@inproceedings {pub6120pub6267,
	title = {Evaluating and Mitigating Fairness Risks in Machine Learning under Structured Missing Data},
	author = {Tim Kochs AND Andrea Castellani AND Felix Lanfermann AND Barbara Hammer},
	year = {2025},
	month = {August},
	abstract = {Handling missing values in machine learning presents significant challenges, impacting both model predictive performance and fairness. 
While unstructured missingness, where values are absent randomly, has been extensively studied, structured missingness remains relatively underexplored. Structured missingness occurs when the absence of one or more features correlates with or influences the absence of others.
In this paper, we investigate the impact of both structured and unstructured missingness on model performance and fairness. 
We introduce a novel approach for simulating structured missingness and propose a training methodology designed to enhance model robustness under these conditions. 
Specifically, we evaluate how different strategies for handling missing data affect fairness and performance, conducting comprehensive experiments using state-of-the-art imputation methods across multiple benchmark tabular datasets.
Our empirical results demonstrate that structured missingness significantly undermines model fairness, particularly when the missingness mechanism affects minority groups. 
Our proposed method achieves up to a 50\% reduction in equalized odds difference while maintaining predictive accuracy.},
	publisher = {CEUR},
	url = {https://ceur-ws.org/Vol-4147/paper12.pdf},
	booktitle = {AEQUITAS 2025 },
	volume = {4147},
	pages = {17},
	series = {AI*IA}
}