Engineering Resilience (Holling)
Engineering resilience is the measure embedded in most mechanical and computational systems: how quickly does a system return to its original state after disturbance? A bridge that deflects under wind load and recovers exhibits engineering resilience. The assumption is that there is one correct state and the system should return to it as rapidly as possible. Holling's 1973 paper argued that this conception is fundamentally inadequate for complex adaptive systems, which exist in multiple possible states and can shift between them. The AI transition is widely analyzed in engineering-resilience terms — how quickly can workers retrain, organizations restructure, markets stabilize — but the dynamics it exhibits are ecological rather than engineering.
In the AI Story
Engineering resilience assumes a single equilibrium. The task is to return to it. This assumption works when disturbances are small, conditions are stable, and the system's structure is well-characterized. It fails when disturbances are large enough to push the system across basins of attraction, when conditions are themselves changing, and when multiple stable configurations are possible.
The triumphalist reading of the AI transition is implicitly an engineering-resilience reading: the system has been perturbed, it will stabilize at a new equilibrium, the task is to accelerate the adjustment. This framing is intuitive — it describes most small-scale disturbances correctly — and dangerously wrong for the current transition.
Ecological resilience, Holling's alternative, asks a different question: not how fast the system returns, but how much disturbance it can absorb before shifting to a qualitatively different regime. The framing matters because the interventions it suggests are different — invest in diversity and redundancy rather than in speed of adjustment.
Origin
Holling introduced the distinction between engineering and ecological resilience in his 1973 paper 'Resilience and Stability of Ecological Systems,' which reframed ecosystem management across multiple disciplines.
Key Ideas
Single equilibrium assumption. Engineering resilience presupposes one correct state and measures speed of return.
Adequate for small disturbances. The framing works for perturbations within the current basin; it fails for perturbations across basins.
Wrong frame for AI. The AI transition is a basin shift, not a deflection — engineering resilience measures the wrong thing.
Appears in the Orange Pill Cycle
Further reading
- Holling, 'Resilience and Stability of Ecological Systems' (1973)
- Holling, 'Engineering Resilience versus Ecological Resilience' (1996)