Equitable AI Audits: evaluating the evaluators in today's world

Abstract

Artificial Intelligence (AI) technologies have triggered substantial transformations across various sectors, yielding benefits and rasing concerns. A significant issue pertains to the unequal distribution of AI advantages among different groups, raising questions about diversity and fundamental rights. This paper proposes equity as a translational mechanism intersecting ethical, legal, and technical dimensions in AI governance. By tracing the evolution of AI regulation from corporate ethical guidelines to forthcoming legal frameworks, this paper sheds light on the growing emphasis on auditing mechanisms to mitigate bias and discrimination, distinguishing equity from the more often-used term fairness. While fairness emphasizes fair processes, equity prioritizes equitable outcomes. We argue that equity offers a value that surpasses the limitations of fairness and adopts a holistic viewpoint, providing a comprehensive framework for tackling socio-technical challenges in AI development and deployment, and preserving fundamental rights. In proposing a definition of equity that draws inspiration from feminist perspectives rather than traditional philosophy, we highlight equity’s contextual nature across diverse global regions and cultures. This framing lets us propose equity as a non-universal value that can be pivotal in curating datasets, fostering community empowerment, and engendering a sense of belonging within the diverse communities affected by the AI tools used in decision-making processes. We further argue that ethics-based auditing serves as a crucial link between technical requirements and legal norms, offering insights into the complex socio-technical dynamics of AI development. This paper critically examines the integration of equity in AI audits, assessing whether current audits adequately address non-discrimination and fundamental rights or merely scratch the surface. With this paper, we wish to contribute to scholarly discussions on AI’s role in preserving fundamental rights, and stress the importance of interdisciplinary collaboration in navigating AI’s multifaceted challenges.

Ähnliche Arbeiten