Martin Pawelczyk

I am a PhD candidate in computer science at the Data Science and Analytics Lab at the University of Tuebingen. Prior to this, I did a MSc in Statistics (Research) at the London School of Economics (LSE), where I focused on statistical learning. In my LSE dissertation, I advanced a nonparametric test for statistical independence that can be applied to infer undirected Gaussian and general graphical models in high-dimensional scenarios when the number of features exceeds the number of observations. Before graduating from LSE, I also studied Econometrics at the University of Edinburgh (MSc) and Economics at the University of Cologne (BSc). During my graduate studies, I was fortunate enough to be generously supported by the German National Merit Foundation.

Research Interests

Fairness and Explainability with guarantees
Generation of counterfactual explanations with guarantees

If you are interested in one of the topics mentioned above and you would like to write your thesis in that area, you are welcome to contact me. Additionally, we regularly publish thesis announcements on our webpage.

Recent Publications

Deep Neural Networks and Tabular Data: A Survey @ IEEE Transactions on Neural Networks and Learning Systems (TNNLS) (Vadim Borisov, Tobias Leemann, Kathrin Sessler, Johannes Haug, Martin Pawelczyk und Gjergji Kasneci)

I Prefer not to Say: Operationalizing Fair and User-guided Data Minimization @ NeuIPS22 Workshop on Algorithmic Fairness through the Lens of Causality and Privacy (Tobias Leemann, Martin Pawelczyk, Christian Thomas Eberle und Gjergji Kasneci)

On the Trade-Off between Actionable Explanations and the Right to be Forgotten @ NeuIPS22 Workshop on Trustworthy and Socially Responsible Machine Learning (Martin Pawelczyk, Tobias Leemann, Asia Biega und Gjergji Kasneci).

Algorithmic Recourse in the Face of Noisy Human Responses @ICLR SRML Workshop 2022 (Martin Pawelczyk, Teresa Datta, Johannes van-den-Heuvel, Gjergji Kasneci, Himabindu Lakkaraju)

Disentangling Algorithmic Recourse @ICLR SRML Workshop 2022 (Martin Pawelczyk, Lea Tiyavorabun, Gjergji Kasneci)

CARLA: A Python Library to Benchmark Algorithmic Recourse and Counterfactual Explanation Algorithms @NeurIPS 2021 (Martin Pawelczyk, Sascha Bielawski, Johannes van-den-Heuvel, Tobias Richter, Gjergji Kasneci)

Leveraging Model Inherent Variable Importance for Stable Online Feature Selection @KDD 2020 (Johannes Haug, Martin Pawelczyk, Klaus Broelemann und Gjergji Kasneci)

On Learning Counterfactual Explanations under Predictive Multiplicity @UAI 2020 (Martin Pawelczyk, Klaus Broelemann und Gjergji Kasneci)

Learning model agnostic counterfactual explanations for tabular data @WWW 2020 (now: The Web Conference) (Martin Pawelczyk, Klaus Broelemann und Gjergji Kasneci)