ExplainabilityTransparencyAccountabilitycs.AI
A Few Good Counterfactuals: Generating Interpretable, Plausible and Diverse Counterfactual Explanations
Barry Smyth, Mark T Keane
January 22, 2021
International Conference on Case-Based Reasoning
37 citations
4 influential
Key Takeaways
-
Counterfactual explanations help people understand AI decisions by showing what would need to change for a different outcome — for example, explaining to a loan applicant "if your credit score were 50 points higher, you would have been approved" — but the paper finds that naturally occurring examples of these explanations are rare in real datasets.
-
Current AI explanation methods often create unrealistic scenarios by modifying features in ways that don't actually occur in the real world, such as suggesting impossible combinations of characteristics; this paper's approach instead uses real feature combinations from existing data to generate more plausible explanations.
-
The method prioritizes sparse explanations (changing fewer features) and diversity (generating multiple different explanations), which makes AI decisions easier for non-experts to understand and act upon, rather than overwhelming users with complex or repetitive explanations.
-
By adapting real examples from datasets rather than generating entirely synthetic ones, the approach produces counterfactuals that represent genuinely achievable scenarios, making the explanations more trustworthy and actionable for people affected by AI decisions.
Abstract
Counterfactual explanations provide a potentially significant solution to the Explainable AI (XAI) problem, but good, native counterfactuals have been shown to rarely occur in most datasets. Hence, the most popular methods generate synthetic counterfactuals using blind perturbation. However, such methods have several shortcomings: the resulting counterfactuals (i) may not be valid data-points (they often use features that do not naturally occur), (ii) may lack the sparsity of good counterfactuals (if they modify too many features), and (iii) may lack diversity (if the generated counterfactuals are minimal variants of one another). We describe a method designed to overcome these problems, one that adapts native counterfactuals in the original dataset, to generate sparse, diverse synthetic counterfactuals from naturally occurring features. A series of experiments are reported that systematically explore parametric variations of this novel method on common datasets to establish the conditions for optimal performance.