A Framework for Feasible Counterfactual Exploration incorporating Causality, Sparsity and Density

TL;DR

This paper proposes a framework that generates feasible counterfactual explanations for machine learning models by preserving causal relations, using VAEs and manifold extraction to ensure relevance and usefulness in real-world scenarios.

Contribution

It introduces a method combining causal constraints, sparsity, and density to produce feasible counterfactuals, validated on benchmark datasets with a novel manifold-based visualization.

Findings

Feasible counterfactuals can be generated while maintaining causal relations.

The approach effectively distinguishes feasible from infeasible examples.

Generated counterfactuals are sparse, causal, and useful for end-users.

Abstract

The imminent need to interpret the output of a Machine Learning model with counterfactual (CF) explanations - via small perturbations to the input - has been notable in the research community. Although the variety of CF examples is important, the aspect of them being feasible at the same time, does not necessarily apply in their entirety. This work uses different benchmark datasets to examine through the preservation of the logical causal relations of their attributes, whether CF examples can be generated after a small amount of changes to the original input, be feasible and actually useful to the end-user in a real-world case. To achieve this, we used a black box model as a classifier, to distinguish the desired from the input class and a Variational Autoencoder (VAE) to generate feasible CF examples. As an extension, we also extracted two-dimensional manifolds (one for each dataset) that located the majority of the feasible examples, a representation that adequately distinguished them from infeasible ones. For our experimentation we used three commonly used datasets and we managed to generate feasible and at the same time sparse, CF examples that satisfy all possible predefined causal constraints, by confirming their importance with the attributes in a dataset.