A general approach to compute the relevance of middle-level input features

TL;DR

This paper introduces a new general framework within XAI to evaluate the relevance of middle-level features in explaining ML model behavior, addressing interpretability challenges in complex models.

Contribution

It proposes the first comprehensive method to assess middle-level feature relevance for ML explanations, improving interpretability over low-level approaches.

Findings

Framework effectively evaluates middle-level feature relevance

Enhances interpretability of ML models in XAI

Addresses limitations of low-level explanations

Abstract

This work proposes a novel general framework, in the context of eXplainable Artificial Intelligence (XAI), to construct explanations for the behaviour of Machine Learning (ML) models in terms of middle-level features. One can isolate two different ways to provide explanations in the context of XAI: low and middle-level explanations. Middle-level explanations have been introduced for alleviating some deficiencies of low-level explanations such as, in the context of image classification, the fact that human users are left with a significant interpretive burden: starting from low-level explanations, one has to identify properties of the overall input that are perceptually salient for the human visual system. However, a general approach to correctly evaluate the elements of middle-level explanations with respect ML model responses has never been proposed in the literature.