Responsible AI and Algorithmic Accountability in Public Administration: Governance Instruments, Oversight, and Democratic Legitimacy

Abstrak

This literature review examines how artificial intelligence and automated decision making are governed in public administration, with particular attention to the accountability instruments used to supervise these systems and their implications for democratic legitimacy. Using a structured qualitative review design, the study synthesises 15 peer reviewed journal articles selected for their relevance to public sector AI, automated administrative decision making, algorithmic accountability, and democratic governance. The analysis proceeds in three stages: identifying recurring governance instruments, mapping them across the lifecycle of public sector AI, and assessing their institutional and democratic implications. The review finds that the literature clusters accountability instruments into normative, procedural, technical, and institutional forms, with the most frequently discussed mechanisms being transparency, explainability, audit, impact assessment, human oversight, and reviewability. However, the evidence also shows that the field remains unevenly developed, with stronger attention to ex ante safeguards and operational oversight than to complaint handling, appeals, and administrative correction. Substantively, the reviewed studies indicate that these instruments do not automatically enhance fairness, due process, administrative justice, or public trust. Their democratic value is conditional on legal design, organisational capacity, and meaningful review mechanisms. The study concludes that algorithmic accountability in public administration should be understood as an integrated institutional architecture rather than a checklist of abstract principles.