Fair Selective Classification Via Sufficiency

Joshua K Lee, Yuheng Bu, Deepta Rajan, Prasanna Sattigeri, Rameswar Panda, Subhro Das, Gregory W Wornell
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:6076-6086, 2021.

Abstract

Selective classification is a powerful tool for decision-making in scenarios where mistakes are costly but abstentions are allowed. In general, by allowing a classifier to abstain, one can improve the performance of a model at the cost of reducing coverage and classifying fewer samples. However, recent work has shown, in some cases, that selective classification can magnify disparities between groups, and has illustrated this phenomenon on multiple real-world datasets. We prove that the sufficiency criterion can be used to mitigate these disparities by ensuring that selective classification increases performance on all groups, and introduce a method for mitigating the disparity in precision across the entire coverage scale based on this criterion. We then provide an upper bound on the conditional mutual information between the class label and sensitive attribute, conditioned on the learned features, which can be used as a regularizer to achieve fairer selective classification. The effectiveness of the method is demonstrated on the Adult, CelebA, Civil Comments, and CheXpert datasets.

Cite this Paper


BibTeX
@InProceedings{pmlr-v139-lee21b, title = {Fair Selective Classification Via Sufficiency}, author = {Lee, Joshua K and Bu, Yuheng and Rajan, Deepta and Sattigeri, Prasanna and Panda, Rameswar and Das, Subhro and Wornell, Gregory W}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {6076--6086}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/lee21b/lee21b.pdf}, url = {https://proceedings.mlr.press/v139/lee21b.html}, abstract = {Selective classification is a powerful tool for decision-making in scenarios where mistakes are costly but abstentions are allowed. In general, by allowing a classifier to abstain, one can improve the performance of a model at the cost of reducing coverage and classifying fewer samples. However, recent work has shown, in some cases, that selective classification can magnify disparities between groups, and has illustrated this phenomenon on multiple real-world datasets. We prove that the sufficiency criterion can be used to mitigate these disparities by ensuring that selective classification increases performance on all groups, and introduce a method for mitigating the disparity in precision across the entire coverage scale based on this criterion. We then provide an upper bound on the conditional mutual information between the class label and sensitive attribute, conditioned on the learned features, which can be used as a regularizer to achieve fairer selective classification. The effectiveness of the method is demonstrated on the Adult, CelebA, Civil Comments, and CheXpert datasets.} }
Endnote
%0 Conference Paper %T Fair Selective Classification Via Sufficiency %A Joshua K Lee %A Yuheng Bu %A Deepta Rajan %A Prasanna Sattigeri %A Rameswar Panda %A Subhro Das %A Gregory W Wornell %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-lee21b %I PMLR %P 6076--6086 %U https://proceedings.mlr.press/v139/lee21b.html %V 139 %X Selective classification is a powerful tool for decision-making in scenarios where mistakes are costly but abstentions are allowed. In general, by allowing a classifier to abstain, one can improve the performance of a model at the cost of reducing coverage and classifying fewer samples. However, recent work has shown, in some cases, that selective classification can magnify disparities between groups, and has illustrated this phenomenon on multiple real-world datasets. We prove that the sufficiency criterion can be used to mitigate these disparities by ensuring that selective classification increases performance on all groups, and introduce a method for mitigating the disparity in precision across the entire coverage scale based on this criterion. We then provide an upper bound on the conditional mutual information between the class label and sensitive attribute, conditioned on the learned features, which can be used as a regularizer to achieve fairer selective classification. The effectiveness of the method is demonstrated on the Adult, CelebA, Civil Comments, and CheXpert datasets.
APA
Lee, J.K., Bu, Y., Rajan, D., Sattigeri, P., Panda, R., Das, S. & Wornell, G.W.. (2021). Fair Selective Classification Via Sufficiency. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:6076-6086 Available from https://proceedings.mlr.press/v139/lee21b.html.

Related Material