Pragmatic Fairness: Developing Policies with Outcome Disparity Control

Limor Gultchin, Siyuan Guo, Alan Malek, Silvia Chiappa, Ricardo Silva
Proceedings of the Third Conference on Causal Learning and Reasoning, PMLR 236:243-264, 2024.

Abstract

We introduce a causal framework for designing optimal policies that satisfy classes of fairness constraints. We take a pragmatic approach asking what we can do with an action space available from historical data, with no further experimentation and novel actions immediately available. We propose two different fairness constraints: a "moderation breaking" constraint which aims at reducing disparity in outcome levels across sensitive attributes to the extent the provided action space permits; and an "equal benefit" constraint which aims at distributing gain from the new and maximized policy equally across sensitive attribute levels, and thus at keeping pre-existing preferential treatment in place or avoiding the introduction of new disparity. We introduce practical methods for implementing the constraints and illustrate their uses on experiments with semi-synthetic models.

Cite this Paper

BibTeX
@InProceedings{pmlr-v236-gultchin24a, title = {Pragmatic Fairness: Developing Policies with Outcome Disparity Control}, author = {Gultchin, Limor and Guo, Siyuan and Malek, Alan and Chiappa, Silvia and Silva, Ricardo}, booktitle = {Proceedings of the Third Conference on Causal Learning and Reasoning}, pages = {243--264}, year = {2024}, editor = {Locatello, Francesco and Didelez, Vanessa}, volume = {236}, series = {Proceedings of Machine Learning Research}, month = {01--03 Apr}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v236/gultchin24a/gultchin24a.pdf}, url = {https://proceedings.mlr.press/v236/gultchin24a.html}, abstract = {We introduce a causal framework for designing optimal policies that satisfy classes of fairness constraints. We take a pragmatic approach asking what we can do with an action space available from historical data, with no further experimentation and novel actions immediately available. We propose two different fairness constraints: a "moderation breaking" constraint which aims at reducing disparity in outcome levels across sensitive attributes to the extent the provided action space permits; and an "equal benefit" constraint which aims at distributing gain from the new and maximized policy equally across sensitive attribute levels, and thus at keeping pre-existing preferential treatment in place or avoiding the introduction of new disparity. We introduce practical methods for implementing the constraints and illustrate their uses on experiments with semi-synthetic models.} }
Endnote
%0 Conference Paper %T Pragmatic Fairness: Developing Policies with Outcome Disparity Control %A Limor Gultchin %A Siyuan Guo %A Alan Malek %A Silvia Chiappa %A Ricardo Silva %B Proceedings of the Third Conference on Causal Learning and Reasoning %C Proceedings of Machine Learning Research %D 2024 %E Francesco Locatello %E Vanessa Didelez %F pmlr-v236-gultchin24a %I PMLR %P 243--264 %U https://proceedings.mlr.press/v236/gultchin24a.html %V 236 %X We introduce a causal framework for designing optimal policies that satisfy classes of fairness constraints. We take a pragmatic approach asking what we can do with an action space available from historical data, with no further experimentation and novel actions immediately available. We propose two different fairness constraints: a "moderation breaking" constraint which aims at reducing disparity in outcome levels across sensitive attributes to the extent the provided action space permits; and an "equal benefit" constraint which aims at distributing gain from the new and maximized policy equally across sensitive attribute levels, and thus at keeping pre-existing preferential treatment in place or avoiding the introduction of new disparity. We introduce practical methods for implementing the constraints and illustrate their uses on experiments with semi-synthetic models.
APA
Gultchin, L., Guo, S., Malek, A., Chiappa, S. & Silva, R.. (2024). Pragmatic Fairness: Developing Policies with Outcome Disparity Control. Proceedings of the Third Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 236:243-264 Available from https://proceedings.mlr.press/v236/gultchin24a.html.

Related Material