The expertise layer · Judgment

When expertise should outweigh consensus

Counting heads is the right way to settle some questions and the wrong way to settle others. The trick is knowing which is which, because using the wrong rule quietly hands the answer to the wrong people.

On questions of fact, informed expertise should weigh more than a simple head count. On questions of value, it should not. That principle decides a great deal about how human judgment over AI ought to work. When the question has a right answer, the engineer who has studied the system should count for more than a crowd that has not. When the question is about what we should want, the expert has no special authority, and equal voice is the honest rule. I am stating a principle here, not a mechanism: how any particular system might act on it is a separate matter, and not the point of this page.

Why consensus is not truth on a factual question

Because reality does not take a vote. On a matter of fact, a hundred confident people can be wrong and one informed person right, and a system that defers to the majority there will reliably mistake popularity for correctness. This is where expertise earns its extra weight: not as status, but as a better track record against a knowable answer.

The line that keeps it honest

Expertise stays in its lane. The moment a question turns from "what is true" to "what should we value," the expert's extra weight evaporates, because there is no fact for them to be expert about. Keeping that line bright is what stops "trust the experts" from becoming a quiet way to let specialists decide questions that belong to everyone.