Appeal to authority

What an appeal to authority is

An appeal to authority is an informal fallacy of relevance in which the standing of a source substitutes for the strength of the case. What distinguishes it from its siblings among the logical fallacies is its respectability: unlike ad hominem, which is the same move with the sign reversed, an appeal to authority usually looks like exactly what good practice recommends. Everyone is told to check their sources, and this fallacy hides inside that instruction.

The Latin name, argumentum ad verecundiam, translates roughly as an argument to modesty or shame, and the name explains the mechanism. It works by making disagreement feel presumptuous. John Locke used the phrase in his Essay Concerning Human Understanding, first published in 1689, for arguments that pressure an opponent by invoking the eminence of those who held the view, so that objecting looks like arrogance rather than like inquiry. That is still the tell: the citation is doing social work, not evidential work.

The shortest accurate definition, and why it is a fallacy at all

The shortest accurate definition is one clause: believing a claim because of the badge on the speaker rather than the reasons in the claim.

It is a fallacy because expertise is a probabilistic guide, not a truth guarantee. A qualified engineer is more likely than a stranger to be right about load bearing walls, and that likelihood is a reason to take the engineer seriously, not a reason to stop checking. The fallacy occurs at the point where the appeal becomes deductive: the expert says so, therefore it is so. That inference is invalid, and it stays invalid however good the expert is. Note the modality carefully, because it is the whole distinction. "An expert says X, so X is probably worth investigating" is sound. "An expert says X, so X is true" is not.

The form a sound expert citation takes: 5 conditions

The form a sound expert citation takes is fixed by 5 conditions, and a citation that fails any one of them is weaker than it looks.

1. Match the field to the claim. The authority's expertise must cover the actual question. A distinguished architect endorsing a soil testing kit is speaking outside their field, which is the commonest failure and the one advertising uses most.
2. Check for consensus. If specialists in the field broadly agree, one expert citation represents that agreement. If they are split, quoting one of them tells you only that the split exists.
3. Check the interest. An authority paid by the party whose claim they support is not disqualified, but the citation carries less weight and the underlying work needs to be visible.
4. Identify the person, not the costume. "Experts say" and "studies show" are not citations. A real citation names who, where and when, and can be looked up.
5. Ask whether the reasons are available. The strongest form of expert testimony comes with its working attached, so that the authority is a shortcut to the evidence rather than a replacement for it.

A worked example, and the repair

Here is a worked example with the repair. A supplier's brochure for a bicycle lock reads: "Recommended by a former national cycling champion." The claim being made is that the lock is hard to defeat.

Test it against the 5 conditions. The field does not match: winning races establishes nothing about resistance to bolt cutters. Consensus is untouched, since no locksmith or security laboratory is mentioned. The interest is undisclosed, and an endorsement is usually paid. The person is named, which is the one condition met. And no reasons are supplied at all: not a rating, not a test standard, not a cutting time.

The repair is to ask what the authority would have to be an authority about, then look for that instead. For a lock, the relevant sources are the published security ratings from independent testing bodies and the tools those tests use. The general form of the repair is a substitution: replace "who said it" with "what would settle it". If nothing would settle it, that is a more interesting finding than the endorsement.

What an appeal to authority is confused with

An appeal to authority is confused with 3 neighboring appeals that share its structure, and with 1 thing that is not a fallacy at all: an ordinary correct citation of a relevant expert. In each of the 3, a property with nothing to do with truth is offered as though it had, with age, origin or popularity standing in for eminence.

Appeal to nature deserves one clarification, since it is often confused with the naturalistic fallacy named by G. E. Moore in Principia Ethica (1903). Moore's target was the attempt to define goodness in terms of a natural property. The appeal to nature is the plainer move of treating natural as a synonym for good or safe. They are related and they are not the same, and a rope's breaking strain is a number either way.

When citing an authority is not a fault

Citing an authority is not a fault in most of life, and a page that only warns about the move is useless. 3 cases are not merely acceptable but correct.

1. Defer where the evidence is inaccessible. No one can personally verify metallurgy, tide tables and building codes at once. Relying on the people who can is how a division of labor works, and it is not a fallacy, it is arithmetic about time.
2. Report consensus as consensus. Saying "the specialists who study this broadly agree" is a claim about the state of a field, and it is either accurate or not. That is a factual report, not an appeal.
3. Use expertise to allocate attention. If a qualified structural engineer says a beam is suspect, the correct response is to have it examined, not to demand the engineer prove it from first principles before anyone looks.

The honest limit runs in the other direction too. Rejecting a claim because an expert made it is not skepticism, it is ad hominem wearing a contrarian coat, and it fails for exactly the same reason.

The test to run on the next expert you are shown

Run this test on any claim that arrives with a credential attached. Ask: what is this person an authority about, and is that the same thing as the claim being made?

Then ask the second question, which catches the polite version of the fallacy: if this citation were removed, would there be any reason left to believe the claim? If the answer is no, you are being asked to defer rather than to be convinced. That may still be reasonable, on a subject where you cannot check, and it is worth knowing which of the two is happening. Confirmation bias makes the difference hard to see from the inside, because an authority who agrees with you feels authoritative and one who does not feels biased.

Appeal to tradition and appeal to nature, the two closest relatives

Two fallacies in the table above deserve their own treatment, because each substitutes a different irrelevant property for evidence and each has a legitimate version that is easy to lose sight of.

Appeal to tradition treats duration as justification: this is how it has been done, therefore this is how it should be done. A factory that has serviced a conveyor every six months since 1974 has a schedule with a long history and no stated reason. The legitimate version is worth keeping, though: a practice that has survived a long time in a demanding environment has been tested by that environment, and the survival is weak evidence that it works. Weak evidence is still evidence, and the honest question is what would have happened if the practice were wrong. If nobody was watching, longevity tells you nothing.

Appeal to nature treats origin as quality: this is natural, therefore it is better or safer. The counterexamples are immediate in both directions, since untreated timber rots and synthetic fiber outperforms natural fiber in most rigging applications. The legitimate version is narrower than the slogan: a material with centuries of use in one specific application has a known failure behavior, and that is a fact about accumulated data, not about naturalness. Ask which property is actually being claimed, then ask for the number that measures it.

Where to go next

FALLACIESLogical fallaciesWhat they are, with examplesFALLACIESAd hominemAttacking the person, not the argumentFALLACIESNon sequiturThe conclusion that does not followBIASESConfirmation biasDefinition and examples