Polycentric Governance (Holling Reading)
Polycentric governance is the institutional realization of adaptive management at scale. Rather than concentrating authority in a single hierarchical body, polycentric systems distribute governance across multiple overlapping institutions operating at different scales — local management bodies, regional coordination mechanisms, national regulatory agencies, international frameworks — connected through information flows and coordinating mechanisms. The architecture matches the multi-scale character of the systems being governed and provides the redundancy, diversity, and cross-scale learning that adaptive capacity requires. AI governance needs polycentric arrangements analogous to those that have emerged around the Great Barrier Reef, the Ostrom common-pool resource literature, and other successful multi-scale governance cases.
In the AI Story
The contrast is with monocentric governance — the model of a single authoritative regulator issuing uniform rules. Monocentric governance is effective under conservation-phase conditions where the system being governed is stable enough that a single optimized framework can outperform a portfolio of approaches. Under release-phase conditions, monocentric governance fails for the same reasons optimization fails: it cannot adapt fast enough, it cannot accommodate diverse conditions, it cannot learn from the experimentation that reorganization requires.
Polycentric arrangements address these failures by design. Local management bodies can respond to conditions that national agencies cannot see. Regional coordination prevents local optimization from producing system-level pathology. National frameworks provide stability and normative coherence. International frameworks address cross-border spillovers. Each level operates at its appropriate scale; information flows between levels enable learning; the architecture as a whole is capable of responses no single level could produce.
For AI governance, this means local experiments in workforce transition, regional educational innovation, national regulatory frameworks, and international coordination mechanisms — connected through channels that propagate learning and prevent pathological capture of governance by any single scale's interests. The current AI governance landscape is dominated by national regulatory efforts (the EU AI Act, American executive orders) operating largely in isolation. The polycentric architecture adequate to the transition does not yet exist; building it is among the most consequential reorganization-phase tasks.
The approach has been developed most thoroughly in the common-pool resource literature associated with Elinor Ostrom's Nobel-winning work, which documented dozens of cases where polycentric arrangements governed complex resources successfully where monocentric arrangements had failed. The translation of these findings into AI governance is ongoing and incomplete.
Key Ideas
Multi-scale architecture matches multi-scale systems. Governance scales should mirror the scales of the phenomena being governed.
Redundancy is a feature. Overlapping authorities provide buffer against failure at any single node.
Learning flows through the architecture. Information and experimentation propagate across scales through coordinating mechanisms.
Monocentric governance fails in release-phase conditions. Single-point optimization cannot adapt fast enough.
AI governance lacks polycentric architecture. Building it is a core reorganization-phase task.
Appears in the Orange Pill Cycle
Further reading
- Ostrom, E. Governing the Commons (Cambridge, 1990)
- Ostrom, V., Tiebout, C.M., Warren, R. The Organization of Government in Metropolitan Areas (American Political Science Review, 1961)
- Folke, C. et al. Adaptive Governance of Social-Ecological Systems (Annual Review of Environment and Resources, 2005)