Evaluating and Correcting Performative Effects of Decision Support Systems via Causal Domain Shift

Philip Boeken, Onno Zoeter, Joris Mooij
Proceedings of the Third Conference on Causal Learning and Reasoning, PMLR 236:551-569, 2024.

Abstract

When predicting a target variable $Y$ from features $X$, the prediction $\hat{Y}$ can be performative: an agent might act on this prediction, affecting the value of $Y$ that we eventually observe. Performative predictions are deliberately prevalent in algorithmic decision support, where a decision support system (DSS) provides a prediction for an agent to affect the value of the target variable. When deploying a DSS in high-stakes settings (e.g. healthcare, law, predictive policing, or child welfare screening) it is imperative to carefully assess the performative effects of the DSS. In the case that the DSS serves as an alarm for a predicted negative outcome, naive retraining of the prediction model is bound to result in a model that underestimates the risk, due to effective workings of the previous model. In this work, we propose to model the deployment of a DSS as causal domain shift and provide novel cross-domain identification results for the conditional expectation $E[Y|X]$, allowing for pre- and post-hoc assessment of the deployment of the DSS, and for re-training of a model that assesses the risk under a baseline policy where the DSS is not deployed. Using a running example, we empirically show that a repeated regression procedure provides a practical framework for estimating these quantities, even when the data is affected by sample selection bias and selective labelling, offering for a practical, unified solution for multiple forms of target variable bias.

Cite this Paper

BibTeX
@InProceedings{pmlr-v236-boeken24a, title = {Evaluating and Correcting Performative Effects of Decision Support Systems via Causal Domain Shift}, author = {Boeken, Philip and Zoeter, Onno and Mooij, Joris}, booktitle = {Proceedings of the Third Conference on Causal Learning and Reasoning}, pages = {551--569}, 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/boeken24a/boeken24a.pdf}, url = {https://proceedings.mlr.press/v236/boeken24a.html}, abstract = {When predicting a target variable $Y$ from features $X$, the prediction $\hat{Y}$ can be performative: an agent might act on this prediction, affecting the value of $Y$ that we eventually observe. Performative predictions are deliberately prevalent in algorithmic decision support, where a decision support system (DSS) provides a prediction for an agent to affect the value of the target variable. When deploying a DSS in high-stakes settings (e.g. healthcare, law, predictive policing, or child welfare screening) it is imperative to carefully assess the performative effects of the DSS. In the case that the DSS serves as an alarm for a predicted negative outcome, naive retraining of the prediction model is bound to result in a model that underestimates the risk, due to effective workings of the previous model. In this work, we propose to model the deployment of a DSS as causal domain shift and provide novel cross-domain identification results for the conditional expectation $E[Y|X]$, allowing for pre- and post-hoc assessment of the deployment of the DSS, and for re-training of a model that assesses the risk under a baseline policy where the DSS is not deployed. Using a running example, we empirically show that a repeated regression procedure provides a practical framework for estimating these quantities, even when the data is affected by sample selection bias and selective labelling, offering for a practical, unified solution for multiple forms of target variable bias.} }
Endnote
%0 Conference Paper %T Evaluating and Correcting Performative Effects of Decision Support Systems via Causal Domain Shift %A Philip Boeken %A Onno Zoeter %A Joris Mooij %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-boeken24a %I PMLR %P 551--569 %U https://proceedings.mlr.press/v236/boeken24a.html %V 236 %X When predicting a target variable $Y$ from features $X$, the prediction $\hat{Y}$ can be performative: an agent might act on this prediction, affecting the value of $Y$ that we eventually observe. Performative predictions are deliberately prevalent in algorithmic decision support, where a decision support system (DSS) provides a prediction for an agent to affect the value of the target variable. When deploying a DSS in high-stakes settings (e.g. healthcare, law, predictive policing, or child welfare screening) it is imperative to carefully assess the performative effects of the DSS. In the case that the DSS serves as an alarm for a predicted negative outcome, naive retraining of the prediction model is bound to result in a model that underestimates the risk, due to effective workings of the previous model. In this work, we propose to model the deployment of a DSS as causal domain shift and provide novel cross-domain identification results for the conditional expectation $E[Y|X]$, allowing for pre- and post-hoc assessment of the deployment of the DSS, and for re-training of a model that assesses the risk under a baseline policy where the DSS is not deployed. Using a running example, we empirically show that a repeated regression procedure provides a practical framework for estimating these quantities, even when the data is affected by sample selection bias and selective labelling, offering for a practical, unified solution for multiple forms of target variable bias.
APA
Boeken, P., Zoeter, O. & Mooij, J.. (2024). Evaluating and Correcting Performative Effects of Decision Support Systems via Causal Domain Shift. Proceedings of the Third Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 236:551-569 Available from https://proceedings.mlr.press/v236/boeken24a.html.

Related Material