Data justice in the “twin objective” of market and risk: How discrimination is formulated in EU's AI policy

Abstract

AbstractBased on a focus on artificial intelligence (AI) policy in the European Union (EU), we explore the dominant approach taken to data justice in policy. More specifically, we ask how the particular issue of discrimination is translated into policy goals and measures as a way to address prominent concerns about AI. Looking at the stage of policy formulation, we provide an analysis of the way (non) discrimination is currently pursued within the EU's AI policy debate through the study of relevant policy documents and public consultations between 2017 and 2023. We argue that whilst the issue of discrimination has moved from the margins to the mainstream in policy debate, it has done so based on an understanding of discrimination as an inevitable risk of AI; such risk is specific to particular situations and the technological features of AI; the nature of this risk can be assessed and managed through a set of procedural safeguards; and such safeguards can be supported by the creation of a trustworthy AI market. Whilst this translation of justice is very important for contending with some of the critique surrounding the advancement of AI, it may also serve to contain and neutralize such critique in the interest of marketization.