A general approach to compute the relevance of middle-level input features
This addresses the interpretive burden in XAI for users, though it appears incremental as it builds on existing middle-level explanation concepts.
The paper tackles the lack of a general method to evaluate middle-level explanations in XAI, proposing a novel framework to compute the relevance of these features for ML model behavior.
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.