Moritz Knolle is a doctoral researcher at the Chair for AI in Healthcare and Medicine at the Technical University of Munich, supervised by Georgios Kaissis, Ben Glocker and Daniel Rueckert. Moritz is affiliated with the TUM University Hospital and the Konrad Zuse School of Excellence in Reliable AI.
Privacy risk is typically measured in aggregate across a dataset/population. Moritz is interested in quantifying what a deployed medical AI model actually leaks about individual patients — and how that risk is distributed across individuals and subgroups.
What do AI models retain about the individual records they were trained on? And what does that mean for the patients whose data they memorised, once those models are used in the clinic?
Trust in medical AI depends on more than accuracy. Trust requires models whose behaviour and performance are: consistent across patients and subgroups, stable under changing conditions and resistant to misuse.
* denotes equal contribution. Full list on Google Scholar.
moritz.knolle[at]tum.de
TranslaTUM, Einsteinstraße 25, 81675 Munich, Germany
