AI is not simply a technology that institutions will use. It is a technology that will reshape institutional authority — who exercises it, how, and with what accountability.
The Authority Shift
Artificial intelligence systems are being deployed in institutional decision-making contexts at a pace that is outrunning the governance frameworks designed to ensure those deployments serve the populations they affect. The AI hiring tool that filters job applications before a human recruiter sees them, the AI credit scoring model that determines loan eligibility before a human loan officer reviews the application, and the AI content moderation system that removes social media content before a human moderator reviews the decision are each examples of the authority shift that AI deployment is producing: the movement of consequential decision-making authority from human institutional actors — accountable for their decisions through the legal, professional, and social mechanisms that govern human behaviour — to algorithmic systems whose accountability is less direct, less transparent, and less well-governed.
The institutional authority question is not whether AI systems make better decisions than human ones — in many specific domains, they do. The institutional authority question is whether the authority that AI systems exercise is governed with the accountability that authority over people's lives requires. The AI system that makes better decisions on average while making worse decisions for specific demographic groups, in ways that a human decision-maker would face legal liability for, exercises institutional authority that is not adequately governed by the current accountability framework. The gap between the authority that AI systems exercise and the accountability framework that governs the exercise of that authority is the governance challenge that AI poses for institutions.
AI's challenge to institutional authority is not that it makes institutions less powerful — it makes them more powerful, by extending their decision-making capacity beyond what human staffing could achieve. Its governance challenge is that the extension of institutional power is not accompanied by the equivalent extension of institutional accountability. That gap is the governance problem that every AI deployment in a consequential institutional context must address.
Discussion