What makes a source worth trusting

L3 said go to the primary source. This lesson is about judging what you find when you get there — because “primary” and “reliable” are not the same thing.

The instincts most people use here are close to useless. A polished site costs very little. A confident tone costs nothing at all. Institutional letterhead tells you an institution exists.

Four properties do considerably better.

1. It shows its evidence

Good sources let you get behind the claim: data, method, the actual document, a link to the thing itself.

Sources that assert without showing may still be right, but you cannot tell, which makes them useless for verification specifically. “According to experts” is not evidence. “According to this dataset, collected this way” is.

2. It corrects itself in public

The single most informative property, and the most overlooked.

Everyone is wrong sometimes. A source that never appears to be wrong is not more accurate than one that publishes corrections — it is one that does not tell you. Visible corrections are the only observable evidence that a source cares about accuracy rather than about appearing accurate.

Look for a corrections page, dated updates, an editor’s note. Their absence across a large body of work is a finding.

3. Its incentives are visible

The fastest single question in this module:

Who benefits if I believe this?

Not a conspiracy test. Everyone has incentives, including good sources. The question is whether they are visible, and whether the claim happens to be exactly what the incentive would predict.

A vendor’s study finding their category effective. An industry body finding regulation harmful. A campaign group finding the problem urgent. None of these are automatically wrong — and all of them need a second source before you rely on them.

Undisclosed funding is worse than obvious funding. A visible interest can be priced in; a hidden one cannot.

4. It separates knowledge from inference

Reliable sources are audibly careful about what they actually established versus what they suspect. We found X. This may suggest Y. We did not test Z.

Sources that present inference in the same register as measurement are telling you they do not track the difference — which means you cannot track it either.

This is the same property E4·L3 asked summaries to preserve, and E5·L7 warned about losing. It runs through the whole curriculum because it is where accuracy mostly lives.

Authority is a weak signal, not no signal

A well-regarded institution is somewhat more likely to be right, and it is much weaker evidence than people treat it as. Institutions have been wrong at length, have incentives, and are cited for things they never said.

Use authority to decide where to look first. Do not use it to decide what is true. That is exactly the E7 pattern: signals point, evidence decides.

Try it now (6 minutes)

Take a source you rely on. Check whether it has a corrections page, whether it shows its evidence, and who funds it.

Then ask who benefits if you believe its main claim. If the answer is “the people who wrote it”, find a second source before using it.

Check your understanding

1. Which property most reliably indicates a source cares about accuracy? A professional design and a recognised name It publishes corrections It never appears to be wrong

2. "Who benefits if I believe this?" is asked in order to: Identify conspiracies Make incentives visible, so a claim matching them gets a second source Decide whether the source is biased and can be dismissed

3. Institutional authority should be used to: Decide what is true Decide where to look first Rank sources definitively

Recap

Judge a source by whether it shows its evidence, corrects itself in public, makes its incentives visible, and distinguishes what it found from what it infers. Ask who benefits if you believe it. Treat authority as a hint about where to look, never as a substitute for looking.