Dr Anjana Wijekoon a.wijekoon1@rgu.ac.uk
Research Fellow B
Dr Anjana Wijekoon a.wijekoon1@rgu.ac.uk
Research Fellow B
Professor Nirmalie Wiratunga n.wiratunga@rgu.ac.uk
Associate Dean for Research
Dr Ikechukwu Nkisi-Orji i.nkisi-orji@rgu.ac.uk
Chancellor's Fellow
Mr Chamath Palihawadana c.palihawadana@rgu.ac.uk
Research Assistant
Dr David Corsar d.corsar1@rgu.ac.uk
Senior Lecturer
Dr Kyle Martin k.martin3@rgu.ac.uk
Lecturer
Mark T. Keane
Editor
Professor Nirmalie Wiratunga n.wiratunga@rgu.ac.uk
Editor
Counterfactual explanations describe how an outcome can be changed to a more desirable one. In XAI, counterfactuals are "actionable" explanations that help users to understand how model decisions can be changed by adapting features of an input. A case-based approach to counterfactual discovery harnesses Nearest-unlike Neighbours as the basis to identify the minimal adaptations needed for outcome change. This paper presents the DisCERN algorithm which uses the query, its NUN and substitution-based adaptation operations to create a counterfactual explanation case. DisCERN uses feature attribution as adaptation knowledge to order substitutions operations and to bring about the desired outcome with as fewer changes as possible. We find our novel approach for Integrated Gradients using the NUN as the baseline against which the feature attributions are calculated outperforms other techniques like LIME and SHAP. DisCERN also uses feature attributions to bring the NUN closer by which the total change needed is further minimised, but the number of feature changes can increase. Overall, DisCERN outperforms other counterfactual algorithms such as DiCE and NICE in generating valid counterfactuals with fewer adaptations.
WIJEKOON, A., WIRATUNGA, N., NKISI-ORJI, I., PALIHAWADANA, C., CORSAR, D. and MARTIN, K. 2022. How close is too close? Role of feature attributions in discovering counterfactual explanations. In Keane, M.T. and Wiratunga, N. (eds.) Case-based reasoning research and development: proceedings of the 30th International conference on case-based reasoning (ICCBR 2022), 12-15 September 2022, Nancy, France. Lecture notes in computer science, 13405. Cham: Springer [online], pages 33-47. Available from: https://doi.org/10.1007/978-3-031-14923-8_3
Conference Name | 30th International conference on case-based reasoning (ICCBR 2022) |
---|---|
Conference Location | Nancy, France |
Start Date | Sep 12, 2022 |
End Date | Sep 15, 2022 |
Acceptance Date | May 30, 2022 |
Online Publication Date | Aug 14, 2022 |
Publication Date | Aug 31, 2022 |
Deposit Date | Jul 6, 2022 |
Publicly Available Date | Aug 15, 2023 |
Publisher | Springer |
Pages | 33-47 |
Series Title | Lecture notes in computer science |
Series Number | 13405 |
Series ISSN | 0302-9743 ; 1611-3349 |
Book Title | Case-based reasoning research and development: proceedings of the 30th International conference on case-based reasoning (ICCBR 2022), 12-15 September 2022, Nancy, France |
ISBN | 9783031149221 |
DOI | https://doi.org/10.1007/978-3-031-14923-8_3 |
Keywords | Counterfactuals (Computing); Explainable artificial intelligence (XAI); Nearest-unlike neighbours (NUN); Machine learning; Artificial intelligence; Counterfactual XAI; Feature attribution; Integrated gradients; Adaptation |
Public URL | https://rgu-repository.worktribe.com/output/1706142 |
WIJEKOON 2022 How close is too close (AAM)
(2.2 Mb)
PDF
Copyright Statement
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG.
iSee: intelligent sharing of explanation experience of users for users.
(2023)
Conference Proceeding
iSee: demonstration video. [video recording]
(2023)
Digital Artefact
Clinical dialogue transcription error correction using Seq2Seq models.
(2022)
Conference Proceeding
Clinical dialogue transcription error correction using Seq2Seq models.
(2022)
Working Paper
DisCERN: discovering counterfactual explanations using relevance features from neighbourhoods.
(2021)
Conference Proceeding
About OpenAIR@RGU
Administrator e-mail: publications@rgu.ac.uk
This application uses the following open-source libraries:
Apache License Version 2.0 (http://www.apache.org/licenses/)
Apache License Version 2.0 (http://www.apache.org/licenses/)
SIL OFL 1.1 (http://scripts.sil.org/OFL)
MIT License (http://opensource.org/licenses/mit-license.html)
CC BY 3.0 ( http://creativecommons.org/licenses/by/3.0/)
Advanced Search