Fairness of Exposure in Stochastic Bandits

Lequn Wang, Yiwei Bai, Wen Sun, Thorsten Joachims
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:10686-10696, 2021.

Abstract

Contextual bandit algorithms have become widely used for recommendation in online systems (e.g. marketplaces, music streaming, news), where they now wield substantial influence on which items get shown to users. This raises questions of fairness to the items — and to the sellers, artists, and writers that benefit from this exposure. We argue that the conventional bandit formulation can lead to an undesirable and unfair winner-takes-all allocation of exposure. To remedy this problem, we propose a new bandit objective that guarantees merit-based fairness of exposure to the items while optimizing utility to the users. We formulate fairness regret and reward regret in this setting and present algorithms for both stochastic multi-armed bandits and stochastic linear bandits. We prove that the algorithms achieve sublinear fairness regret and reward regret. Beyond the theoretical analysis, we also provide empirical evidence that these algorithms can allocate exposure to different arms effectively.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-wang21b, title = {Fairness of Exposure in Stochastic Bandits}, author = {Wang, Lequn and Bai, Yiwei and Sun, Wen and Joachims, Thorsten}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {10686--10696}, 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/wang21b/wang21b.pdf}, url = {https://proceedings.mlr.press/v139/wang21b.html}, abstract = {Contextual bandit algorithms have become widely used for recommendation in online systems (e.g. marketplaces, music streaming, news), where they now wield substantial influence on which items get shown to users. This raises questions of fairness to the items — and to the sellers, artists, and writers that benefit from this exposure. We argue that the conventional bandit formulation can lead to an undesirable and unfair winner-takes-all allocation of exposure. To remedy this problem, we propose a new bandit objective that guarantees merit-based fairness of exposure to the items while optimizing utility to the users. We formulate fairness regret and reward regret in this setting and present algorithms for both stochastic multi-armed bandits and stochastic linear bandits. We prove that the algorithms achieve sublinear fairness regret and reward regret. Beyond the theoretical analysis, we also provide empirical evidence that these algorithms can allocate exposure to different arms effectively.} }
Endnote
%0 Conference Paper %T Fairness of Exposure in Stochastic Bandits %A Lequn Wang %A Yiwei Bai %A Wen Sun %A Thorsten Joachims %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-wang21b %I PMLR %P 10686--10696 %U https://proceedings.mlr.press/v139/wang21b.html %V 139 %X Contextual bandit algorithms have become widely used for recommendation in online systems (e.g. marketplaces, music streaming, news), where they now wield substantial influence on which items get shown to users. This raises questions of fairness to the items — and to the sellers, artists, and writers that benefit from this exposure. We argue that the conventional bandit formulation can lead to an undesirable and unfair winner-takes-all allocation of exposure. To remedy this problem, we propose a new bandit objective that guarantees merit-based fairness of exposure to the items while optimizing utility to the users. We formulate fairness regret and reward regret in this setting and present algorithms for both stochastic multi-armed bandits and stochastic linear bandits. We prove that the algorithms achieve sublinear fairness regret and reward regret. Beyond the theoretical analysis, we also provide empirical evidence that these algorithms can allocate exposure to different arms effectively.
APA
Wang, L., Bai, Y., Sun, W. & Joachims, T.. (2021). Fairness of Exposure in Stochastic Bandits. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:10686-10696 Available from https://proceedings.mlr.press/v139/wang21b.html.

Related Material