When Centers Fail
Center-finding works when there is a center. Meet the conditions, independence, bounded variation, a typical the samples surround, and averaging lands on the truth, tighter with every sample. This asks the other question. What happens where a center does not exist. Several things lack one, and they are not edge cases. They are most of what matters.
Heavy tails. Some quantities have no finite typical. Wealth, city size, the size of a war. The largest sample is not an outlier around a mean. It is comparable to the sum of all the rest. The average does not settle. It lurches with each new extreme, because the extreme is the thing, not noise around the thing.
Multimodality. Two kinds wearing one name. Measure the time to recover across what is really two populations, and the mean sits in the empty valley between them, describing no one. It is a number placed exactly in the gap where no case lives.
Non-stationarity. The thing moves while you sample it. The center you find belongs to a quantity that no longer exists by the time you have found it. Yesterday's mean, computed perfectly, is a fact about a world that is gone.
Constructed quantities. Numbers projected onto what was never numeric. Intelligence scored, sentiment indexed, merit ranked. The apparatus runs, since you can average anything that has been assigned a number. But the average is the center of the projection, not of the thing, and the thing had no center because it was never a magnitude.
The singular. The thing that happens once. This marriage, this death, this life. The variation is not spread around a typical case. The variation is the case. There is no population. Averaging here is a category error, statistics worn as a costume over a question that has no statistical shape.
In every case the machinery does not object. It returns a number, with an error bar, and the bar tightens as you add data, and the tightening is the trap. It reads as confidence in a center that is not there. The apparatus cannot tell you that it does not apply. Only the seeing can. Deciding whether a thing has a center is not a computation. It is the prior act, and it belongs to intelligence, not to thinking.
This is the demarcation chaos draws. Where the recurring holds, thinking works, and the machine works better than we do. Where the thing is singular and happens once, no center exists to find, and the apparatus stays silent about its own silence. That silence is where intelligence lives.