A Conceptual View on the Design and Properties of Explainable AI Systems for Legal Settings

Martijn van Otterlo, Martin Atzmueller

Abstract:

The recent advances in AI have broad implications, e. g., onto explainability,
accountability, and responsibility – in particular in legal settings. In this
paper, we provide a conceptual view on design properties of such respective AI
systems, where we focus on two specific approaches to explainable artificial intelligence
(AI) for legal settings. The first approach aims at designing explainable
AI using legal requirements – in a research and design methodology; the second
one deals with a design strategy for ensuring requirements of computational ethical
reasoning systems in and for which explainability is a core element.