LMU Researcher Receives DFG Funding for Explainable AI

Johannes Paetzold, a researcher at Ludwig-Maximilians-Universität München (LMU), has been awarded approximately €2.1 million in funding from the German Research Foundation's (DFG) Emmy Noether Programme. The funding, allocated over six years, will support the development of artificial intelligence (AI) systems capable of making their decision-making processes understandable to humans.

While current AI models achieve impressive results in analyzing images and texts, the mechanisms behind their decisions often remain opaque. This lack of transparency poses a significant challenge, particularly in critical fields such as medicine, hindering reliable implementation. Paetzold's work aims to address this by enhancing the comprehensibility of AI models.

Paetzold's approach integrates methods from geometry, topology, and graph theory with contemporary AI architectures. The core concept is to embed principles familiar to human cognition, such as finding the shortest path or recognizing objects irrespective of their orientation, into AI systems. This is expected to result in more robust and interpretable AI.

The research will initially focus on medical applications, including early detection of eye, vascular, and cancerous diseases. However, the developed methods are also applicable to other domains, such as satellite imagery analysis. The long-term objective is to create AI tools that provide reliable decision support for healthcare professionals while remaining understandable to patients.