The reason an example is not evidence but persuades as one

The report contains a table covering eleven thousand cases and one interview with a named woman in Kyiv. Six weeks later the table is gone from memory and she is what the report said. She was chosen as an illustration of the table.

What everyone sees

An example is understood as a way of making data accessible, subordinate to the aggregate it illustrates. In practice the ranking inverts. The concrete case is encoded with imagery and causal structure and the aggregate is a number, and when they are recalled together the case supplies the interpretation.

What is actually happening

Hamill, Wilson and Nisbett gave participants a single vivid case together with explicit information that it was unrepresentative, and found generalisation to the wider population barely reduced by the warning — the sample-bias information failing to blunt the inference drawn from the case. Borgida and Nisbett found the complementary asymmetry in decision-making, where concrete first-hand accounts influenced choices more than statistically superior aggregate summaries covering the same ground. Neither result says examples should be avoided. It says the example is doing the persuading, so the honest use is to select it at random from the distribution rather than for how well it illustrates the point.

Why it stays hidden

The substitution hides because the aggregate is present. The table is in the document, the methodology is described, and anyone can check that the claim is supported. What cannot be checked from the page is which element the reader retained, and the retained element was chosen for vividness by whoever wrote it.