Reconciling Legal and Technical Approaches to Algorithmic Bias

Alice Xiang

Tennessee Law Review



In recent years, there has been a proliferation of papers in the algorithmic fairness literature proposing various technical definitions of algorithmic bias and methods to mitigate bias. Whether these algorithmic bias mitigation methods would be permissible from a legal perspective is a complex but increasingly pressing question at a time when there are growing concerns about the potential for algorithmic decision-making to exacerbate societal inequities. In particular, there is a tension around the use of protected class variables: most algorithmic bias mitigation techniques utilize these variables or proxies, but anti-discrimination doctrine has a strong preference for decisions that are blind to them. This Article analyzes the extent to which technical approaches to algorithmic bias are compatible with U.S. anti-discrimination law and recommends a path toward greater compatibility. This question is vital to address because a lack of legal compatibility creates the possibility that biased algorithms might be considered legally permissible while approaches designed to correct for bias might be considered illegally discriminatory. For example, a recent proposed rule from the Department of Housing and Urban Development (“HUD”), which would have established the first instance of a U.S. regulatory definition for algorithmic discrimination, would have created a safe harbor from disparate impact liability for housing-related algorithms that do not use protected class variables or close proxies. An abundance of recent scholarship has shown, however, that simply removing protected class variables and close proxies does little to ensure that the algorithm will not be biased. In fact, this approach, known as “fairness through unawareness” in the machine learning community, is widely considered naive. While the language around algorithms was removed in the final rule, this focus on the visibility of protected attributes in decision-making is central in U.S. anti-discrimination law. Causal inference provides a potential way to reconcile algorithmic fairness techniques with anti-discrimination law. In U.S. law, discrimination is generally thought of as making decisions “because of” a protected class variable. In fact, in Texas Department of Housing and Community Affairs v. Inclusive Communities Project, Inc., the case that motivated the HUD proposed rule, the Court required a “causal connection” between the decision-making process and the disproportionate outcomes. Instead of examining whether protected class variables appear in the algorithm, causal inference would allow for techniques that use protected class variables with the intent of negating causal relationships in the data tied with race. While moving from correlation to causation is challenging—particularly in machine learning, where leveraging correlations to make accurate predictions is typically the goal—doing so offers a way to reconcile technical feasibility and legal precedence while providing protections against algorithmic bias.

Related Publications

Causality for Temporal Unfairness Evaluation and Mitigation

NeurIPS, 2022
Aida Rahmattalabi, Alice Xiang

Recent interests in causality for fair decision-making systems has been accompanied with great skepticism due to practical and epistemological challenges with applying existing causal fairness approaches. Existing works mainly seek to remove the causal effect of social categ…

Men Also Do Laundry: Multi-Attribute Bias Amplification

NeurIPS, 2022
Dora Zhao, Jerone T. A. Andrews, Alice Xiang

As computer vision systems become more widely deployed, there is increasing concern from both the research community and the public that these systems are not only reproducing but amplifying harmful social biases. The phenomenon of bias amplification, which is the focus of t…

A View From Somewhere: Human-Centric Face Representations

NeurIPS, 2022
Jerone T. A. Andrews, Przemyslaw Joniak*, Alice Xiang

We propose to implicitly learn a set of continuous face-varying dimensions, without ever asking an annotator to explicitly categorize a person. We uncover the dimensions by learning on a novel dataset of 638,180 human judgments of face similarity (FAX). We demonstrate the ut…

  • HOME
  • Publications
  • Reconciling Legal and Technical Approaches to Algorithmic Bias


Shape the Future of AI with Sony AI

We want to hear from those of you who have a strong desire
to shape the future of AI.