Fairness and machine learning
Limitations and Opportunities
Solon Barocas, Moritz Hardt, Arvind Narayanan
This online textbook is an incomplete work in progress. Essential chapters are still missing. In the spirit of open review, we solicit broad feedback that will influence existing chapters, as well as the development of later material. Please contact us at contact@fairmlbook.org.
| CONTENTS | ||
| Preface | ||
| Acknowledgments | ||
| 1 | Introduction | |
| 2 | When is automated decision making legitimate? | |
| We explore what makes automated decision making a matter of normative concern, situated in bureaucratic decision making and its mechanical application of formalized rules. | ||
| 3 | Classification | |
| We introduce formal non-discrimination criteria in a decision-theoretic setting, establish their relationships, and illustrate their limitations. | ||
| 4 | Relative notions of fairness | |
| We introduce formal non-discrimination criteria in a decision-theoretic setting, establish their relationships, and illustrate their limitations. | ||
| 5 | Causality | |
| We dive into the rich technical repertoire of causal inference and how it helps articulate and address shortcomings of the classification paradigm, while raising new conceptual and normative questions. | ||
| 6 | Understanding United States anti-discrimination law | |
| We discuss what United States anti-discrimination law is and isn’t, how it navigates tradeoffs, its limits, and how it applies to machine learning. | ||
| 7 | Testing discrimination in practice | |
| We systematize tests of discrimination and discuss the practical complexities of applying them, both to traditional decision-making systems and to algorithmic systems. | ||
| 8 | A broader view of discrimination | |
| We review structural, organizational, and interpersonal discrimination in society, how machine learning interacts with them, and discuss a broad set of potential interventions. | ||
| 9 | Datasets | |
| Datasets are the backbone of machine learning research and development. We critically examine their role, the harms associated with data, and survey improvements in data practices. | ||
| 10 | Algorithmic interventions | |
| In this online-only chapter, we survey and systematize a burgeoning set of algorithmic interventions aimed at promoting fairness, while highlighting the limitations of this paradigm. | ||
Video tutorials
- Fairness and Machine Learning (Part 1, Part 2) (MLSS 2020)
- Fairness in machine learning (NeurIPS 2017)
- 21 fairness definitions and their politics (FAccT 2018)
Course materials
Citations
To cite this book, please use this bibtex entry:
@book{barocas-hardt-narayanan,
title = {Fairness and Machine Learning},
author = {Solon Barocas and Moritz Hardt and Arvind Narayanan},
publisher = {fairmlbook.org},
note = {\url{http://www.fairmlbook.org}},
year = {2019}
}
Last updated: Tue Aug 2 04:45:10 PDT 2022