Algorithmic Shura: Consultation, Consent, and the Advice–Decision Distinction in Public-Sector AI

Abstract

When artificial intelligence (AI) systems rank, flag, and recommend inside public administration, the philosophically interesting question is not only whether their outputs are accurate but where advice ends and authoritative decision begins. Existing accounts, including the threat of “algocracy”, the responsibility gap, and meaningful human control, diagnose how machine outputs can erode the conditions of legitimate public authority. This article contributes a conceptual resource from outside the usual Euro-American canon: the Islamic principle of consultation ( shura ). Read alongside consent-based theories of legitimacy, shura yields what we call algorithmic shura : a set of three jointly necessary conditions for government AI that informs significant decisions. First, meaningful consultation among accountable humans must precede deployment. Second, machine output must be institutionally treated as counsel rather than command, with an advice–decision wall that makes refusal no costlier than acceptance. Third, authorisation must be renewable rather than fossilised at launch. We reconstruct the framework, distinguish it from claims that machines can be consultative partners, and test it against AI governance instruments in the United Arab Emirates and neighbouring Gulf states, which already articulate fragments of the view. We argue that algorithmic shura sharpens familiar human-oversight principles by locating the core problem in the political relation between counsel and authority, and that it offers a model for cross-traditional work in the philosophy of technology.