Accountability Models for Inclusive AI Governance in Public Institutions

Abstract

AI's growing presence in government agencies has opened up opportunities for enhancing their service delivery, accelerating administrative processes, and data-driven decision-making. While the incorporation of AI into public governance has significant potential, it does pose key equity, justice, transparency, public trust, and accountability issues as well. This study explores participatory governance as a practical tool for formulating accountability designs which will help keep AI systems responsive to citizens, particularly marginalised, under-represented groups within public institutions. It examines the concept of public accountability in decision-making processes where AI plays a role, and how stakeholder involvement, institutional monitoring, transparent decision-making, ethical review systems and feedback-driven governance can contribute to this. The findings reveal that participatory approaches foster greater levels of trust, perceived fairness and enabling institutional processes. The results also indicate that merely boosting technical audits or legal frameworks is insufficient to ensure accountability in public-sector AI—a need to involve citizens on an ongoing basis, as well as consult with multiple stakeholders on a frequent basis, and establish clear responsibility throughout the entire process of AI design, deployment and evaluation. By drawing on the study, inclusive AI governance models aim to minimize algorithmic bias, uphold citizens' rights and strengthen the legitimacy of governance-making processes by utilizing democratic elements and transparency regarding technical aspects of AI.