Algorithmic Fairness and Color-blind Racism: Navigating the Intersection

TL;DR

This paper addresses the gap between algorithmic fairness research and race theory by applying color-blind racism as an analytic lens to diagnose and navigate the cross-disciplinary space. It analyzes how racial concepts are often decoupled from oppression in fairness discourse and identifies roadblocks and detours that impede meaningful integration. The authors offer guiding principles and critique metric choices, data practices, and race-neutral rhetoric to promote more explicit, context-aware research. By foregrounding lived experience, historical harms, and power structures, the work aims to enhance the social relevance and accountability of AI fairness interventions.

Abstract

Our focus lies at the intersection between two broader research perspectives: (1) the scientific study of algorithms and (2) the scholarship on race and racism. Many streams of research related to algorithmic fairness have been born out of interest at this intersection. We think about this intersection as the product of work derived from both sides. From (1) algorithms to (2) racism, the starting place might be an algorithmic question or method connected to a conceptualization of racism. On the other hand, from (2) racism to (1) algorithms, the starting place could be recognizing a setting where a legacy of racism is known to persist and drawing connections between that legacy and the introduction of algorithms into this setting. In either direction, meaningful disconnection can occur when conducting research at the intersection of racism and algorithms. The present paper urges collective reflection on research directions at this intersection. Despite being primarily motivated by instances of racial bias, research in algorithmic fairness remains mostly disconnected from scholarship on racism. In particular, there has not been an examination connecting algorithmic fairness discussions directly to the ideology of color-blind racism; we aim to fill this gap. We begin with a review of an essential account of color-blind racism then we review racial discourse within algorithmic fairness research and underline significant patterns, shifts and disconnects. Ultimately, we argue that researchers can improve the navigation of the landscape at the intersection by recognizing ideological shifts as such and iteratively re-orienting towards maintaining meaningful connections across interdisciplinary lines.