This bachelor’s thesis examines the integration of large language models into medical questionnaires using the example of a checklist from the Deutsche Krebsgesellschaft for recording hereditary risk factors for breast and ovarian cancer. The aim is to develop and evaluate an LLM-assisted user interface that supports users in answering complex questions, thereby improving usability and user experience.

As part of a human-centered design process, a proof of concept was designed, developed, and tested in a qualitative study with six participants. The user interface includes artificial intelligence support from two LLM agents that communicate with each other and generate appropriate outputs for users. The results of the study show improved transparency, guidance, and efficiency through LLM assistance, while reducing the overall complexity of the questionnaire.

The work confirms the potential of large language models to support medical survey instruments, but also provides an outlook on necessary improvements and future research areas for the further development of such applications.