Causal Learning and AI
Computer Science
1 Career
- Since 2026 Advisory board member, Emmi AI
- 2024 – 2026 AI Resident, Chan Zuckerberg Initiative
- Since 2023 Assistant Professor, Institute of Science and Technology Austria (ISTA)
- 2021 – 2023 Senior Applied Scientist, Science lead and manager of the Causal Representation Learning team, Amazon Web Services, Tübingen, Germany
- 2020 Research Software Engineer Intern, Google Brain, Amsterdam, the Netherlands
- 2018 – 2020 Research Consultant, ETH Zürich / Max Planck Institute, MSRA partnership with Google Brain Zürich, Switzerland
- 2016 – 2020 PhD, ETH Zürich, Max Planck ETH Center for Learning Systems
- 2016 Research Assistant, ETH Zürich, Data Analytics Lab
1.1 Selected Distinctions
- Program Chair of the Neural Information Processing Systems (NeurIPS) 2024 and 2025 Datasets and Benchmarks Track
- 2024 Google Research Scholar Award
- 2023 Hector Foundation Award, Heidelberg Academy of Science
- 2023 Full Professorship (W3-chair) for Machine Learning at Saarland University – declined
- 2019 Best Paper Award, International Conference on Machine Learning (ICML)
1.2 Selected Grants
- 2026 – 2031 Schmidt Futures, Water from the mountains: global reanalysis and future tipping points, Coordinators: F. Pellicciotti (ISTA), Immerzeel (Utrecht University). Awarded overall: $ 9,990,429. Awarded to RG: € ∼250,000.
- 2024 – 2027 FFG Tiks, Building Energy Systems on Causal Reasoning, Coordinator: G. Schweiger (TU Wien). Awarded overall: € 486,804. Awarded to RG: € 150,000.
- 2024 – 2026 Chan Zuckerberg Initiative AI residency. Awarded to RG: $ 1,100,000.
1.3 Tech Transfer (Patents, Spin-Offs)
N/A
1.4 Selected Publications
- T. Mencattini∗, R. Cadei∗, and F. Locatello, “Exploratory causal inference in saence,” in International Conference on Learning Representations (ICLR) – Oral, 2026.
- R. Cadei et al., “Prediction powered causal inference,” in Advances in Neural Information Processing Systems (NeurIPS), 2025.
- D. Yao, D. Rancati, R. Cadei, M. Fumero, and F. Locatello, “Unifying causal representation learning with the invariance principle,” in International Conference on Learning Representations (ICLR), 2025.
- D. Yao, C. Muller, and F. Locatello, “Marrying causal representation learning with dynamical systems for science,” in Advances in Neural Information Processing Systems (NeurIPS), 2024.
- D. Yao et al., “Multi-view causal representation learning with partial observability,” in International Conference on Learning Representations (ICLR) – Spotlight, 2024.
2 Group Members
- Riccardo Cadei, PhD Student since 2025
- Berker Demirel, PhD Student since 2024
- Walter Nelson, PhD Student since 2025
- Dario Rancati, PhD Student since 2025
- Dingling Yao, PhD Student since 2023
- Clémentine Dominé, Postdoc since 2025
- Marco Fumero, Postdoc since 2024
- Shimeng Huang, Postdoc since 2025
- Valentino Maiorca, Postdoc since 2026
- Francesco Montagna, Postdoc since 2025
- Adeel Pervez, Postdoc since 2024
3 Research
The number of storks and newborn babies in a country are correlated, but one doesn’t cause the other. Distinguishing coincidence from cause-and-effect relationships is a central challenge in science, yet it’s becoming harder as the data required to answer scientific questions grows ever larger. To continue progress, we rely increasingly on artificial intelligence. However, today’s AI systems primarily detect correlations, not causality, which can lead to misleading or even false conclusions against experimental evidence. In the Causal Learning and AI group, we turn AI into a scientific instrument that uncovers causes and effects, plans experiments, and anticipates the outcome of its actions. To do so, we build upon the framework of statistical causal models to draw causal insights from large amounts of high-dimensional, temporal data
3.1 Current Scientific Projects/Directions
- Data-driven theory for cloud dynamics
- Epistemology for the AI age at the example of virtual cells
- BOSS: Building Energy Systems on causal reasoning
- Water from the mountains: global reanalysis and future tipping points (MountAInWater)
3.2 Research in Focus
Causal representation learning (CRL) aims at recovering latent causal variables from high-dimensional observations to solve causal downstream tasks, such as predicting the effect of new interventions or more robust classification. A plethora of methods have been developed, each tackling carefully crafted problem settings that lead to different types of identifiability. These different settings are widely assumed to be important because they are often linked to different rungs of Pearl’s causal hierarchy, even though this correspondence is not always exact. This work shows that instead of strictly conforming to this hierarchical mapping, many causal representation learning approaches methodologically align their representations with inherent data symmetries that can be assumed. Identification of causal variables is guided by invariance principles that are not necessarily causal. This result allows us to unify many existing approaches in a single method that can mix and match different assumptions, including non-causal ones, based on the invariance relevant to the problem at hand. It also significantly benefits applicability, which we demonstrate by improving treatment effect estimation on real-world high-dimensional ecological data. Overall, our paper clarified the role of causal assumptions in the discovery of causal variables and shifted the focus to preserving data symmetries. This work was published in the International Conference on Learning Representation: D. Yao et al., International Conference on Learning Representation (2025).