Contrastive Explanation Method (CEM)

• Values: { explainability }
  • Explanation type: { contrastive }
• Categories: { model-agnostic }
• Stage: { post-processing }
• Repository: https://github.com/IBM/Contrastive-Explanation-Method
• Tasks: { classification }
• Input data: { image }
• Licence: Apache License 2.0
• Languages: { Python }
• References:
  • Dhurandhar et al. - Explanations based on the Missing: Towards Contrastive Explanations with Pertinent Negatives

Dhurandhar et al. support a type of contrastive explanation based on what they call pertinent negatives. A contrastive explanation answers the question: “Why P, rather than Q”?

CEM supports such an explanation by finding the minimal set of features that lead to prediction P (a pertinent positive that resembles an anchor explanation), and additionally a minimal set of features that should be absent to maintain decision P instead of the decision for closest class Q (a pertinent negative that is somewhat similar to a counterfactual ).

CEM is also implemented in the more encompassing Alibi .