Decoupled Classifiers for Group-Fair and Efficient Machine Learning

Cynthia Dwork, Nicole Immorlica, Adam Tauman Kalai, Max Leiserson
Proceedings of the 1st Conference on Fairness, Accountability and Transparency, PMLR 81:119-133, 2018.

Abstract

When it is ethical and legal to use a sensitive attribute (such as gender or race) in machine learning systems, the question remains how to do so. We show that the naive application of machine learning algorithms using sensitive attributes leads to an inherent tradeoff in accuracy between groups. We provide a simple and efficient decoupling technique, that can be added on top of any black-box machine learning algorithm, to learn different classifiers for different groups. Transfer learning is used to mitigate the problem of having too little data on any one group.

Cite this Paper

BibTeX
@InProceedings{pmlr-v81-dwork18a, title = {Decoupled Classifiers for Group-Fair and Efficient Machine Learning}, author = {Dwork, Cynthia and Immorlica, Nicole and Kalai, Adam Tauman and Leiserson, Max}, booktitle = {Proceedings of the 1st Conference on Fairness, Accountability and Transparency}, pages = {119--133}, year = {2018}, editor = {Friedler, Sorelle A. and Wilson, Christo}, volume = {81}, series = {Proceedings of Machine Learning Research}, month = {23--24 Feb}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v81/dwork18a/dwork18a.pdf}, url = {https://proceedings.mlr.press/v81/dwork18a.html}, abstract = {When it is ethical and legal to use a sensitive attribute (such as gender or race) in machine learning systems, the question remains how to do so. We show that the naive application of machine learning algorithms using sensitive attributes leads to an inherent tradeoff in accuracy between groups. We provide a simple and efficient decoupling technique, that can be added on top of any black-box machine learning algorithm, to learn different classifiers for different groups. Transfer learning is used to mitigate the problem of having too little data on any one group.} }
Endnote
%0 Conference Paper %T Decoupled Classifiers for Group-Fair and Efficient Machine Learning %A Cynthia Dwork %A Nicole Immorlica %A Adam Tauman Kalai %A Max Leiserson %B Proceedings of the 1st Conference on Fairness, Accountability and Transparency %C Proceedings of Machine Learning Research %D 2018 %E Sorelle A. Friedler %E Christo Wilson %F pmlr-v81-dwork18a %I PMLR %P 119--133 %U https://proceedings.mlr.press/v81/dwork18a.html %V 81 %X When it is ethical and legal to use a sensitive attribute (such as gender or race) in machine learning systems, the question remains how to do so. We show that the naive application of machine learning algorithms using sensitive attributes leads to an inherent tradeoff in accuracy between groups. We provide a simple and efficient decoupling technique, that can be added on top of any black-box machine learning algorithm, to learn different classifiers for different groups. Transfer learning is used to mitigate the problem of having too little data on any one group.
APA
Dwork, C., Immorlica, N., Kalai, A.T. & Leiserson, M.. (2018). Decoupled Classifiers for Group-Fair and Efficient Machine Learning. Proceedings of the 1st Conference on Fairness, Accountability and Transparency, in Proceedings of Machine Learning Research 81:119-133 Available from https://proceedings.mlr.press/v81/dwork18a.html.

Related Material