portrait

Home / Research / Causeries

Sebastian Weichwald.

📢 News / 👥 Group / 🔍 Research with us / 🏛 Contact

I aim to understand and resolve what holds causal modelling back in practice by clarifying its foundations (specifying what is being modelled, how it can be tested, which metrics reflect progress) and by prioritising ready-to-use implementations. I am fortunate to work with a relentlessly curious group and inspiring collaborators.

We argue that the common interpretation of actions as interventions renders causal model predictions circular and thus non-falsifiable (2025) and make the case for time in causal graphs (2025). These challenges relate to our earlier work on causal model transformations and abstractions (UAI 2017 & 2021), and have become increasingly urgent with growing interest in causal representation learning. Our success in the NeurIPS Causality 4 Climate competition (PMLR 2020) motivated our work uncovering how common causal discovery benchmarks may be gamed and fail to capture meaningful progress (NeurIPS 2021 & 2023). We have developed 𝚐𝚊𝚍𝚓𝚒𝚍 for evaluating learned causal graphs via adjustment identification distances (UAI 2024) and CIfly for fast algorithm development in graphical causal inference (2025).

At the University of Copenhagen, I am part of the Copenhagen Causality Lab (CoCaLa), a faculty member in the Department of Mathematical Sciences, and co-lead for causality at the Pioneer Centre for AI (P1). I also serve on the fellowship evaluation committee of the Danish Data Science Academy (DDSA). Previously, I was a postdoc in Copenhagen and a PhD student at the Max Planck Institute for Intelligent Systems and ETH Zurich.

News.

2025-Sep.
Harald is joining Anton and me as a new PhD student. He is curious about causal discovery and what emerges when non-linear functions stack deep. Looking forward to puzzling through these questions together.
2025-Jun.
CIfly is ready to fly 🪁 Now on CRAN for R and PyPI for Python! It is a declarative framework for prototyping graphical causal inference algorithms using rule tables 🛠️. Check out CIfly.dev and the preprint.
2025-May.
Emanuele presented our joint work at AISTATS. The paper, ‘All or None: Identifiable Linear Properties of Next-token Predictors in Language Modeling’, explores when linear properties persist across distribution-equivalent next-token predictors. The collaboration grew from his research visit and made for a great exchange of ideas.
2025-Feb.
Two preprints up on arXiv discussing foundations of causal (acyclic) models: we ask ‘What is causal about causal models and representations?’ and we make the case for time in causal DAGs.

Group at CoCaLa.

We are part of the Copenhagen Causality Lab, where I feel fortunate to work closely with these curious and inspiring people who make our research environment so engaging. If you would like to explore researching or collaborating with us, check out News and Research with us for current opportunities and ways to get involved.

Alumni: Simon Bing (2025, PhD exchange), Emanuele Marconato (2024, PhD exchange), Alexander G Reisach (2024, PhD exchange; 2020/21, MSc exchange), Theo Würtzen (2023/24, RA)

Research with us.

For official contact details, see my University of Copenhagen profile.

We are always happy to hear from curious minds who enjoy academic challenges and want to engage with the kinds of causal questions we explore. Whether you are a prospective postdoc, PhD student, visitor, or MSc student, you are welcome to get in touch. Please read the relevant section below and use the appropriate subject prefix when emailing. We receive many inquiries and may not be able to respond to all messages. Please do not read silence as a lack of interest. It usually means we are at the blackboard or immersed in research discussions with the group 📚. If your message is still relevant after a few weeks, do feel free to resend it. We won’t mind at all 📬.

Postdoc and PhD positions.

Postdoc positions are announced annually (applications due November 15) and PhD positions twice a year (applications due April 1 or November 15) through official department calls.

Applications must be submitted through the official calls. To keep the process fair for everyone and to manage volume, we cannot pre‑assess applications or provide feedback before the official process, and we cannot consider application materials sent by email outside these calls.

In your application, describe your own ideas and explain why they fit well with our group and, if relevant, with the specific project mentioned in the call. Even when a project is listed, you are always welcome to pitch your own research agenda. If you are unsure whether your background or interests are a perfect match, we still encourage you to apply. We value curiosity and originality, and many details can be discussed during the process.

If you plan to apply for your own funding (for example, DDSA, DARA, MSCA), email me using the subject prefix “[initiative]” and include a concise sketch of your project and why you believe our group offers the right environment for it to thrive.

Additional resources: PhD programme overview, salary conditions (PhDs usually start at step 4 after a 3+2 BSc+MSc education), pointers for prospective postdocs or PhD students on Jun Yang’s website

Research visits and exchanges.

We have had great experiences hosting research visitors and PhD exchange students. If you are interested in visiting, please email me using the subject prefix “[visit]” and include a note about your background, what you would like to research with us, and what draws you to Copenhagen or the Copenhagen Causality Lab. DDSA visit grants may be a funding option for short visits.

Collaborations.

We welcome collaborations, whether that means a joint research project, organising a workshop, or inviting us for a talk. If you have an idea that aligns with our interests, feel free to reach out with the subject prefix “[collaboration]”.

MSc thesis and student projects.

If you are a student at the University of Copenhagen and would like to do your thesis project with me, please stop by my office or email me from your university address using the subject prefix “[UCPH]”. To help us start the conversation, include one concept, topic, or paper from a relevant course that sparked your curiosity and a question it raised that you found interesting. Also tell me which aspects of the work you enjoy most (for example theory, proofs, literature research, coding, data analysis, or writing) and what your ambitions are for the project.

Olds.

Imprint & Credits