By Mael Pegny (Université de Paris 1)

Abstract

Citizens have a right to an explanation of the decisions affecting them. However, if AIs are to be used in decision procedures, how can we explain complex AI systems to the general public, while so many computer scientists find them inscrutably opaque? In this presentation, I will try to present an optimistic approach to this seemingly hopeless question, which I call input-output black box reasoning. In a nutshell, this is the idea that simple explanations  can be produced for practical purposes by discarding  the inner workings of the system and focusing on the relations between inputs and outputs. While I defend this is a good methodological starting point, I will also try to define some of the tough challenges ahead.

Tuesday 4 June 2019 @11.15 room BC 420 (see map)