The Precision-Without-Source Signal for Fabricated Specifics

A fast filter for a slow problem

Verifying a statistic from scratch is expensive; you have to find the real number and confirm it. The Claude Certified Associate - Foundations (CCAO-F) exam teaches a cheaper first-pass filter for the specific case of fabricated figures: look at whether the precise number carries a source. This turns the fabricated-specific pattern from the general taxonomy into a concrete, repeatable check you can run in seconds before deciding whether a slower verification is even needed.

The insight is about how real statistics behave in the wild. A figure precise enough to be quoted to a decimal place almost always came from somewhere specific, and in professional writing it is reported with that somewhere attached. So the presence of precision with the absence of a source is an anomaly, and anomalies are exactly what a first-pass filter should flag.

Why precision plus no source is the tell

Precision reads as authority. A statement that "roughly a majority" of firms did something invites scrutiny; a statement that "63.4 percent" of firms did it feels settled. That psychological weight is precisely why fabricated specifics are dangerous, and precisely why the missing source matters. When a figure is that exact, the reader's instinct is to trust it, so the one thing that would let the reader check it, the citation, is the thing most worth demanding. Its absence, next to that much precision, is the signature of a number-shaped guess rather than a reported fact.

The citation check comes before verification

The reason this is a filter and not the whole job is efficiency. Independently verifying a number means research; checking whether a citation is even attached takes a glance. Running the glance first triages your effort: figures that arrive with a genuine, traceable source can be spot-checked, and figures that arrive precise but uncited are flagged for either a demand-for-source re-prompt or removal. You do the expensive verification only where the cheap filter says you must, and you never let a precise-but-uncited figure through on the strength of its precision alone.

What the CCAO-F exam trips candidates on

The first trap is trusting a precise-sounding statistic more than a vague one, when the reverse caution is warranted without a source. The exam deliberately dresses a fabricated figure in extra precision to test whether you read that as rigour or as risk. The credited answer treats an uncited precise number as more suspicious than a hedged general one, because the precision is the persuasion, not the proof.

The second trap is accepting a citation that exists in name but cannot actually be located or opened as sufficient grounding. The filter is only satisfied by a source you can trace, not by a source-shaped string of text. A reference that looks like a citation but leads nowhere is itself a fabricated specific in the format of rigour, which is developed further in auditable versus untraceable citations.

Common misreadings to avoid

How this shows up on the exam

Domain 2 questions on this knowledge point present a precise figure and ask whether it can be trusted or what the risk is. The reliable move is to run the precision-without-source filter first: an exact number with no traceable citation is flagged as a likely fabricated specific, treated with more caution than a vague claim, and either re-grounded or verified before use.

This knowledge point sharpens the fabricated-specific signature from the hallucination pattern taxonomy into a concrete filter, and it hands off to auditable versus untraceable citations for what counts as a real source and to code execution for numeric verification for the numbers you must actually compute. It also feeds diagnosing failure patterns in outputs.