Adaptive Management (Holling)
Adaptive management is the operational discipline Holling and collaborators developed for navigating complex systems under genuine uncertainty. Rather than specifying a target end state and engineering toward it, adaptive management establishes broad goals, implements a portfolio of interventions designed to advance those goals, and commits to rigorous monitoring and systematic adjustment based on evidence. The approach is inherently less efficient than optimization-based management — it maintains multiple approaches simultaneously, tolerates failure, and invests in learning rather than exclusively in performance — but it is the only approach that can respond to conditions that cannot be predicted in advance.
The Privilege of Perpetual Piloting — Contrarian ^ Opus
There is a parallel reading that begins from the lived experience of communities subjected to this framework. Adaptive management—treat every intervention as an experiment, monitor outcomes, adjust course—sounds responsible when articulated by research institutions. It feels different when you are the experiment. Indigenous communities in the Pacific Northwest experienced decades of "adaptive" salmon management that treated their fisheries, their food security, and their treaty rights as experimental variables. Each iteration was presented as learning; each adjustment required them to absorb another round of disruption while managers collected data. The framework systematically transfers risk downward—the designers experiment, the affected populations bear the cost of learning.
Applied to AI displacement, this becomes particularly sharp. Adaptive governance for the AI transition means workers, students, and communities navigating "experimental" labor markets, "experimental" education redesigns, "experimental" social safety nets—all while researchers monitor outcomes and adjust parameters. The Everglades didn't have to pay rent during restoration experiments. The people most affected by AI's reorganization do. The framework assumes resources sufficient to absorb learning costs, timeframes compatible with institutional patience, and political arrangements where experimental subjects have meaningful input into experimental design. These conditions hold for ecosystems managed by well-funded agencies. They rarely hold for labor transitions affecting millions of people with immediate subsistence needs and negligible influence over the institutions running the experiments.
In the AI Story
Adaptive management emerged from Holling's 1970s work on environmental assessment and was articulated in Adaptive Environmental Assessment and Management (1978). The core insight: in complex systems where outcomes cannot be predicted, the manager's task is not to execute the right plan but to learn the right plan through disciplined experimentation.
The Everglades restoration provides the canonical contemporary example. After decades of command-and-control water management that met engineering objectives while destroying the ecosystem, the Comprehensive Everglades Restoration Plan adopted an explicitly adaptive framework. It was slower, less efficient, less certain about specific outcomes — and more effective, because it could respond to surprises.
Applied to AI, adaptive management suggests governance mechanisms that establish goals (broadly distributed benefit, protection against displacement, maintenance of judgment-development infrastructure) and implement interventions as experiments rather than specifications. The adaptive governance framework operationalizes this for the AI transition.
Adaptive management requires institutional and political cultures that can tolerate experimentation and failure. This tolerance is rare. Most institutions reward the appearance of certainty and punish visible failure, creating incentive structures opposed to adaptive learning. The reorganization window is the period during which institutional design choices that embed or foreclose adaptive capacity are being made.
Origin
Holling and colleagues at UBC developed adaptive management in the 1970s, articulated in Adaptive Environmental Assessment and Management (1978). Extended by Carl Walters, Kai Lee, and the Resilience Alliance.
Key Ideas
Intervention as experiment. Every action is a test of hypotheses about system behavior; monitoring is the experimental measurement.
Portfolio over single model. Maintain multiple approaches simultaneously; let outcomes select among them.
Tolerance for failure. Failure is data; institutional structures must treat it as learning rather than as violation.
Not the absence of management. Adaptive management is more demanding than command management, not less.
Appears in the Orange Pill Cycle
Experiments Require Power Symmetry — Arbitrator ^ Opus
The substantive validity of adaptive management depends entirely on who holds experimental authority and who bears experimental costs. When these align—when experimenters and affected populations substantially overlap, or when political arrangements give affected communities genuine veto power over interventions—adaptive management delivers what Holling describes: disciplined learning under irreducible uncertainty, superior long-term outcomes despite lower short-term efficiency. The Everglades case works because the ecosystem can't relocate and organized constituencies successfully imposed adaptive frameworks on agencies. This represents maybe 15% of adaptive management applications.
The contrarian reading dominates (75%) in most actual deployments: adaptive management as implemented typically runs experiments on populations with insufficient power to refuse participation or shape protocols. This isn't Holling's framework failing—it's institutions adopting the vocabulary while maintaining command-and-control power relations. The framework itself is neutral regarding this distribution; the political economy of its application is not. AI governance faces this directly: adaptive approaches to labor transition could mean worker organizations running experiments on training and safety net designs with genuine authority to terminate failed interventions, or it could mean policy institutions treating displaced workers as experimental subjects with "stakeholder input."
The synthetic frame: adaptive management is the correct technical approach to irreducible uncertainty, and it requires political arrangements most societies don't have. The reorganization window is when these arrangements might be built—or when adaptive management becomes a legitimation vocabulary for risk transfer. Which emerges depends on whether affected populations gain institutional authority proportional to their exposure, not on the technical merits of the framework itself.
Further reading
- Holling (ed.), Adaptive Environmental Assessment and Management (1978)
- Walters, Adaptive Management of Renewable Resources (1986)
- Lee, Compass and Gyroscope (1993)