Addressing algorithmic discrimination in healthcare through an intersectional lens - a comparative legal perspective

The growing use of artificial intelligence (AI) in clinical decision-making offers new powerful tools for addressing pressing health challenges. However, clinical algorithms are not neutral and can mirror prevailing patterns of power and disadvantage in healthcare. The nature of these disparities and discriminatory practices is often intersectional, meaning that the disadvantage is based on a complex synergy of protected grounds. These intersecting axes of inequality can be entrenched and exacerbated by algorithms at various points of their lifecycle. However, intersectional discrimination continues to escape the scrutiny of antidiscrimination law which is predominantly based on specific protected grounds considered in isolation. Similarly, fairness interventions proposed in the fast-growing computer science literature are mostly designed around single categorical attributes. By assuming that persons having a protected identity create homogenous groups, both the law and the technology reinforce a single-axis approach, failing to protect patients with intersectional identities against algorithmic discrimination. To address this problem, the present thesis proposes an intersectional approach to the development and regulation of AI-based clinical algorithms. It begins by reconstructing the main elements of the intersectionality theory and operationalising it in the context of health disparities research. Subsequently, the thesis investigates how the concept of intersectionality has been developed in computer science literature concerning algorithmic fairness and the law. The geographical scope of the legal analysis is limited to the EU and the US, two global leaders in AI regulation and innovation. It offers a comparative perspective on the EU and US regulatory landscapes regarding intersectional discrimination in healthcare and the regulation of bias in AI-based clinical tools. Finally, the thesis proposes an intersectional fairness assessment framework, which is intended to support researchers, clinical AI developers and regulators, fostering trans-Atlantic collaboration on fair clinical AI.