Artifcial intelligence (AI) offers challenges and benefts to the public sector. We present an ethical framework to analyze the effects of AI in public organizations, guide empirical and theoretical research in public administration, and inform practitioner deliberation and decision making on AI adoption. We put forward six propositions on how the use of AI by public organizations may facilitate or prevent unnecessary harm. The framework builds on the theory of administrative evil and contributes to it in two ways. First, we interpret the theory of administrative evil through the lens of agency theory. We examine how the mechanisms stipulated by the former relate to the underlying mechanisms of the latter. Specifcally, we highlight how mechanisms of administrative evil can be analyzed as information problems in the form of adverse selection and moral hazard. Second, we describe possible causal pathways of the theory of administrative evil and associate each with a level of analysis: individual (micro), organizational (meso), and cultural (macro). We then develop both descriptive and normative propositions on AI's potential to increase or decrease the risk of administrative evil. The article hence contributes an institutional and public administration lens to the growing literature on AI safety and value alignment.
ASJC Scopus subject areas
- Public Administration