Part V · Chapter 10 · Behaviour · lesson 3

The mechanism behind the meme

11 min

Builds on Why investors trail their own investments

Eleven people are in a room. Ten of them have been told to say that the shorter of two lines is the longer one. The eleventh has been told nothing. The eleventh can see the lines perfectly well.

You already know what a decent proportion of eleventh people do, because you have been the eleventh person, in a meeting, about something you knew.

The interesting question is not why they cave. It is why caving is usually a good idea.

If forty people are running in one direction, the cheapest available hypothesis is that forty people know something you do not. Acting on that hypothesis costs you almost nothing when they are wrong and saves your life when they are right.

Copying is not laziness. It is a genuinely efficient way to acquire information you have no other way of getting, and every social animal does it. The problem is narrower than "people are sheep": copying works when the crowd's information is independent of yours, and stops working the moment the crowd is also copying.

The cascade

Here is the mechanism, with numbers small enough to hold in your head.

Ten people each have a private, slightly unreliable opinion about whether a thing is worth buying — say each of them is right about seven times in ten. They decide one at a time, in public, and each can see what everyone before them did.

Person one has nothing to go on but their own read, so they act on it. Person two has their own read plus one observation. Person three, if the first two agreed, faces a choice between their own single opinion and two other opinions pointing the other way, and rationally goes with the two.

At which point person three's own information has left the system. They acted on the crowd, so their action tells person four nothing new. Person four sees three agreeing and follows for the same reason. So does everyone after.

Predict

Eight people in a row have now bought. How much private information about the thing is contained in those eight decisions?

Eight people's worth — that is what makes it convincingTwo people's worth, because everyone after the second was copyingNone — group decisions carry no information at allIt depends on how sure each person was

Why money makes it worse

In most settings the crowd is just information. In a market the crowd is also the price.

If everyone buys, the price goes up. The price going up is then read as confirmation that buying was correct, which brings in more buyers, which raises the price again. The feedback runs the same way in reverse. Nothing in that loop requires anybody to be irrational at any individual step, which is why "people were being stupid" is a poor explanation of any particular episode.

Not the same thing

Information cascade

Copying others because their actions are evidence about something you cannot observe directly.

Measured in: How much independent private information the crowd still contains
For example: Choosing the busy restaurant over the empty one on an unfamiliar street.
Chase it alone and: The crowd's confidence grows while its evidence does not, so a wrong first mover can be amplified indefinitely.

Fear of missing out

Acting because other people are getting a reward you are not, regardless of what their actions imply.

Measured in: The size of the gap between their outcome and yours
For example: Buying after a colleague mentions what their holding has done this year.
Chase it alone and: It responds to the outcomes you were shown rather than to any evidence, so it peaks exactly when the winners are most visible.

Which is which? Put each one on a side.

Picking the queue with more people in it at an unfamiliar airport
Buying because three people at a party have made money on it and you have not
Joining a bank queue after seeing a queue form, believing the bank may be in trouble

Fear of missing out is a sampling problem

The cascade is about information. The other half of this is about which outcomes you get to see, and it is a different mechanism with a similar effect.

You hear about the person who bought early. You do not hear about the person who bought late, because that story is not fun to tell and nobody asks. The people around you are not a random sample of people who tried the thing — they are the ones who came out of it with a story.

That is enough on its own to make almost anything look better than it is, and the effect gets stronger the more people are involved, because a bigger crowd produces more extreme winners to hear about. The track after next is entirely about this.

Check

What actually distinguishes a crowd worth following from one worth ignoring?

The size of the crowd — a larger one is more likely to be rightWhether the members formed their views independently of each otherHow confident the crowd soundsHow long the crowd has been growing

The useful version of the question

Naming the bias does not tell you what to do, and the honest position is that following a crowd is often correct. The question that does some work is narrower:

What do the people already in this know that I do not, and how would I check?

If the answer is something specific — they have used the product, they work in the industry, they read the filings — then the crowd is carrying information and copying is reasonable. If the answer is that they are in it because other people are in it, then the crowd is carrying its own reflection, and the number of participants tells you nothing about whether it is a good idea.

Why this is worse online than it was in a villageOptional

A village crowd was limited by how many people you could physically see, and those people mostly had independent information about local conditions.

A feed is not size-limited and is not independent. It shows you what is already being engaged with, which means the visible crowd is selected for having already grown. Both properties of a healthy crowd — bounded size, independent members — are removed by the same mechanism, and it is worth being aware that the mechanism was optimised for something other than the accuracy of your beliefs.

Worth remembering

Copying others is an efficient way to acquire information and works well when their views were formed independently of yours.
In an information cascade, people rationally stop acting on their own information, so the crowd stops accumulating evidence while continuing to accumulate members.
In a market the crowd sets the price, and the price is then read as confirmation, which closes the loop.
Fear of missing out responds to outcomes you were shown rather than to evidence, and the people who lost are usually not the ones telling stories.
The question worth asking is what the people already in it know that you do not, and how you would check.