Abstract
Artificial intelligence (AI) is increasingly used in public administration, yet its evaluation often remains centered on performance, efficiency, transparency, or explainability. This chapter argues that such approaches are insufficient for legally consequential administrative action because they do not adequately address whether AI-assisted decisions remain lawfully justified, intelligible, and open to review. To address this gap, the chapter develops the Judicially Informed Decision Framework, a five-step framework for evaluating AI-assisted public decision-making through the lens of administrative justification. The framework asks whether an administrative process can adequately define the case, identify the correct legal basis, reason through the case, write reasons that faithfully reflect the grounds of decision, and preserve meaningful opportunities for review and challenge. The chapter argues that AI in public administration should be treated as a bounded support technology rather than as a substitute for administrative reasoning.

This publication has 24 references indexed in Scilit: