Moritz Willig

Hi! I'm Moritz.
I'm a PhD student at the AIML Lab at TU Darmstadt,
teaching AI models 'genuine' (causal) reasoning.

I am on the academic job market (starting mid-2026) for Postdoc or Research Scientist positions. If you are interested in my research, please feel free to reach out.

About Me

I am a final-year PhD student at the AIML Lab at TU Darmstadt, Germany, working together with Kristian Kersting and Devendra Singh Dhami, trying to understand how causal (and general reflective) reasoning can be integrated and learned by modern LLM and similar AI systems. I am furthermore interested in how causal systems evolve and change. With our recent works on meta-causal models we investigate how to model and analyze qualitative changes in dynamic environments for outcome prediction and decision making.

News

Jan. 2026:	I was invited to give a guest lecture "Causality for AI & ML: Beyond Causal Parrots" at LMU, Munich. [Slides]
Oct. 2025:	I gave a contributed talk "Beyond Causal Parrots: The Role of Meta-Causality for Genuine Causal Understanding" at the Winter School on Causality and Explainable AI at Sorbonne University, Paris. [Slides]
Oct. 2025:	I was recognized as a top reviewer at NeurIPS 2025.
Sept. 2025:	My paper "When Causal Dynamics Matter: Adapting Causal Strategies through Meta-Aware Interventions." has been accepted at NeurIPS 2025.
Sept. 2025:	Florian's paper "Structural Causal Circuits: Probabilistic Circuits Climbing All Rungs of Pearl's Ladder of Causation." has been accepted at TMLR.
May 2025:	I was recognized as a notable reviewer at ICLR 2025.
May 2025:	I was recognized as a top reviewer at ICML 2025.
Jan. 2025:	My paper "Systems with switching causal relations: A meta-causal perspective." has been accepted as an ICLR spotlight paper.

Research Topics

These are the topics I'm working on at the moment. See the full list of publications below.

Causality and LLMs

Are LLM truly able to perform causal reasoning or did they only learn to 'parrot' memorized patterns? We where upon the first to test causal discovery and reasoning capabilities of LLM.

Paper: "Causal Parrots: Large Language Models May Talk Causality But Are Not Causal". TMLR 2023

Meta-Causal Models

How do causal relations evolve under dynamical systems? We abstract away from structural equations and model
qualitative edge dynamics via meta-causal models.

Paper: "Systems with switching causal relations: A meta-causal perspective". ICLR 2025 [Spotlight]

Model Consolidation

Can causal models be consolidated while still being able to intervene? We formalize conditions for intervention-preserving consolidation of SCM and discuss the utility of resulting models.

Paper: "Do not marginalize mechanisms, rather consolidate!". Neurips 2023

Publications

Google Scholar.

Conference |

Journal |

Workshop |

Preprint | *: indicates shared first-authorship.

Preprints

Florian Peter Busch*, Moritz Willig*, Kristian Kersting and Devendra Singh Dhami.
Tagged for Direction: Pinning Down Causal Edge Directions with Precision.
ArXiv, 2025. Link: arXiv

Tim Woydt, Moritz Willig, Antonia Wüst, Lukas Helff, Wolfgang Stammer, Constantin A. Rothkopf and Kristian Kersting.
Fodor and Pylyshyn’s Legacy – Still No Human-like Systematic Compositionality in Neural Networks.
ArXiv 2025, Link: arXiv.

Nicholas Tagliapietra, Juergen Luettin, Lavdim Halilaj, Moritz Willig, Tim Pychynski, Kristian Kersting.
CausalMan: A physics-based simulator for large-scale causality.
ArXiv 2025, Link: arXiv.

