Enhancing textual counterfactual explanation intelligibility through Counterfactual Feature Importance - Archive ouverte HAL Access content directly
Conference Papers Year : 2023

Enhancing textual counterfactual explanation intelligibility through Counterfactual Feature Importance

Milan Bhan
  • Function : Author
  • PersonId : 1317469
Nicolas Chesneau
  • Function : Author
  • PersonId : 1317470

Abstract

Textual counterfactual examples explain a prediction by modifying the tokens of an initial instance in order to flip the outcome of a classifier. Even under sparsity constraint, counterfactual generation can lead to numerous changes from the initial text, making the explanation hard to understand. We propose Counterfactual Feature Importance, a method to make non-sparse counterfactual explanations more intelligible. Counterfactual Feature Importance assesses token change importance between an instance to explain and its counterfactual example. We develop two ways of computing Counterfactual Feature Importance, respectively based on classifier gradient computation and counterfactual generator loss evolution during counterfactual search. Then we design a global version of Counterfactual Feature Importance, providing rich information about semantic fields globally impacting classifier predictions. Counterfactual Feature Importance enables to focus on impacting parts of counterfactual explanations, making counterfactual explanations involving numerous changes more understandable.
Fichier principal
Vignette du fichier
CFI___TrustNLP_ACL_2023 (1).pdf (861.71 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-04311780 , version 1 (11-12-2023)

Identifiers

Cite

Milan Bhan, Jean-Noël Vittaut, Nicolas Chesneau, Marie-Jeanne Lesot. Enhancing textual counterfactual explanation intelligibility through Counterfactual Feature Importance. 3rd Workshop on Trustworthy Natural Language Processing (TrustNLP 2023), Jul 2023, Toronto, Canada. pp.221-231, ⟨10.18653/v1/2023.trustnlp-1.19⟩. ⟨hal-04311780⟩
25 View
16 Download

Altmetric

Share

Gmail Facebook X LinkedIn More