Towards a Generic Framework for Black-box Explanations of Algorithmic Decision Systems

Clement Henin (1, 2) , Daniel Le Métayer (2)

1 ENPC - École des Ponts ParisTech
2 PRIVATICS - Privacy Models, Architectures and Tools for the Information Society

The main goal of this paper is to define a generic framework for black-box explanation methods in order to make it easier to compare and classify different approaches. We focus on two components of this framework, called respectively “Sampling” and “Generation”, which are characterized formally and used to build a taxonomy of explanation methods. This document gives details on how this framework can be used to describe black-box explanation methods found in the literature.

Mots clés: Automatic decision system; black-box model; algorithm transparency; explanability; machine-learning; artificial intelligence

Domaines: Informatique [cs]; Intelligence artificielle [cs.AI]

Clement Henin, Daniel Le Métayer. Towards a Generic Framework for Black-box Explanations of Algorithmic Decision Systems. [Research Report] RR-9276, Inria Grenoble Rhône-Alpes; Ecole des Ponts ParisTech. 2019. ⟨hal-02131174v1⟩