Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Preliminary Framework for Intersectionality in ML Pipelines

Michelle Nashla Turcios, Alicia E. Boyd, Angela D. R. Smith, Brittany Johnson

TL;DR

The paper tackles biases in ML systems by advocating a principled, theory-grounded use of intersectionality to capture complex social identities. It develops a framework anchored in Crenshaw, the Combahee River Collective, and Collins, and applies it to critically analyze three case studies, extracting five guiding principles (G1–G5) to assess relationality, power dynamics, historical specificity, feature methods, and ethics. Through a replication study and a second experimental phase with dataset variations, the authors reveal substantial misalignments in prior applications and demonstrate how data choices and methodological decisions shape outcomes. The work offers concrete guidelines and emphasizes reflexivity, power, and interdisciplinarity as essential for translating intersectionality into equitable ML practice, with practical implications for researchers and practitioners seeking responsible, transparent, and just pipeline designs.

Abstract

Machine learning (ML) has become a go-to solution for improving how we use, experience, and interact with technology (and the world around us). Unfortunately, studies have repeatedly shown that machine learning technologies may not provide adequate support for societal identities and experiences. Intersectionality is a sociological framework that provides a mechanism for explicitly considering complex social identities, focusing on social justice and power. While the framework of intersectionality can support the development of technologies that acknowledge and support all members of society, it has been adopted and adapted in ways that are not always true to its foundations, thereby weakening its potential for impact. To support the appropriate adoption and use of intersectionality for more equitable technological outcomes, we amplify the foundational intersectionality scholarship--Crenshaw, Combahee, and Collins (three C's), to create a socially relevant preliminary framework in developing machine-learning solutions. We use this framework to evaluate and report on the (mis)alignments of intersectionality application in machine learning literature.

A Preliminary Framework for Intersectionality in ML Pipelines

TL;DR

The paper tackles biases in ML systems by advocating a principled, theory-grounded use of intersectionality to capture complex social identities. It develops a framework anchored in Crenshaw, the Combahee River Collective, and Collins, and applies it to critically analyze three case studies, extracting five guiding principles (G1–G5) to assess relationality, power dynamics, historical specificity, feature methods, and ethics. Through a replication study and a second experimental phase with dataset variations, the authors reveal substantial misalignments in prior applications and demonstrate how data choices and methodological decisions shape outcomes. The work offers concrete guidelines and emphasizes reflexivity, power, and interdisciplinarity as essential for translating intersectionality into equitable ML practice, with practical implications for researchers and practitioners seeking responsible, transparent, and just pipeline designs.

Abstract

Machine learning (ML) has become a go-to solution for improving how we use, experience, and interact with technology (and the world around us). Unfortunately, studies have repeatedly shown that machine learning technologies may not provide adequate support for societal identities and experiences. Intersectionality is a sociological framework that provides a mechanism for explicitly considering complex social identities, focusing on social justice and power. While the framework of intersectionality can support the development of technologies that acknowledge and support all members of society, it has been adopted and adapted in ways that are not always true to its foundations, thereby weakening its potential for impact. To support the appropriate adoption and use of intersectionality for more equitable technological outcomes, we amplify the foundational intersectionality scholarship--Crenshaw, Combahee, and Collins (three C's), to create a socially relevant preliminary framework in developing machine-learning solutions. We use this framework to evaluate and report on the (mis)alignments of intersectionality application in machine learning literature.
Paper Structure (32 sections, 14 figures, 1 table)

This paper contains 32 sections, 14 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background
  3. Intersectionality & Machine Learning
  4. Intersectionality & Data Processes
  5. Intersectionality as a Methodology
  6. Additive vs. Intersectional Approaches
  7. Intra-Anti-Inter Categorical Approach
  8. Intersectionality & Colonialism
  9. Methodology
  10. Case Study
  11. Analyzing Alignment with Theory
  12. Intersectionality as a Relational Analysis (G1)
  13. Social Formations of Complex Social Inequalities (G2)
  14. Historical and Cultural Specificity (G3)
  15. Feature Engineering and Statistical Methods (G4)
  16. ...and 17 more sections

Figures (14)

  • Figure 1: Average Cosine Similarity Between Four SocialC ategories and Fifty Most Similar Words to Four Social Institutions
  • Figure 2: Differences Between Mean Cosine Similarity for Six Pairwise Social Categories across Four Social Institutions(No Changes)
  • Figure 3: First Two Dimensions from PCA of 50 Words With Highest Cosine Similarity to Each of Four Institutions(NoChanges)
  • Figure 4: Words with the Highest Cosine Similarity to Four Social Institution Vectors(NoChanges)
  • Figure 5: Average Cosine Similarity Between Four Social Categories and Fifty Most Similar Words to Four Social Institutions(OriginalData)
  • ...and 9 more figures