AI and bureaucratic discretion

Abstract

Algorithmic decision-making has the potential to radically reshape policy-making and policy implementation. Many of the moral examinations of AI in government take AI to be a neutral epistemic tool or the value-driven analogue of a policymaker. In this paper, I argue that AI systems in public administration are often better analogised to a street-level bureaucrat. Doing so opens up a host of questions about the moral dispositions of such AI systems. I argue that AI systems in public administration often act as indifferent bureaucrats, and that this can introduce a problematic homogeneity in the moral dispositions in administrative agencies.