Trustworthy AI in the Public Sector: Opportunities, Risks, and Recommendations

Abstract

As shown in this pandemic, the digitalization in (local) governments is a severe problem that affects the lives and health of citizens. Artificial intelligences (AI) are applied to various cases, from police activities to nursery matching. Even if it is technically possible to use AI, in some cases, we should not use it from the viewpoint of normativity. However, few studies have focused on this gap in public sectors. The purpose of this study was to assess several use cases from the perspective of discretionary control in administration such as elimination of arbitrariness. It clarified some points related data bias, algorithmic auditing, and objection procedures.