“Causality” professorship at the Research Center Trustworthy Data Science and Security appointed

The new professorship is already the seventh at the research center founded in 2021, which deals with the trustworthiness of intelligent systems in safety-critical applications. Alexander Marx focuses on causal and non-causal relationships: Users of AI regularly encounter the problem that, unlike humans, AI does not intuitively understand the concept of cause and effect, i.e. causality. For example, we understand that children's height and intelligence depend on their age. An AI, on the other hand, could draw the erroneous conclusion that IQ influences weight. It is good at recognizing patterns in a large amount of data, but often confuses statistical correlations with causality.

In order to avoid such errors and make pattern recognition processes transparent, Marx wants to research how AI models recognize and map causal networks in observational data. His work in this area is interdisciplinary and involves collaboration with medical professionals, for example: When looking at gene activity data, it is often unclear whether certain genes influence each other or whether their simultaneous activity has other reasons. By making the way AI works more transparent and precise, users can make more accurate predictions when researching genetic diseases.

Avoiding false assumptions of AI

Marx also focuses on developing robust and trustworthy machine learning models. The algorithms are trained using data of different types, such as text, video or images, but sometimes they do not learn the actual cause-and-effect relationships, but also non-causal correlations, resulting in false assumptions. Marx addresses the question of how such models, which are used in autonomous vehicles and medicine, for example, must be designed so that no false conclusions arise and we can trust them. "The research field of causality has many interfaces with other areas," says Alexander Marx about his field of research. "I'm really looking forward to working with colleagues from statistics and information science and also on interdisciplinary projects with partners from medicine, biology and psychology."

Marx previously worked in Saarbrücken, Amsterdam, Zurich and Basel. For his doctorate in computer science, which he completed in 2021, he conducted research at the Max Planck Institute for Informatics and the CISPA Helmholtz Center for Information Security in Saarbrücken and completed a guest stay at the University of Amsterdam. He then moved to ETH Zurich, where he worked at the AI Center on the topics of causal discovery, causal representation learning and information theory as well as interdisciplinary projects at the interface between machine learning and medicine. Prior to his appointment, Marx analyzed machine learning models for biomedical applications in HIV disease at the Computational Biology Group at ETH Zurich. He has also been a member of the European Laboratory for Learning and Intelligent Systems (ELLIS) since 2024.

Back