Causal Machine Learning for Reliable Decision-Making - Dissertation

Sep 17, 2026

Uploaded by Noah

Data-driven decision-making has become increasingly crucial across healthcare, economics, and policy-making, where large-scale datasets can guide critical choices. The dissertation focuses on causal machine learning methods that aim to produce reliable, decision-relevant insights rather than purely predictive signals. Emphasis is placed on learning causal effects under realistic conditions, leveraging modern machine-learning tools while maintaining the validity of conclusions. Applications include areas such as medical treatment optimization and evidence-driven public policy design.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages218

Reading Time9 h 9 min

AccessFree to download

Acknowledgements

Abstract

Causal Machine Learning for Reliable Decision-Making Dennis Frauen München, den 29.09.2025 Causal Machine Learning for Reliable Decision-Making Dennis Frauen Dissertation an der Fakultät für Mathematik, Informatik, und Statistik der Ludwig-Maximilians-Universität München vorgelegt von Dennis Frauen München, den 29.09.2025 Erstgutachter: Prof. Dr. Stefan Feuerriegel Zweitgutachterin: Prof. Dr. Mihaela van der Schaar Drittgutachter: Prof. Dr. Uri Shalit Tag der mündlichen Prüfung: 24.03.2026 Acknowledgements First and foremost, I am profoundly grateful to my advisor, Prof. Stefan Feuerriegel, for his support throughout my entire PhD. Thank you Stefan for giving me the freedom to explore my own ideas while at the same time pushing me to grow as a researcher. I am especially grateful for your guidance in writing, your encouragement to collaborate, and for fostering an environment in which research could thrive. I am thankful to my second advisor, Prof. Mihaela van der Schaar, for giving me the opportunity to spend multiple research stays in Cambridge and for enabling collaborations within the ELLIS PhD program. Your creative input on our projects and your broader research vision have been deeply inspiring. I would also like to sincerely thank Prof. Uri Shalit for taking the time to serve as a referee for my PhD thesis and for participating in my defense. My thanks also go to Prof. Volker Tresp for acting as chair of my PhD committee and for leading the discussion during my defense. I am very grateful to my academic collaborators. In particular, I would like to thank Prof. Niki Kilbertus for many thought-provoking discussions and for several exciting project collaborations. I thank Alicia Curth and Fergus Imrie for welcoming me in Cambridge and for a fruitful collaboration. I am also grateful to Henri Arno and Ruizi Yan for visiting, for your enthusiasm about our research, and for the inspiring exchanges. My thanks also go to Milan Kuzmanovic and Tobias Hatt for their support and collaboration at the beginning of my PhD, and to Lars van der Laan for patiently answering all my causal inference questions and for an amazing summer in California. I would also like to thank my industry mentors and collaborators at Netflix, including Kevin Mercurio, Sherry Xie, and Grace Huang, and at Meta, including Samir Khan, Roxana Pop, Israel Nir, and Ariel Evnine. Thank you for introducing me to applications of causal inference in industry and for showing me real-world impact of my research. A very special thank you goes to my colleagues at LMU, to whom I owe much of my academic success over the last years. You supported me not only as lab mates, but also as friends. I am especially grateful to Valentyn Melnychuk for spending the Covid period together in the office, for studying causal inference with me, for welcoming me to Munich, for brainstorming research ideas together, and for helping lay the foundations of what has become today’s causal ML lab. I thank Jonas Schweisthal for both excellent research discussions and gym advice, and for the many nights spent together in the office, at beer halls, or at conference events. I am also very grateful to Maresa Schröder, Emil Javurek, Konstantin Heß, Yuxin Wang, and Yuchen Ma for so many collaborations, random discussions, support, and shared memories, from trips and conferences to almost dying on a volcano hike. I also want to thank all other lab members who shaped this journey, whether through collaboration or simply by sharing the latest gossip in and out of the office: Dominique Geißler, Abdurahman Maarouf, Dominik Bär, Simon Schallmoser, Octavia Ciora, Marie Brockschmidt, Kerstin Forster, Jonas Schröder, Sebastian Maier, and Haorui Ma. Last but not least, thank you to Shuk-ki Lam for doing an amazing job keeping our backs free for research. You were all part of this journey and one of the main reasons why these years were both so successful and so much fu

FAQs

What problem does the dissertation address?

It studies how to use data-driven methods for reliable decision-making across domains like healthcare, economics, and policy-making.

What role do large-scale datasets play?

They provide information to support critical decisions, such as optimizing medical treatments and designing effective public policies.

Which main approach does the dissertation emphasize?

Causal machine learning methods that learn causal effects to support decision-relevant conclusions rather than relying only on prediction.

Sep 17, 2026

Uploaded by Noah

Causal Machine Learning for Reliable Decision-Making - Dissertation_Page 1

Open in WPS to view all 218 pages