Our Method
A description of how material here is produced, precise enough that you could tell whether we followed it.
The production sequence
- Question first. Every piece starts with a question a reader would actually ask, written down before drafting. If the question cannot be stated in one sentence, the piece is not ready to write.
- Evidence gathering. Sources are collected and read before drafting, not found afterwards to support a written argument. Where a claim requires external evidence and none is available, the claim is cut rather than softened.
- Structure. Instructional material follows a fixed shape: objective, plain explanation, worked example, exercise, checklist. Reports follow a different fixed shape: question, judgments with confidence, methods, limitations, references.
- Drafting. Written for a reader at roughly an eighth to tenth grade reading level for core instructional content, with specialized terms defined at first use.
- Review. Checked against the editorial standards, with particular attention to whether each claim is supported at the strength it is stated.
- Publication and revision. Published with a date. Substantive changes are dated separately and the original publication date is never altered to make material appear new.
Analytic techniques we use
The reasoning material on this site is drawn from structured analytic techniques used in professional analysis, simplified for general use without changing what they do.
- Analysis of competing hypotheses. Generating several explanations and scoring evidence for inconsistency rather than support. The everyday version appears inWhat Else Could Explain This?
- Key assumptions check. Identifying which unstated premise a conclusion cannot survive without. This is the basis of theassumption audit worksheet.
- Calibrated confidence language. Confidence words used consistently and defined, with the evidence they rest on stated. Covered inHow Confident Should You Be?
- Lateral reading. Evaluating a source by leaving it, drawn from research comparing professional fact checkers with other skilled readers.
How we handle uncertainty
Claims are stated at the strength the evidence supports and no higher. Where the evidence is thin, the page says so. Where an important question is unresolved, the page says what would resolve it rather than picking a side and presenting it as settled.
Reports carry an explicit confidence level. Low means thin, single-sourced, or indirect evidence. Moderate means reasonable evidence with a known gap or unresolved alternative. High means several independent lines of evidence, a stated mechanism, and no substantial contrary evidence. None of these means certainty.
What we do not do
- We do not rate outlets. We teach a procedure for evaluating a source rather than publishing a list of approved and disapproved publishers. A rating substitutes our judgment for the reader's, which is the opposite of the aim.
- We do not adjudicate live political disputes. Not because they do not matter, but because a public education initiative that does so becomes a participant, and its materials stop working for half its audience.
- We do not treat ordinary disagreement as misinformation. People who share our facts and reach different conclusions are disagreeing, which is a normal and legitimate thing.
- We do not use fear as a motivator. Alarm is easy to produce and does not durably change behaviour.
Fictional and real examples
Every Think Lab scenario is fictional and labelled as fictional on the page. Worked examples in lessons are constructed unless attributed to a source.
We do not publish a factual case study unless its sources can be verified and linked. A real example with an unverifiable source is a worse teaching tool than a clearly labelled constructed one, because it teaches the reader to accept unverified material when the framing is authoritative.
Related pages
Editorial standards covers sourcing and citation rules in detail. Corrections covers what happens when we get something wrong.