Strategies for Using Agentic “Co-Scientist” AI Systems in Medicine

Target group and structure

This workshop series for doctoral candidates of the Medical Faculty consists of three hands-on workshops held on different days. Participation in all three workshops is strongly recommended, as they build on one another.

Workshop contents

The Planning Workshop (1^st day) introduces participants to the principles of agentic large language models (LLMs), with a particular focus on Co-Scientist systems, key data protection considerations, and the University of Freiburg’s comprehensive AI infrastructure, which provides seamless access to AI tools. Foundational AI literacy is developed through a hands-on session in which AI serves as a tutoring assistant to help participants clarify and sharpen their research question. Using AI as a sparring partner, they challenge and refine their ideas during the development of a structured planning document, while retaining full responsibility for formulating the research question.

Building on this foundation, the Synthetic Data Workshop (2^nd day) focuses on the specification and generation of hypothetical datasets. This process encourages participants to further refine and operationalize study outcomes, study design, and analysis objectives, based on the specifications from the previous session. Through discussions with peers and Institute staff, participants strengthen their AI literacy in a problem-oriented setting. Working on their own research questions, they learn to critically evaluate and utilize AI-generated suggestions while supporting one another through teamwork.

The Data Analysis Workshop (3^rd day) builds on the results from preceding sessions and guides participants through their first data analyses using agentic AI tools. Data protection requirements are addressed through the use of synthetic datasets and local model infrastructures. At the same time, participants are encouraged to critically reflect on their own use of AI throughout the analytical process.

Requirements

1. During registration, those interested are asked to upload a machine-readable outline of their own study or research project (within the scope of an “elevator pitch”). This document must also be brought to the first workshop and is intended as a starting point for the group work. Additional and more detailed documents about the participants’ studies would be useful, if available, but are not necessary to explore the AI agent workflows.
2. The participants will be asked to install necessary software on their laptops in preparation of the first workshop. A setup guide will be sent to all participants after registration.

For the second and third workshops (Synthetic Data Workshop, Data Analysis Workshop), participants must bring the results from the previous workshops, as each workshop builds on the previous one.