Reciprocal public AI governance: public AI dividend infrastructure (PAIDI) for public return, data agency and redress

Abstract

This paper aims to address a reciprocity gap in public artificial intelligence (AI) governance: existing frameworks increasingly classify, document and audit AI systems but say less about how governments can make AI-enabled transformation visibly reciprocal through public return, low-burden citizen agency and credible redress. The study develops a conceptual framework through problem-driven theory synthesis. It uses purposive conceptual sampling across AI risk governance, legitimacy, procedural justice, administrative burden, data governance, digital public infrastructure, market concentration and redress and explains how these literatures are combined. This paper theorises public AI dividend infrastructure (PAIDI) as a meso-level architecture linking three layers: public return, citizen agency and assurance/redress. It derives these layers from distributive legibility, actionable agency and remedial credibility deficits. This paper is conceptual and offers propositions for empirical and comparative testing. The framework identifies policy levers in registries, procurement, interoperability, reporting, assisted agency and claims administration. The contribution is architectural rather than component-wise. PAIDI differs from risk-based AI governance, digital public infrastructure, data-intermediary and redress models by making one covered-system architecture link public-return reporting, low-burden citizen agency and pre-funded remedy. The added value lies in the coupling of familiar instruments through registries, procurement duties, citizen-facing interfaces, incident records and assurance mechanisms.