Confidence

Confidence is coverage

The confidence beside a score is the share of the methodology's total weight that the score was actually able to read. If every domain reported, confidence is 1. If a region can only report part of the ladder, confidence falls by the weight of what is missing.

This is the single most important property of the number, and it is the reason the weights total 1.18 rather than 1. Missing data lowers confidence. It does not move the score.

Why that choice matters

The alternative is to treat an absent domain as neutral, or to rescale silently and present the result as though nothing were missing. Both make a thin reading look like a thick one. Regions with sparse observation, the polar ones in particular, report fewer domains than the rest, and their confidence is visibly lower for it. That is the system working, not failing.

Confidence is never upgraded to look better

The same rule governs AIEDS: use the confidence the evidence supports, and never raise it because an estimate looks plausible. A domain fed by a representative sample rather than a live observation does not become more certain because the resulting figure is reasonable.