Instrumental Variable Methods for Sequential Experiment Design

MCML Authors

Elisabeth Ailer

* Former Member

→ Group Niki Kilbertus
 Ethics in Systems Design and Machine Learning

Abstract

This thesis develops instrumental variable (IV) techniques for sequential experimental design to quantify causal relationships in high-dimensional systems with hidden confounders. Motivated by challenges in biological datasets—such as high dimensionality, hidden confounding, nonlinear effects, and limited resources—we address these issues by extending IV methods to underspecified linear and nonlinear settings, refining bounds for causal effects, and optimizing experiment selection.

phdthesis Ail25

---

Dissertation

TU München. Dec. 2025

---

Authors

E. Ailer

---

Links

URL

---

Research Area

 A3 | Computational Models