KR2023Proceedings of the 20th International Conference on Principles of Knowledge Representation and ReasoningProceedings of the 20th International Conference on Principles of Knowledge Representation and Reasoning

Rhodes, Greece. September 2-8, 2023.

Edited by
  1. Pierre Marquis,
  2. Tran Cao Son
  3. Gabriele Kern-Isberner

ISSN: 2334-1033
ISBN: 978-1-956792-02-7

Sponsored by
Published by

Copyright © 2023 International Joint Conferences on Artificial Intelligence Organization

Counterfactual Explanations and Model Multiplicity: a Relational Verification View

  1. Francesco Leofante(Imperial College London)
  2. Elena Botoeva(University of Kent)
  3. Vineet Rajani(University of Kent)
PDF
BibTeX
https://doi.org/10.24963/kr.2023/78

Keywords

  1. Explainable AI
  2. KR and machine learning, inductive logic programming, knowledge acquisition

Abstract

We study the interplay between counterfactual explanations and model multiplicity in the context of neural network classifiers. We show that current explanation methods often produce counterfactuals whose validity is not preserved under model multiplicity. We then study the problem of generating counterfactuals that are guaranteed to be robust to model multiplicity, characterise its complexity and propose an approach to solve this problem using ideas from relational verification.