On the Hardness of Computing Counterfactual and Semifactual Explanations in XAI

TL;DR

This paper analyzes the computational difficulty of generating counterfactual and semi-factual explanations in XAI, revealing that producing these explanations is often computationally hard and hard to approximate, impacting AI transparency and regulation.

Contribution

It provides new inapproximability results demonstrating the inherent computational hardness of generating and approximating explanations in XAI.

Findings

Generating explanations is often computationally hard.

Explanations are hard to approximate under certain assumptions.

Implications affect AI transparency and policy regulation.

Abstract

Providing clear explanations to the choices of machine learning models is essential for these models to be deployed in crucial applications. Counterfactual and semi-factual explanations have emerged as two mechanisms for providing users with insights into the outputs of their models. We provide an overview of the computational complexity results in the literature for generating these explanations, finding that in many cases, generating explanations is computationally hard. We strengthen the argument for this considerably by further contributing our own inapproximability results showing that not only are explanations often hard to generate, but under certain assumptions, they are also hard to approximate. We discuss the implications of these complexity results for the XAI community and for policymakers seeking to regulate explanations in AI.