2025

Moritz Willig, Tim Nelson Woydt, Devendra Singh Dhami and Kristian Kersting.
When Causal Dynamics Matter: Adapting Causal Strategies through Meta-Aware Interventions.
Thirty-ninth Conference on Neural Information Processing Systems (NeurIPS 2025). Link: OpenReview

Florian Peter Busch, Moritz Willig, Matej Zečević, Kristian Kersting and Devendra Singh Dhami.
Structural Causal Circuits: Probabilistic Circuits Climbing All Rungs of Pearl's Ladder of Causation.
Transactions on Machine Learning Research (TMLR), 2025. Link: OpenReview

Florian Peter Busch, Moritz Willig, Matej Zečević, Kristian Kersting and Devendra Singh Dhami.
Counterfactual Sum-Product Networks.
UAI 2025 Workshop on Tractable Probabilistic Models, 2025. Link: OpenReview

Moritz Willig, Tim Nelson Tobiasch, Florian Peter Busch, Jonas Seng, Devendra Singh Dhami and Kristian Kersting.
Systems with switching causal relations: A meta-causal perspective.
The Thirteenth International Conference on Learning Representations (ICLR), 2025. Link: OpenReview
Spotlight ⭐

2024

Moritz Willig, Tim Nelson Tobiasch, Florian Peter Busch, Jonas Seng, Devendra Singh Dhami and Kristian Kersting.
Systems with switching causal relations: A meta-causal perspective.
NeurIPS2024 Causal Representation Learning Workshop, 2024. Link: OpenReview.

Florian Peter Busch, Moritz Willig, Jonas Seng, Kristian Kersting and Devendra Singh Dhami.
psiNet: Efficient causal modeling at scale.
International Conference on Probabilistic Graphical Models (PGM), pages 452–469. PMLR, 2024. Link: PMLR.

Harsh Poonia, Moritz Willig, Zhongjie Yu, Matej Zečević, Kristian Kersting and Devendra Singh Dhami.
chiSPN: Characteristic interventional sum-product networks for causal inference in hybrid domains.
The 40th Conference on Uncertainty in Artificial Intelligence (UAI), 2024. Link: OpenReview.

Moritz Willig, Matej Zečević, and Kristian Kersting.
“Do not disturb my circles!” Identifying the type of counterfactual at hand.
Conference on Advances in Robust Argumentation Machines (RATIO), pages 266–275. Springer, 2024. Link: Springer.

Anita Keshmirian, Moritz Willig, Babak Hemmatian, Kristian Kersting, Ulrike Hahn and Tobias Gerstenberg.
Chain versus common cause: Biased causal strength judgments in humans and large language models.
Proceedings of the Annual Meeting of the Cognitive Science Society (CogSci), volume 46, 2024. Link: eScholarship.

2023

Moritz Willig, Matej Zečević, Devendra Singh Dhami and Kristian Kersting.
Do not marginalize mechanisms, rather consolidate!
Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS), 2023. Link: OpenReview.

Matej Zečević*, Moritz Willig*, Devendra Singh Dhami and Kristian Kersting.
Causal parrots: Large language models may talk causality but are not causal.
Transactions on Machine Learning Research (TMLR), 2023. Link: OpenReview.

Matej Zečević, Moritz Willig, Devendra Singh Dhami and Kristian Kersting.
Identifying challenges for generalizing to the pearl causal hierarchy on images.
ICLR 2023 Workshop on Domain Generalization, 2023. Link: OpenReview.

2022

Moritz Willig*, Matej Zečević*, Devendra Singh Dhami and Kristian Kersting.
Can foundation models talk causality?
UAI 2022 Workshop on Causal Representation Learning, 2022. Link: ArXiv.

2020

Johannes Czech, Moritz Willig, Alena Beyer, Kristian Kersting and Johannes Fürnkranz.
Learning to play the chess variant crazyhouse above world champion level with deep neural networks and human data.
Frontiers in Artificial Intelligence, 3:24, 2020. Link: FrontiersIn.

Professional Service

I am a regular reviewer for various international ML conferences, journals and workshops:
NeurIPS 2024, 2025 (top reviewer); AAAI 2024, 2025; ICLR 2025 (notable reviewer), 2026; ICML 2022, 2025 (top reviewer); AISTATS 2022 (top reviewer; top 10%), 2024-2026; TMLR 2024-current; Conference on Causal Learning and Reasoning (CLeaR) 2025, 2026.

Contact

Find me on
Twitter
Bluesky
LinkedIn
Github
or reach out via e-mail
moritz.willig@cs.tu-darmstadt.de
moritz@moritz-willig.de