Local Explanations via Necessity and Sufficiency: Unifying Theory and Practice

TL;DR

This paper unifies the concepts of necessity and sufficiency within a formal framework for explainable AI, providing a new algorithm for computing explanations that outperforms existing methods.

Contribution

It develops a comprehensive theoretical foundation for necessity and sufficiency in XAI and introduces a novel, sound, and complete algorithm for explanation computation.

Findings

Algorithm demonstrates competitive performance on various tasks.

Unified framework clarifies the role of necessity and sufficiency in explanations.

Provides a solid theoretical basis for future XAI research.

Abstract

Necessity and sufficiency are the building blocks of all successful explanations. Yet despite their importance, these notions have been conceptually underdeveloped and inconsistently applied in explainable artificial intelligence (XAI), a fast-growing research area that is so far lacking in firm theoretical foundations. Building on work in logic, probability, and causality, we establish the central role of necessity and sufficiency in XAI, unifying seemingly disparate methods in a single formal framework. We provide a sound and complete algorithm for computing explanatory factors with respect to a given context, and demonstrate its flexibility and competitive performance against state of the art alternatives on various tasks.