The use of LLM-based conversational agents in psychiatric diagnostics opens up new possibilities for making clinical interviews more scalable and accessible. However, the way patients disclose information is strongly influenced by design decisions, such as the design of the avatar, the chosen interaction modality, or the framing of the system. To investigate these factors systematically in experimental settings, a modular and configurable interface is required that can reliably represent different experimental conditions. The aim of this thesis is to conceptualize and develop a prototype of such an experimental interface following the Design Science Research approach.

More information is available here in the call for applications