Can I read Issues with Post-hoc Counterfactual Explanations: a Discussion on EtoBox?

Issues with Post-hoc Counterfactual Explanations: a Discussion by Laugel, Thibault; Lesot, Marie-Jeanne; Marsala, Christophe; Detyniecki, Marcin is a scholarly article available to read on EtoBox.

What is Issues with Post-hoc Counterfactual Explanations: a Discussion about?

Counterfactual post-hoc interpretability approaches have been proven to be useful tools to generate explanations for the predictions of a trained blackbox classifier. However, the assumptions they make about the data and the classifier make them unreliable in many contexts. In this paper, we discuss three desirable properties and approaches to quantify them: proximity, connectedness and stability. In addition, we illustrate that there is a risk for post-hoc counterfactual approaches to not satisfy these properties.

Author

Laugel, Thibault; Lesot, Marie-Jeanne; Marsala, Christophe; Detyniecki, Marcin

Published

2019

Language

EN