Measuring and Mitigating Bias for Tabular Datasets with Multiple Protected Attributes

arXiv:2405.19300v1 Announce Type: new
Abstract: Motivated by the recital (67) of the current corrigendum of the AI Act in the European Union, we propose and present measures and mitigation strategies for discrimination in tabular datasets. We specifically focus on datasets that contain multiple protected attributes, such as nationality, age, and sex. This makes measuring and mitigating bias more challenging, as many existing methods are designed for a single protected attribute. This paper comes with a twofold contribution: Firstly, new discrimination …

abstract act age ai act arxiv attributes bias cs.ai cs.lg current datasets discrimination european union focus measuring multiple strategies tabular type union