Fairness Beyond the Algorithmic Frame: Actionable Recommendations for an Intersectional Approach

ACM Conference on Fairness, Accountability, and Transparency (FAccT), pp. 3276–3290

Abstract

Achieving fair use of AI systems is a multi-faceted challenge.Intersectionality, rooted in Black Feminist movements, is increasingly used to address the interconnected nature of discrimination such as racism, ableism, and sexism.Yet in AI research, intersectionality is often reduced to a narrow technical lens, focused on algorithmic bias between subgroups defined by protected attributes and addressed through fairness metrics.This algorithmic frame sidelines key aspects of intersectionality, such as power relations, social justice, and structural inequality.Still, AI experts play a central role in development and deployment, and therefore should act to limit unjust outcomes.This study offers actionable guidance for AI experts, grounded in a broader intersectional perspective.Through a thematic analysis of AI fairness papers on key aspects of intersectionality, evaluated through community engagement, we identify five themes with concrete recommendations: 1) insisting on collaboration in interdisciplinary teams, 2) embedding reflection and recognizing positionality, 3) approaching communities and facilitating co-ownership, 4) engaging with power dynamics and social context, and 5) assessing the framing and nuance of data and metrics.Participating experts noted barriers such as tech-optimism and fear of insufficient knowledge.Still, they valued the recommendations for communicating the importance of intersectionality and initiating more just AI practices.We call on AI experts to meet this challenge through interdisciplinary collaboration with diverse communities.

Authors 4

1. Steven Vethman

   Netherlands Organisation for Applied Scientific Research · Institut d'Etudes Politiques de Paris

   Affiliation as printed

   Sciences Po Law School, Paris, France and TNO, The Hague, Netherlands

2. Quirine T. S. Smit

   Netherlands Organisation for Applied Scientific Research

   Affiliation as printed

   TNO, The Hague, Netherlands

3. Nina M. van Liebergen

   Netherlands Organisation for Applied Scientific Research

   Affiliation as printed

   TNO, The Hague, Netherlands

4. Cor J. Veenman Aachen

   Leiden University · Netherlands Organisation for Applied Scientific Research

   Affiliation as printed

   LIACS, Leiden University, Leiden, Netherlands and TNO, The Hague, Netherlands