Causality Concepts in Machine Learning: Heterogeneous Treatment Effect Estimation With Machine Learning and Model Interpretation With Counterfactual and Semi-Factual Explanations
MCML Authors
Susanne Dandl
Dr.
* Former Member
Abstract
Susanne Dandl
Dr.
* Former Member
Abstract
This thesis explores the growing intersection of machine learning and causality through seven articles, offering new insights into how these fields can enhance one another. It addresses key topics, including adapting machine learning algorithms for heterogeneous treatment effect estimation, where combining causal and model-based forest elements improves performance across diverse datasets. Additionally, the thesis introduces advanced interpretability tools, proposing methods to generate multiple counterfactual and semi-factual explanations that aid in fairness assessments and address interpretability challenges. A modular R package developed in this work provides accessible tools for researchers to apply and compare counterfactual explanation methods, further bridging machine learning and causal inference for practical applications. (Shortened).
BibTeXKey: Dan23