ABSTRACT

Artificial Intelligence (AI) applications fundamentally impact individuals, the environment, and society as a whole. Many applications of this transformative technology pose ethical challenges for developers, policy-makers, and society. Ethics and AI meet in various ways. In this talk, we identify three central pillars: i) novel ethical concerns specific to AI; ii) AI-tailored policy-making; and iii) implementation of formal reasoning with norms and values. Particular emphasis is put on the third, due to the increasing awareness that AI systems must be aligned with human values, ethics, and laws. Besides determining what these values must be, the immediate question here is, how do values and norms influence formal reasoning processes? One of the key challenges for such reasoning is the fact that norms and values are highly conflict-sensitive. Early formal studies of normative reasoning (often umbrellaed under the name “deontic logic”) emerged in the 1950s, whereas Knowledge Representation approaches in the 2000s led to the introduction of defeasible normative reasoning systems adopting conflict resolution mechanisms. Whereas traditional methods in the field have focused on showing that certain obligations/rights/permissions follow from a given normative knowledge base, developments in AI shifted focus to explainability, additionally requiring proof of why certain conclusions follow. Concerning the latter, recent work will be discussed that leverages computational argumentation and dialogue models to provide explanatory normative reasoning, also identifying several key future research challenges.