Fair Learning with Private Demographic Data

Hussein Mozannar, Mesrob Ohannessian, Nathan Srebro
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:7066-7075, 2020.

Abstract

Sensitive attributes such as race are rarely available to learners in real world settings as their collection is often restricted by laws and regulations. We give a scheme that allows individuals to release their sensitive information privately while still allowing any downstream entity to learn non-discriminatory predictors. We show how to adapt non-discriminatory learners to work with privatized protected attributes giving theoretical guarantees on performance. Finally, we highlight how the methodology could apply to learning fair predictors in settings where protected attributes are only available for a subset of the data.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-mozannar20a, title = {Fair Learning with Private Demographic Data}, author = {Mozannar, Hussein and Ohannessian, Mesrob and Srebro, Nathan}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {7066--7075}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/mozannar20a/mozannar20a.pdf}, url = {https://proceedings.mlr.press/v119/mozannar20a.html}, abstract = {Sensitive attributes such as race are rarely available to learners in real world settings as their collection is often restricted by laws and regulations. We give a scheme that allows individuals to release their sensitive information privately while still allowing any downstream entity to learn non-discriminatory predictors. We show how to adapt non-discriminatory learners to work with privatized protected attributes giving theoretical guarantees on performance. Finally, we highlight how the methodology could apply to learning fair predictors in settings where protected attributes are only available for a subset of the data.} }
Endnote
%0 Conference Paper %T Fair Learning with Private Demographic Data %A Hussein Mozannar %A Mesrob Ohannessian %A Nathan Srebro %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-mozannar20a %I PMLR %P 7066--7075 %U https://proceedings.mlr.press/v119/mozannar20a.html %V 119 %X Sensitive attributes such as race are rarely available to learners in real world settings as their collection is often restricted by laws and regulations. We give a scheme that allows individuals to release their sensitive information privately while still allowing any downstream entity to learn non-discriminatory predictors. We show how to adapt non-discriminatory learners to work with privatized protected attributes giving theoretical guarantees on performance. Finally, we highlight how the methodology could apply to learning fair predictors in settings where protected attributes are only available for a subset of the data.
APA
Mozannar, H., Ohannessian, M. & Srebro, N.. (2020). Fair Learning with Private Demographic Data. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:7066-7075 Available from https://proceedings.mlr.press/v119/mozannar20a.html.

Related Material