Explainers · Bad evidence

Appeal to Authority

also called argumentum ad verecundiam, the credential as proof

The reason to believe it is who said it. Nothing the expert actually knows has been passed along — only the fact that they’re the sort of person who would know.

Send it instead of typing it out again.

The short version

A claim needs support. What arrives instead is a person — a doctorate, a title, a bestseller, a household name — and the implication that the question is now closed.

The odd thing about this one is how nearly it works. Most of the time, deferring to somebody who knows more than you is exactly the right thing to do. The problem isn’t trusting experts. It’s accepting the credential in place of the thing the credential was evidence of.

Where the name comes from

Argumentum ad verecundiam — “argument to modesty.” It’s Locke’s, and his version is sharper than the one we inherited. The move works, he thought, by making you feel immodest for pressing: not here is why this is true, but who are you to ask?

That’s still the engine. The name isn’t doing logical work. It’s setting a social price on the follow-up question.

What it looks like

What’s claimed

This supplement prevents the illness.

Why we should believe it No study, no mechanism, no numbers.
The support offered

The man who wrote the book has a doctorate.

What’s claimed

This is perfectly safe for children.

Why we should believe it The prize was for physics.
The support offered

A Nobel laureate said so.

What’s claimed

The market will have recovered by spring.

Why we should believe it He has been wrong about four of the last six springs.
The support offered

He’s the most famous investor alive.

Why it doesn’t work

Two different failures wear the same suit, and they’re worth keeping apart.

The credential in the wrong field. Expertise is narrow and doesn’t travel. A brilliant physicist holds no special standing on nutrition; a Nobel Prize is a fact about one specific body of work, not a general licence to be right. What’s really happening is the halo effect — competence observed in one place gets read as competence at large, because that’s very much cheaper than assessing each domain separately.

The credential in the right field, standing in for the reasoning. Subtler, and far more common. The doctorate is genuinely relevant. The person genuinely does know more than you. And still nothing has been transmitted. A qualification is evidence that somebody probably has good reasons. It is not the reasons. Offered instead of them, it’s an invitation to buy the box without the contents.

There’s a structural problem underneath both. Any field large enough to have a consensus is also large enough to contain credentialed dissenters — which means a name can be produced for very nearly any position. “A doctor said” is almost always true of something. The existence of one expert tells you approximately nothing; the distribution of them tells you a great deal, and it’s the distribution that never gets cited.

And it works best, as Locke noticed, on exactly the people being careful. If the reply to a name is a question, the questioner has to be willing to look arrogant. Most people aren’t, so most people stop, and the claim goes through unexamined.

This is not an appeal to authority

  • Deferring to expert consensus. You cannot personally verify the safety of a bridge, the dosage on a label, or the state of a literature you can’t read. Trusting the settled view of a field is not a fallacy — it’s the only way anybody gets through a day, and it has an excellent track record.
  • Citing an authority who gave reasons. “The report concluded this, section four” points at an argument. The authority is a signpost, not a substitute.
  • Weighting testimony by track record. Where the expert’s judgement genuinely is the evidence — a diagnosis, an assessment — their standing bears directly on how much weight it deserves.
  • Claims about what the experts hold. “Most epidemiologists think X” is a checkable statement about a population, and a sound place to start.

The test is what survives a push. Ask why they think it, and see whether there’s anything underneath the name. A real expert nearly always has an answer and is usually pleased to give it. A citation that can’t be unpacked is a citation being used as a wall.

How to call it out

  • “I’ll grant she’d know. What’s the actual reason?”
  • “Is that her field? What has she published on it?”
  • “Is that most people who study this, or one person who does?”

The third is the strongest, because it converts a name into a distribution, and distributions can be looked up. It’s also the fairest: it doesn’t ask you to doubt anybody’s competence, only to find out how alone they are.

The version worth running on yourself is quieter. Notice the things you believe because of who told you, where you could not now reconstruct the reason. That isn’t a failing — there is far too much to know, and delegating most of it is unavoidable. It’s just worth being able to tell which of your beliefs you checked and which ones you merely received.