Publications

Formalising causal inference as prediction on a target population

A different way of modelling causal inference.

Practical Foundations for Probability: Prediction Methods and Calibration

How to understand probability.

We should avoid the assumption of data-generating probability distributions in social settings

We should model ML without true distributions, at least in social settings.

Position: The Categorization of Race in ML is a Flawed Premise

ML research oversimplifies race and we need to learn how to move beyond categories.

Doh, M., Höltgen, B., Riccio, P., Oliver, N.: "Position: The categorization of race in ML is a flawed premise." ICML. 2025.

Reconsidering Fairness Through Unawareness From the Perspective of Model Multiplicity

Using protected attributes can increase disparate impact without increasing accuracy.

Höltgen, B., Oliver, N.: "Reconsidering fairness through unawareness from the perspective of model multiplicity." ACM EAAMO. 2025.

On the Richness of Calibration

Exploring the concept of calibration and different ways of measuring it.

Höltgen, B., Williamson, R.C.: "On the richness of calibration." ACM FAccT. 2023.

Prioritized Training on Points that are Learnable, Worth Learning, and Not Yet Learned

An online batch selection algorithm for time-efficient training of large ML models.

Mindermann, S., Brauner, J., Razzak, M., Sharma, M., Kirsch, W., Xu, W., Höltgen, B., Gomez, A.N., Morisot, A., Farquhar, S., Gal, Y.: "Prioritized training on points that are learnable, worth learning, and not yet learned." ICML. 2022.

DeDUCE: Generating Counterfactual Explanations Efficiently

An algorithm for generating counterfactual explanations. My second master thesis.

Höltgen, B., Schut, L., Brauner, J., Gal, Y.: "DeDUCE: Generating counterfactual explanations efficiently." NeurIPS Workshop: eXplainable AI approaches for debugging and diagnosis. 2021.

Encoding Causal Macrovariables

An algorithm for causal representation learning and a reflection on causal variables. My first master thesis.

Höltgen, B.: "Encoding causal macrovariables." NeurIPS Workshop: Causal Inference & Machine Learning: Why now?. 2021.

Moral Discourse Boosts Confidence in Moral Judgments

The title says it all.

Heinzelmann, N., Höltgen, B., and Tran, V.: "Moral discourse boosts confidence in moral judgments." Philosophical Psychology 4:8, 1192-1216. 2021.

Structure-sensitive Testimonial Norms

Listening to scientific rockstars is not good for science.

Höltgen, B.: "Structure-sensitive testimonial norms." European Journal for Philosophy of Science 11:80. 2021.