G11: Contested Epistemic Authority

The same agency, twice

Answer two questions quickly, before you think about them too hard.

Suppose a national health agency announces that a common medical product is safe, and a popular podcaster with no medical training tells millions of listeners that the agency is wrong and they should research it for themselves. Should an ordinary listener defer to the agency?

Now suppose that a few years later the same agency, under new leadership you distrust, announces that the same product is unsafe, and a group of credentialed specialists publicly says the agency is wrong. Should an ordinary listener defer to the agency now?

If you answered "defer" to one question and "look at the evidence yourself" to the other, notice what decided it. A fixed view about when to trust official experts would give the same answer both times; a view about which experts are on the right side would not. Whichever way you switched, you have just met this grammar in yourself.

Source still being found. research showing that people judge whether someone counts as an expert, or whether a consensus exists, partly by whether the expert's conclusion fits their own views; needed before the page can say this switch is common rather than merely possible

Try one more. Imagine a woman with a chronic illness says a treatment changed her life, while a review of trials finds it no better than a placebo. Imagine a parent says a school's reading curriculum is failing her child, while researchers say it is well supported. In which case does the person with direct experience deserve the last word? If your answer differs, ask what changed: the evidence, or the answer you wanted.

What the grammar is

Almost everything any of us believes, we believe because someone told us. Few people have checked for themselves that smoking causes cancer or that a country they have never visited exists. This is not a failing; nobody can verify everything. It means the most practical question about knowledge, most of the time, is not "what is true?" but "whom should I believe, and about what?"

This grammar is what happens when that question is fought over instead of asked. Several rival sources of authority compete to be believed: formal, credentialed expertise; the lived experience of the people a question affects; the "do your own research" of the independent investigator; the crowd, whether in the form of prediction markets or crowd-written fact-checks; and, increasingly, the algorithmic feed and the AI system that answers in confident paragraphs. Each faction runs its own way of deciding what counts as knowledge, and each accuses the others of gullibility. The deep question — when should you defer, to whom, and about what? — almost never gets asked directly. The grammar is sharpest when several institutions reach different conclusions from what should be the same underlying facts, because then even someone who wants to defer has to choose whom to defer to.

This site calls a recurring pattern of argument like this a grammar: a shape the same mistake takes across many different subjects, so that once you can see it in one argument you can see it in others. This one has two poles, and each is a real failure.

Pole A, the credentialists, treat a credential as if it settled every question near it. They appeal to authority beyond the authority's domain, as when a scientist's standing in one field is spent on a policy question in another. They turn "trust the science" into a slogan that misrepresents how science works, as if science were a set of pronouncements rather than a method of revising them. And they overclaim, speaking with more certainty than the evidence supports, which spends down the public trust that expertise depends on. They are right that expertise is real. They are wrong when they treat trust as owed rather than earned.

Pole B, the epistemic populists, treat a credential as a reason for suspicion. Their do-it-yourself research is often motivated curation: collecting the sources that agree with what they already believe and calling the collection an investigation. They generalize from a few expert failures to never deferring at all, as if "the experts were wrong once" meant "never defer to experts again." And they practice what might be called expertise substitution: the confident podcaster stands in for the boring specialist who reports error bars. They are right that experts err and institutions can be captured or careless. They are wrong when they treat every failure as proof that no one knows better than they do.

Taking the question apart

Undoing the grammar means asking the question it hides. Four questions do most of the work.

Whose field is this, exactly? Expertise is bounded. A virologist knows about viruses, not necessarily about school closures, which also involve child development, economics and law. A patient knows what her illness feels like, not necessarily whether a drug works across a population. Ask which part of the question each claimant actually has special access to.

How strong is the agreement, and among whom? "Experts say" can mean a finding replicated over decades, or one press release. Ask whether the people agreeing are inside the relevant field and how many independent lines of evidence they rely on.

What kind of claim is it? Some claims are about what a person experienced: this treatment made me feel better. On those, the person who lived it often has the best evidence there is. Other claims are about causes and populations: this treatment works. On those, one experience is a single data point, however vivid. Many fights in this grammar are two people answering different kinds of claim.

What is the track record, on this kind of question? An institution wrong about one thing is not thereby wrong about everything, and one right before is not thereby right now. Ask how reliable a source has been on questions like this one, not whether it has ever failed.

None of these tells you whom to believe. They turn "trust the experts" and its opposite from slogans into questions with answers in particular cases.

Why it is so persistent

The first reason is that deference is unavoidable and uncomfortable. No one likes to admit how much they take on trust, so we describe our own trust as judgment and other people's trust as gullibility. Each side sees the other as sheep, and itself as the flock that happens to be right.

The second reason is that trust is not evenly distributed. People tend to trust sources that share their values, and institutions are not equally trusted by every community. Someone whose community has real reasons to distrust an institution will reasonably discount it; someone whose community has done well by it will reasonably extend it credit. Both are calibrating trust on experience. The trouble starts when that calibration hardens into a rule applied to every question.

