Exploring Counterfactual Explanations Through the Lens of Adversarial Examples: A Theoretical and Empirical Analysis

Martin Pawelczyk, Chirag Agarwal, Shalmali Joshi, Sohini Upadhyay, Himabindu Lakkaraju
Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, PMLR 151:4574-4594, 2022.

Abstract

As machine learning (ML) models becomemore widely deployed in high-stakes applications, counterfactual explanations have emerged as key tools for providing actionable model explanations in practice. Despite the growing popularity of counterfactual explanations, the theoretical understanding of these explanations is still lacking behind. In this work, we systematically analyze counterfactual explanations through the lens of adversarial examples. We do so by formalizing the similarities between popular counterfactual explanation and adversarial example generation methods identifying conditions when they are equivalent. We then derive upper bounds between the solutions output by counterfactual explanation and adversarial example generation methods, which we validate on several real world data sets. By establishing these theoretical and empirical similarities between counterfactual explanations and adversarial examples, our work raises fundamental questions about the design and development of existing counterfactual explanation algorithms.

Cite this Paper


BibTeX
@InProceedings{pmlr-v151-pawelczyk22a, title = { Exploring Counterfactual Explanations Through the Lens of Adversarial Examples: A Theoretical and Empirical Analysis }, author = {Pawelczyk, Martin and Agarwal, Chirag and Joshi, Shalmali and Upadhyay, Sohini and Lakkaraju, Himabindu}, booktitle = {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics}, pages = {4574--4594}, year = {2022}, editor = {Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel}, volume = {151}, series = {Proceedings of Machine Learning Research}, month = {28--30 Mar}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v151/pawelczyk22a/pawelczyk22a.pdf}, url = {https://proceedings.mlr.press/v151/pawelczyk22a.html}, abstract = { As machine learning (ML) models becomemore widely deployed in high-stakes applications, counterfactual explanations have emerged as key tools for providing actionable model explanations in practice. Despite the growing popularity of counterfactual explanations, the theoretical understanding of these explanations is still lacking behind. In this work, we systematically analyze counterfactual explanations through the lens of adversarial examples. We do so by formalizing the similarities between popular counterfactual explanation and adversarial example generation methods identifying conditions when they are equivalent. We then derive upper bounds between the solutions output by counterfactual explanation and adversarial example generation methods, which we validate on several real world data sets. By establishing these theoretical and empirical similarities between counterfactual explanations and adversarial examples, our work raises fundamental questions about the design and development of existing counterfactual explanation algorithms. } }
Endnote
%0 Conference Paper %T Exploring Counterfactual Explanations Through the Lens of Adversarial Examples: A Theoretical and Empirical Analysis %A Martin Pawelczyk %A Chirag Agarwal %A Shalmali Joshi %A Sohini Upadhyay %A Himabindu Lakkaraju %B Proceedings of The 25th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2022 %E Gustau Camps-Valls %E Francisco J. R. Ruiz %E Isabel Valera %F pmlr-v151-pawelczyk22a %I PMLR %P 4574--4594 %U https://proceedings.mlr.press/v151/pawelczyk22a.html %V 151 %X As machine learning (ML) models becomemore widely deployed in high-stakes applications, counterfactual explanations have emerged as key tools for providing actionable model explanations in practice. Despite the growing popularity of counterfactual explanations, the theoretical understanding of these explanations is still lacking behind. In this work, we systematically analyze counterfactual explanations through the lens of adversarial examples. We do so by formalizing the similarities between popular counterfactual explanation and adversarial example generation methods identifying conditions when they are equivalent. We then derive upper bounds between the solutions output by counterfactual explanation and adversarial example generation methods, which we validate on several real world data sets. By establishing these theoretical and empirical similarities between counterfactual explanations and adversarial examples, our work raises fundamental questions about the design and development of existing counterfactual explanation algorithms.
APA
Pawelczyk, M., Agarwal, C., Joshi, S., Upadhyay, S. & Lakkaraju, H.. (2022). Exploring Counterfactual Explanations Through the Lens of Adversarial Examples: A Theoretical and Empirical Analysis . Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 151:4574-4594 Available from https://proceedings.mlr.press/v151/pawelczyk22a.html.

Related Material