You are hungry in an unfamiliar city, standing between two restaurants. One has a queue snaking out the door. The other is empty. You know nothing else — same prices, same photos in the window, no reviews on your phone. Which do you pick?
Almost everyone picks the busy one, and here is the uncomfortable thing: that is the rational choice. The queue is a stack of signals. Each of those people knew something you don’t, chose this place, and their choice is evidence. Copying them is a sensible bet. But now notice what your choice does to the person arriving after you. They see a queue one longer — partly because of information you never actually had. You copied the crowd, and in doing so you made the crowd look wiser than it is.
That loop — rational people copying rational people until nobody is adding any new information — is an information cascade, and it is one of the most powerful and least intuitive models in the whole latticework. It explains why bestseller lists are self-fulfilling, why some apps and songs and stocks run away from equally-good rivals, why standing ovations sweep a theatre, why scientific fields chase the same fashionable idea for a decade, and why a confident crowd is so often confidently wrong.
What you’ll actually walk away with
By the end you’ll be able to tell three things apart that look identical from the outside — a crowd that has pooled its knowledge, a crowd that is merely conforming to social pressure, and a crowd caught in an information cascade — and you’ll understand why the third kind is rational for each person yet fragile and often wrong for the group. Here’s the map:
- The urn game — the canonical experiment, worked with real Bayesian numbers, where you can see exactly when a rational person should ignore their own evidence. This lesson hands you an interactive line of deciders you drive yourself.
- Rational yet fragile — why cascades aggregate almost no private information, and why that makes them shatter and flip at the first credible nudge.
- Cascades in the wild — the real faces: queues, bestseller lists, viral hits, bank-run links, research and citation bandwagons, and reputational herding.
- Cascades, common knowledge & critical mass — how a cascade relates to the models beneath it, and how a public signal can start or break one.
- Breaking a cascade — where the model lies to you, and the cure: keeping independent signals alive so a crowd can genuinely pool what it knows.
The one-sentence version
An information cascade is what happens when it becomes rational to ignore your own information and copy the crowd — so the crowd stops learning, and a whim can lock everyone onto a choice that may well be wrong.
Before you read — take a guess
Before we start — take a guess. What is the defining feature of an information cascade (as opposed to plain peer pressure)?
A quick taste of the idea
Here is the whole model in miniature. Imagine a line of people guessing whether an urn is mostly-blue or mostly-red. Each person peeks at one ball (their private signal), then announces a guess out loud — but everyone can hear all the earlier guesses. The first person just says what they saw. The second, if they saw the same colour, agrees; if they saw the opposite, it’s a toss-up. But suppose the first two both guess “blue.” Now the third person draws a red ball. What should they say?
If they are thinking clearly: “blue.” Two independent “blue” guesses outweigh their single “red” ball. So they say blue — and here’s the trap — their guess now tells the fourth person nothing, because everyone can work out they’d have said blue no matter what they drew. From person three onward, every guess is just an echo. The crowd has stopped gathering information, even though it looks more and more confident. And if those first two happened to be wrong, the entire line marches off a cliff together.
Where we're headed
That frozen “everyone-says-blue” line is the entire course in miniature. Each person is rational. The herd is fragile — it rests on two early draws and pools almost nothing else — which is why a single new public signal, or one credible contrarian, can shatter and reverse it in an instant. The rest of the course is this idea made precise, then let loose on markets, fashions, science, and crowds.
In the urn line, why does the third person's guess of 'blue' (after drawing red) tell the fourth person nothing?
How to use this course
Every lesson opens with a quick guess, teaches the idea through a concrete story and — where it helps — real numbers, then checks that it stuck. Don’t skip the guesses: trying to answer before you know is one of the most reliable ways to actually remember. There’s a little arithmetic here (the urn lesson walks the Bayesian steps by hand, no formulas memorised), and one lesson hands you an interactive line of deciders whose signal quality and public signals you control.
This is an expert-tier course, so it assumes you’ve met a few earlier models — you’ll get far more from it if you’re comfortable with Bayesian updating (revising a belief as evidence arrives), common knowledge (what everyone knows that everyone knows), and critical mass (the threshold of early movers that tips the rest). When you’ve finished the five teaching lessons, a graded final exam pulls it all together. It’s one-way — once you submit an answer it’s locked — so treat the practice questions along the way as exactly that: practice.
One habit to build as you go
Whenever you see a crowd converge — a hot restaurant, a runaway bestseller, a stock everyone is piling into — train yourself to ask not “what do they all know?” but “how much of this is real information, and how much is just people copying the person in front?” That second question is the one this course teaches you to answer.
Ready? The next lesson builds the urn game carefully — the exact moment a rational person should ignore their own eyes — and shows why that moment is where the crowd stops learning.