Source still being found. research on partisan or group differences in trust in scientists, doctors or public-health institutions, ideally showing that trust varies by group and by issue rather than uniformly

The third reason is a habit the novelist Michael Crichton named Gell-Mann amnesia: noticing the press get your own field wrong, then trusting it on everything else. The same habit runs in reverse. Someone who has seen experts get one thing wrong, especially something in their own life, may stop trusting them about anything at all. Both habits are failures to grade trust by domain.

The fourth reason is that the difference between skepticism and cynicism is easy to blur. Skepticism produces a graded verdict: this claim is well supported, that one is not. Cynicism produces the same verdict whatever the evidence, and calls it wisdom. Cynicism feels like independence and costs nothing, because it never has to evaluate anything.

Source still being found. a scholarly account of the difference between calibrated skepticism and generalized distrust, or of how misinformation strategies aim to produce general distrust rather than belief in any particular claim

How the argument has been had before

Each time, something changed who could claim to know, and both sides were partly right.

The priesthood of all believers. In the Protestant Reformation of the sixteenth century, reformers argued that ordinary believers could read and interpret scripture for themselves, without a priest to mediate it. The reformers were right that a closed class of interpreters can serve its own interests. Their opponents were right that reading a text is not the same as interpreting it well, as the many competing readings that followed, each claiming to be the plain meaning, made clear.

Source still being found. a reliable historical account of the "priesthood of all believers" doctrine, its origins with Martin Luther, and the debate over lay interpretation of scripture

The patent-medicine era. In the nineteenth century, remedies with confident claims and secret ingredients were sold directly to the public, in competition with a medical profession still uneven in what it knew. Skeptics of the profession were not wrong that much of its practice was poorly founded; buyers of patent medicines were wrong that a confident label was a substitute.

Source still being found. a historical account of how reliable mainstream nineteenth-century medicine was, to support the claim that much of its practice was poorly founded at the time

Example still being found. a real, quoted, dated patent-medicine advertisement or seller's claim that set itself against doctors or the medical establishment, and a sourced account of the reform that followed, such as the US Pure Food and Drug Act of 1906

Self-diagnosis online. With medical websites such as WebMD, patients began arriving at appointments with their own research. Doctors were right that a search result is not a diagnosis; patients were right that they were entitled to understand their conditions, and sometimes found what a hurried appointment had missed.

Source still being found. research or reporting on how online symptom-checking changed the doctor–patient relationship, with any evidence on accuracy

COVID's expert whiplash. During the COVID-19 pandemic that began in 2020, public-health guidance changed repeatedly as evidence accumulated, and different institutions sometimes gave different advice. Those who defended the experts were right that changing guidance as evidence changes is how science is meant to work. Those who distrusted them were right that some guidance had been stated with more certainty than the evidence supported, and that this made later corrections look like reversals.

Example still being found. a real, quoted, dated instance of public-health guidance during COVID that was stated with high confidence and later revised, with the revision, from a sourced record

How each pole fails

The two poles fail in mirror-image ways.

The credentialist answers a question about evidence with a fact about status. The expert's title is cited in place of the expert's reasons, and the title is stretched to cover questions outside the expert's field. When an expert overstates certainty and is later shown wrong, the credentialist treats the resulting loss of trust as the public's irrationality rather than as a cost the overstatement incurred. At its deepest, it treats disagreement with experts as bad faith, so a reasonable question and a crank theory get the same answer.

Example still being found. a real, quoted, dated instance of Pole A, an expert or institution claiming authority beyond its domain or overstating certainty, from a source associated with the political left or with established institutions

Example still being found. a real, quoted, dated instance of Pole A from the political right, for example an appeal to a credentialed authority used to close a question outside that authority's field

The epistemic populist makes the opposite moves. The populist answers a question about evidence with a fact about motive: the experts are paid, captured or arrogant, so their conclusion can be ignored. The populist's own research is selective in exactly the way the experts are accused of being. One expert failure becomes a general verdict, and one credentialed dissenter becomes proof that the consensus has collapsed. At its deepest, it is cynicism mistaken for independence: because no one can be fully trusted, every source is treated as equally untrustworthy, which in practice means trusting whoever sounds most sure.

Example still being found. a real, quoted, dated instance of Pole B from the political right, such as a "do your own research" claim or a generalization from expert failure to never deferring

Example still being found. a real, quoted, dated instance of Pole B from the political left, such as lived experience or community knowledge used to dismiss contrary evidence on a question of causes or averages

Both poles share one move: each settles the question of whom to believe before looking at the question being asked. The credentialist settles it by title, the populist by suspicion of titles, and each feels the other is the one who has stopped thinking.

Where it borders other grammars

This grammar rarely travels alone.

With G10: Victimhood as Standing it shares the confusion between who may speak and whether a claim is true. "You have no standing to say this" and "you have no credentials to say this" are the same move, aimed from different directions.

