"Expert systems fail when they ignore local knowledge.
Central dashboards and policy can look like expertise from a distance. Up close, they can miss the constraints that only the domain owners and on-call engineers see, and the system pays for that gap.
Hayek’s point is that systems fail when decisions are made far from the work. A metric proposal that changes incentives will be gamed or resisted if it ignores local knowledge: where the bottlenecks are, what is brittle, and what is already overloaded. The metric doesn’t just measure; it routes attention, and teams will optimize what survives review.
Before you ship the metric, surface the assumption it makes about the work and name the first signal that it is distorting behavior. Then decide what you will do when that signal appears, so local reality can correct the expert plan.