From Pseudorandomness to Multi-Group Fairness and Back

Cynthia Dwork, Daniel Lee, Huijia Lin, Pranay Tankala
Proceedings of Thirty Sixth Conference on Learning Theory, PMLR 195:3566-3614, 2023.

Abstract

We identify and explore connections between the recent literature on multi-group fairness for prediction algorithms and the pseudorandomness notions of leakage-resilience and graph regularity. We frame our investigation using new, statistical distance-based variants of multicalibration that are closely related to the concept of outcome indistinguishability. Adopting this perspective leads us naturally not only to our graph theoretic results, but also to new, more efficient algorithms for multicalibration in certain parameter regimes and a novel proof of a hardcore lemma for real-valued functions.

Cite this Paper

BibTeX
@InProceedings{pmlr-v195-dwork23a, title = {From Pseudorandomness to Multi-Group Fairness and Back}, author = {Dwork, Cynthia and Lee, Daniel and Lin, Huijia and Tankala, Pranay}, booktitle = {Proceedings of Thirty Sixth Conference on Learning Theory}, pages = {3566--3614}, year = {2023}, editor = {Neu, Gergely and Rosasco, Lorenzo}, volume = {195}, series = {Proceedings of Machine Learning Research}, month = {12--15 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v195/dwork23a/dwork23a.pdf}, url = {https://proceedings.mlr.press/v195/dwork23a.html}, abstract = {We identify and explore connections between the recent literature on multi-group fairness for prediction algorithms and the pseudorandomness notions of leakage-resilience and graph regularity. We frame our investigation using new, statistical distance-based variants of multicalibration that are closely related to the concept of outcome indistinguishability. Adopting this perspective leads us naturally not only to our graph theoretic results, but also to new, more efficient algorithms for multicalibration in certain parameter regimes and a novel proof of a hardcore lemma for real-valued functions.} }
Endnote
%0 Conference Paper %T From Pseudorandomness to Multi-Group Fairness and Back %A Cynthia Dwork %A Daniel Lee %A Huijia Lin %A Pranay Tankala %B Proceedings of Thirty Sixth Conference on Learning Theory %C Proceedings of Machine Learning Research %D 2023 %E Gergely Neu %E Lorenzo Rosasco %F pmlr-v195-dwork23a %I PMLR %P 3566--3614 %U https://proceedings.mlr.press/v195/dwork23a.html %V 195 %X We identify and explore connections between the recent literature on multi-group fairness for prediction algorithms and the pseudorandomness notions of leakage-resilience and graph regularity. We frame our investigation using new, statistical distance-based variants of multicalibration that are closely related to the concept of outcome indistinguishability. Adopting this perspective leads us naturally not only to our graph theoretic results, but also to new, more efficient algorithms for multicalibration in certain parameter regimes and a novel proof of a hardcore lemma for real-valued functions.
APA
Dwork, C., Lee, D., Lin, H. & Tankala, P.. (2023). From Pseudorandomness to Multi-Group Fairness and Back. Proceedings of Thirty Sixth Conference on Learning Theory, in Proceedings of Machine Learning Research 195:3566-3614 Available from https://proceedings.mlr.press/v195/dwork23a.html.

Related Material