Actionable feature discovery in counterfactuals using feature relevance explainers.

Wiratunga, Nirmalie; Wijekoon, Anjana; Nkisi-Orji, Ikechukwu; Martin, Kyle; Palihawadana, Chamath; Corsar, David

Abstract

Counterfactual explanations focus on 'actionable knowledge' to help end-users understand how a Machine Learning model outcome could be changed to a more desirable outcome. For this purpose a counterfactual explainer needs to be able to reason with similarity knowledge in order to discover input dependencies that relate to outcome changes. Identifying the minimum subset of feature changes to action a change in the decision is an interesting challenge for counterfactual explainers. In this paper we show how feature relevance based explainers (i.e. LIME, SHAP), can inform a counterfactual explainer to identify the minimum subset of 'actionable features'. We demonstrate our DisCERN (Discovering Counterfactual Explanations using Relevance Features from Neighbourhoods) algorithm on three datasets and compare against the widely used counterfactual approach DiCE. Our preliminary results show that DisCERN to be a viable strategy that should be adopted to minimise the actionable changes.

Citation

WIRATUNGA, N., WIJEKOON, A., NKISI-ORJI, I., MARTIN, K., PALIHAWADANA, C. and CORSAR, D. 2021. Actionable feature discovery in counterfactuals using feature relevance explainers. In Borck, H., Eisenstadt, V., Sánchez-Ruiz, A. and Floyd, M. (eds.) Workshop proceedings of the 29th International conference on case-based reasoning (ICCBR-WS 2021), 13-16 September 2021, [virtual event]. CEUR workshop proceedings, 3017. Aachen: CEUR-WS [online], pages 63-74. Available from: http://ceur-ws.org/Vol-3017/101.pdf

• Presentation Conference Type: Conference Paper (published)
• Conference Name: Workshops of the 29th International conference on case-based reasoning (ICCBR-WS 2021)
• Start Date: Sep 13, 2021
• End Date: Sep 16, 2021
• Acceptance Date: Jun 11, 2021
• Online Publication Date: Sep 16, 2021
• Publication Date: Nov 24, 2021
• Deposit Date: Jan 6, 2022
• Publicly Available Date: Jan 6, 2022
• Publisher: CEUR-WS
• Peer Reviewed: Peer Reviewed
• Pages: 63-74
• Series Title: CEUR workshop proceedings
• Series Number: 3017
• Series ISSN: 1613-0073
• Keywords: Explainable AI; Counterfactual; Feature relevance; Actionable features
• Public URL: https://rgu-repository.worktribe.com/output/1563535
• Publisher URL: http://ceur-ws.org/Vol-3017/101.pdf