a trigger-to-response guideline that can inform the design of adaptive medical diagnosis chatbots.

Tasks

• Literature review (chatbot design, dialog management, conversational agents in healthcare)
• Study design and execution
• Derivation of a "signal → preferred characteristics” mapping and design recommendations for adaptive chatbots

Preferred Prerequisites

• Interest in Human-AI Interaction, chatbots, UX, interaction design
• Experience with survey tools, quantitative user research methods
• Familiarity with chatbot interfaces or AI tools

Kontakt: rutuja.joshi@tum.de