Informational influence for people who want the details

You change your mind because somebody else appeared to know more than you. That is informational influence, and it is the most defensible kind there is. It is also the easiest to fake, because the counterfeit and the real thing feel identical while they are happening. Both arrive as the pleasant sensation of learning something.

The distinction is not in the feeling. It is in whether there is any evidence at the end of the chain.

The research behind the idea

Muzafer Sherif showed in the 1930s that when a judgment is genuinely ambiguous — a point of light in a dark room that seems to drift — people's estimates converge on a shared figure, and each person then keeps using that figure even when tested alone. Solomon Asch's line experiments in the 1950s ran the other way: a large share of people gave an answer they could see was wrong in order to match a unanimous group.

Deutsch and Gerard (1955) separated the two forces those experiments exposed. Informational influence is treating what other people say and do as evidence about how the world actually is. Normative influence is going along with them to meet their expectations, whatever you privately believe. The first tends to produce private acceptance, a genuine change of mind; the second produces public compliance, agreement that lasts exactly as long as the audience does.

A quieter relative is pluralistic ignorance: nearly everyone privately rejects a belief while assuming that everyone else holds it, so nobody speaks and the false consensus props itself up with no pressure applied at all.

What to take away

When deferring is the right move

You cannot verify most of what you need to act on. The alternative to trusting other people is not independence, it is paralysis, so the question is not whether to defer but to whom and on what.

Deferring is sound when three things hold. The other party has access you do not, through position, training, or having simply been there. Their interests are not opposed to yours in this particular matter, or the conflict is visible and priced in. And there is some mechanism by which they would be corrected if they were wrong.

That third condition does most of the work and gets the least attention. Expertise never tested against outcomes drifts: confidence grows while accuracy does not.

Someone who makes checkable predictions in public and is corrected when they miss is a different kind of source from someone whose claims never meet a result, even when both are equally qualified on paper.

The failure that looks like consensus

Here is how a large number of people come to believe something with nobody having checked it.

The first person has weak evidence and acts on it. The second has their own weak evidence, sees the first person's choice, and reasonably treats it as information. Now two people have acted, and the third sees two.

By the fourth or fifth, the observed behavior is stronger evidence than anything an individual holds, so everyone sensibly ignores their own private signal and follows. What accumulates is not evidence. It is a queue, and each person joined it for a good reason.

This is an information cascade, and its two properties are what make it dangerous. It can be started by very little, and it is fragile: because almost nobody contributed independent information, one credible contrary source can reverse the whole thing quickly. A belief that flips overnight was probably a cascade, and the speed of the reversal is the clue.

The formal account comes from economics rather than psychology. Bikhchandani, Hirshleifer and Welch, and independently Abhijit Banerjee, set out in 1992 how entirely rational people stop using their own information once enough others have already acted.

The publishing version is the same shape. One account is written, several outlets summarize it, later writers cite the summaries, and eventually the original is unavailable while its descendants remain. Everything looks corroborated. In fact it is circular reporting, and the number of sources tells you about copying rather than about truth.

Testing whether a crowd knows anything

The practical version of the table is one question, asked of any apparent consensus: what did the people who agree actually see? If the answer is that they saw each other, you have a queue and not a finding.

The counterfeit

Because deference to better information is legitimate, the profitable thing to fake is the appearance of better information. That is where most of the serious deception in this area lives, and it takes a few recognizable forms.

Manufactured agreement, where a large number of apparently independent voices are one interested party. Volume standing in for evidence, where a claim is repeated until repetition is mistaken for confirmation. Borrowed authority, where a credential in one field is presented as relevant to another.

Precision without provenance, where a very specific number is offered because specificity reads as rigour, and no source is given because there is not one. And selective completeness, where everything stated is true and the omitted part is the part that would change your answer.

One test covers all of them. Someone who genuinely knows more loses nothing when you find out how they persuaded you: their information was the argument, and it is still there afterwards. Someone whose case rested on the appearance of consensus loses everything the moment you trace the sources, which is exactly why the tracing is discouraged.

Habits that hold up

When you are the informed one

The obligations run in the other direction too, and they follow from the same distinction between giving evidence and applying pressure set out in social influence basics.

Say where your information comes from, and say it early enough to check before deciding, not after. Give your confidence honestly, even when it is low. Volunteer what would change your mind.

Distinguish what you have seen, what you were told, and what you infer. Listeners cannot make that distinction for you, and they will collapse all three into fact.

And when you do not know, say so. The temptation is to answer anyway, because uncertainty reads as weakness in the moment. It is the single most reliable way to start a cascade you will later have to stop.

Common questions

If everyone believes something, is that not evidence?

It is evidence when the believers reached the view separately. When they reached it by watching each other, the number of believers is a fact about the group and not about the claim.

How can I tell an expert from someone confident?

Look for correction. An expert has a record of being wrong about specific things and adjusting. Confidence without any visible history of being corrected is a presentation style, and it is available to anybody.

Is it wrong to cite a lot of sources?

Not at all, provided they are independent. Five sources tracing to one origin should be cited as one, and saying so openly is more persuasive to a careful reader than the longer list.

What do I do when I cannot check anything myself?

Judge the source rather than the claim: their access, their interests, and whether they face any consequence for being wrong. That is a weaker position than checking, and it is honest about being weaker, which is better than a false confidence built on counting how many places you saw it.