How honest institutions end up publishing claims they cannot support

Nobody decides to overstate. The drift happens one comparative adjective at a time.

The sequence is familiar. A draft says the institution has small class sizes. Somebody suggests very small. Somebody else, wanting warmth, offers exceptionally small. A later revision, working from the previous version rather than the original fact, arrives at among the smallest anywhere. At no point did a person intend to mislead. Each edit was a small strengthening of the one before, and because the source fact was never carried alongside the sentence, there was nothing to check the strengthening against.

Two conditions make this drift almost inevitable. The first is writing from previous copy instead of from evidence, which means every revision compounds the last one. The second is reviewing for tone rather than for accuracy: the question in the room is whether the sentence sounds confident, not whether the underlying claim is still the one the records support. Both conditions are cultural rather than technical, and both are fixable by changing what the review meeting asks.

The repair is to attach the source to the sentence. Beside every factual claim in a draft, a note: where this came from, who holds it, when it was last true. Claims that cannot carry a note get rewritten or removed. This slows the first draft and speeds everything afterwards, because when somebody eventually asks where a number came from, the answer exists rather than having to be reconstructed by a person who was not there.