Legality, Legitimacy, and Instrumental Possibility in Human and Computational Governance for the Public Sector

Artificial Intelligence (AI) can have a significant beneficial or harmful impact when used in public policies and services. This paper provides a reflection on the EU regulatory initiatives towards regulating governance and risk management of AI, elaborating on the entrenchment of the dimensions of legality, legitimacy and instrumental possibility. We argue that only against this conceptual backdrop we can possibly unpack what a legal, legitimate, trustworthy, fair and accountable approach to designing and implementing AI in public service is. As an outcome, we identify a few crucial architectural requirements, both at normative and at computational level.