With G6: The Hidden Hand it shares the populist's explanation for expert error. If the experts were wrong, someone must have wanted them to be, and the failure becomes evidence of a cover-up.

With G4: Loyalty, Heresy, and Betrayal it shares the treatment of dissenters. A credentialed expert who breaks with a consensus may be cast as a heretic by one side and as a hero by the other, with neither side asking whether the dissent is sound.

With G9: Language as Territory it shares a fight over words. Terms such as "misinformation," "the science" and "lived experience" can be used to describe a claim or to rule it out of bounds before it is examined.

When naming the pattern becomes a weapon

The vocabulary on this page can be used to end an argument instead of examining it.

"Appeal to authority" is the obvious example. Deferring to someone who knows more than you about a question in their field is not a fallacy; it is how almost all knowledge works. Calling every reference to experts an "appeal to authority," or "credentialism," dismisses the most reliable evidence available without evaluating it.

The weapon also points the other way. "Conspiracy theorist," "anti-science" and "misinformation" each name something real, but each can be stretched to cover a reasonable question about an institution's record or a legitimate disagreement among experts. A patient who says her doctor dismissed her symptoms is not anti-science; she may be right.

Example still being found. a real, quoted, dated instance of "anti-science," "misinformation" or a similar label applied to a question that turned out to be reasonable, and a matching instance of "appeal to authority" or "credentialism" used to dismiss well-supported expert evidence, from different sides

A label that names a real failure in general does not settle whether the failure is happening here. Naming the grammar should open the question of whom to trust about what, not answer it.

Catching it in yourself

A few questions help.

Did my rule change, or did my side change? Go back to the questions at the top of this page. If you trusted an institution when it agreed with you and distrusted it when it didn't, your rule about deference was really a rule about conclusions.

Which experts do I quote, and which do I explain away? When an expert agrees with you, do you cite their credentials? When an expert disagrees, do you look for their funding?

Whose lived experience counts for me? When someone with direct experience says something you agree with, do you treat it as testimony? When they say something you disagree with, do you call it an anecdote?

Is my doubt graded or total? Can you name something the institutions you distrust probably get right, and something the ones you trust probably get wrong? If you can't, you may have stopped evaluating.

Am I defending territory, or mapping it? When you "do your own research," are you looking for what would change your mind, or for confirmation of what you already think?

Tools that help

• C3: The Consensus Audit. Everyone defers to someone, so check the quality of your deference. Ask whether a claimed consensus is inside the relevant field and how strong the agreement really is. Its misuse runs both ways: stretching "the science is settled" over unsettled questions, and hunting for one credentialed contrarian to declare the consensus broken.
• C5: The Cynicism Firewall. Skepticism grades claims; cynicism rejects everything and calls it wisdom. It lets you lower your confidence in a particular class of unreliable claims without abandoning the well-supported core. It also helps you notice when someone is flooding a debate with confusion until people stop believing anything can be known.
• F2: The Standing/Soundness Split. Who may speak and whether the claim is true are different questions; never let one answer the other. It gives lived experience its due on what a person directly experienced while still testing their broader claims on the evidence, and it does the same for credentials. Its misuse is "you can't speak on this" used against a sound argument, or "facts don't care about your feelings" used to dismiss testimony that is itself the relevant fact.

Four tools work against every grammar, this one included: A4: The Mirror Check, C2: The Steelman Protocol, E5: The Precision Covenant and F4: The Scout Report. The full list is at /tools.

Where it shows up now

As of October 2026, the site has not yet examined this grammar in a live argument. These are places to look, not findings.

• Health populism under Robert F. Kennedy Jr. as US health secretary. Not yet examined.
• "Lived experience" as a trump card, and the backlash against it. Not yet examined.
• AI-generated confidence versus human expertise. Not yet examined.
• Community Notes (crowd-written fact-checks, first used on X) versus professional fact-checkers, as rival ways of deciding what is true. Not yet examined.
• Homeschooling and unschooling, and their ideas about who can teach and what counts as knowing. Not yet examined.
• Financial advice on TikTok. Not yet examined.

Related concepts

• Epistemic trust calibration. Matching how much you trust a source to how reliable it has actually been on the kind of question at hand.
• Domain-boundedness of expertise. The fact that expertise in one field does not carry over to another.
• Gell-Mann amnesia. Michael Crichton's name for noticing the press get your own field wrong, then trusting it on everything else.
• Consensus auditing. Checking who agrees, in which field, how strongly, and on what evidence, before treating a consensus as settled.
• Skepticism and cynicism. Skepticism grades claims by their evidence; cynicism rejects them all regardless.
• Appeal to authority. Citing someone's standing as evidence for a claim; a fallacy when the authority is outside their field or the appeal replaces evidence, and ordinary good sense when it is not.
• Standing and soundness. The difference between whether someone may speak on a question and whether what they say is true.