Causal Strategic Linear Regression

Yonadav Shavit, Benjamin Edelman, Brian Axelrod
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:8676-8686, 2020.

Abstract

In many predictive decision-making scenarios, such as credit scoring and academic testing, a decision-maker must construct a model that accounts for agents’ propensity to “game” the decision rule by changing their features so as to receive better decisions. Whereas the strategic classification literature has previously assumed that agents’ outcomes are not causally affected by their features (and thus that strategic agents’ goal is deceiving the decision-maker), we join concurrent work in modeling agents’ outcomes as a function of their changeable attributes. As our main contribution, we provide efficient algorithms for learning decision rules that optimize three distinct decision-maker objectives in a realizable linear setting: accurately predicting agents’ post-gaming outcomes (prediction risk minimization), incentivizing agents to improve these outcomes (agent outcome maximization), and estimating the coefficients of the true underlying model (parameter estimation). Our algorithms circumvent a hardness result of Miller et al. (2019) by allowing the decision maker to test a sequence of decision rules and observe agents’ responses, in effect performing causal interventions through the decision rules.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-shavit20a, title = {Causal Strategic Linear Regression}, author = {Shavit, Yonadav and Edelman, Benjamin and Axelrod, Brian}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {8676--8686}, 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/shavit20a/shavit20a.pdf}, url = {https://proceedings.mlr.press/v119/shavit20a.html}, abstract = {In many predictive decision-making scenarios, such as credit scoring and academic testing, a decision-maker must construct a model that accounts for agents’ propensity to “game” the decision rule by changing their features so as to receive better decisions. Whereas the strategic classification literature has previously assumed that agents’ outcomes are not causally affected by their features (and thus that strategic agents’ goal is deceiving the decision-maker), we join concurrent work in modeling agents’ outcomes as a function of their changeable attributes. As our main contribution, we provide efficient algorithms for learning decision rules that optimize three distinct decision-maker objectives in a realizable linear setting: accurately predicting agents’ post-gaming outcomes (prediction risk minimization), incentivizing agents to improve these outcomes (agent outcome maximization), and estimating the coefficients of the true underlying model (parameter estimation). Our algorithms circumvent a hardness result of Miller et al. (2019) by allowing the decision maker to test a sequence of decision rules and observe agents’ responses, in effect performing causal interventions through the decision rules.} }
Endnote
%0 Conference Paper %T Causal Strategic Linear Regression %A Yonadav Shavit %A Benjamin Edelman %A Brian Axelrod %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-shavit20a %I PMLR %P 8676--8686 %U https://proceedings.mlr.press/v119/shavit20a.html %V 119 %X In many predictive decision-making scenarios, such as credit scoring and academic testing, a decision-maker must construct a model that accounts for agents’ propensity to “game” the decision rule by changing their features so as to receive better decisions. Whereas the strategic classification literature has previously assumed that agents’ outcomes are not causally affected by their features (and thus that strategic agents’ goal is deceiving the decision-maker), we join concurrent work in modeling agents’ outcomes as a function of their changeable attributes. As our main contribution, we provide efficient algorithms for learning decision rules that optimize three distinct decision-maker objectives in a realizable linear setting: accurately predicting agents’ post-gaming outcomes (prediction risk minimization), incentivizing agents to improve these outcomes (agent outcome maximization), and estimating the coefficients of the true underlying model (parameter estimation). Our algorithms circumvent a hardness result of Miller et al. (2019) by allowing the decision maker to test a sequence of decision rules and observe agents’ responses, in effect performing causal interventions through the decision rules.
APA
Shavit, Y., Edelman, B. & Axelrod, B.. (2020). Causal Strategic Linear Regression. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:8676-8686 Available from https://proceedings.mlr.press/v119/shavit20a.html.

Related Material