Show the work. Show the limits.
A label is the beginning of an explanation, not a substitute for one.
1. Define the exact statement
Identify the population, time, place and terms that matter. Separate bundled claims. Decide whether the question is empirical, mathematical, historical, interpretive or not empirically testable.
2. Find the strongest available evidence
Prefer original measurements, research, documents and data. Use authoritative syntheses when the claim spans a large field. Check authorship, publication dates, methods, corrections and the connection between the source and the exact statement.
3. Look for corroboration and conflict
Independent observations are stronger than many sites repeating one source. Examine credible conflicting evidence, confounders, uncertainty and the limits of the methods. An institution’s name alone is not proof.
4. Use the right form of reasoning
A mathematical identity needs a proof under stated definitions. A historical claim needs authenticated records and corroboration. An empirical claim needs relevant observation or experiment. A theological or subjective proposition may not be resolvable by empirical testing.
5. Draft, review and approve
AI-assisted assessments enter the review queue. The owner checks the statement, sources, classification and caveats before approving publication. The initial collection was published following the owner’s explicit approval on October 8, 2026. New visitor submissions never publish automatically.
6. Keep the conclusion accountable
Show the review date and revision history. Explain what could change the conclusion. Investigate correction reports and record substantive amendments. Classifications describe the evidence for the exact statement; they are not probabilities or votes.