A rigged wheel and a question about Africa
In the early 1970s, two psychologists sat people down in front of a wheel of fortune — the big carnival kind, numbers painted around the rim. The volunteer gave it a spin, watched it clatter to a stop, and wrote down the number it landed on. Perfectly ordinary, except for one detail the volunteers never learned: the wheel was rigged. It only ever stopped in one of two places — 10 or 65.
Then came the real question: what percentage of countries in the United Nations are in Africa?
Nobody in the room believed a spinning wheel knew geography. The number was visibly, insultingly random — a prop, a distraction, a thing that had just landed in front of them. And yet:
- People who watched the wheel stop on 10 guessed, on average, about 25%.
- People who watched it stop on 65 guessed, on average, about 45%.
Read that again. A number produced by a rigged carnival wheel — a number the participants generated with their own hands and knew to be meaningless — shoved the average estimate 20 percentage points in the direction it pointed. The truth, by the way, sits around 28%. But the truth barely got a vote. The wheel voted louder.
That is the whole course in one experiment. Welcome to anchoring.
What just happened: the lens
Anchoring is the tendency for the first number you encounter to become a starting point — an anchor — that your final estimate stays stuck close to, even when that number is arbitrary, irrelevant, or transparently planted. You don’t weigh the anchor and reject it. It quietly reaches into the estimate and pulls.
The finding comes from Amos Tversky and Daniel Kahneman in 1974, in the paper that launched the whole heuristics-and-biases program. Anchoring is one of the three founding shortcuts they named (the other two, availability and representativeness, get their own courses on this ladder). And it is not a fragile lab curiosity. Anchoring is among the most replicated findings in all of psychology. It survives when you warn people it’s coming. It survives when you pay them for accuracy. It survives when the people doing the estimating are experts on the exact thing being estimated — judges, real-estate agents, negotiators. Telling someone about the anchor and then showing them one still moves their number.
If you took the earlier courses, you already know why this should bother you. In thinking in probabilities, a good estimate is supposed to come from the evidence — your prior, updated by what’s actually informative. Anchoring smuggles a fake data point into that process: a number with zero information content gets treated as if it were a clue. And from incentives, you know that paying people for accuracy usually sharpens them up. Anchoring shrugs at the money. That combination — immune to knowledge, immune to warning, immune to reward — is exactly what makes it worth a whole course.
Before you read — take a guess
A friend is guessing the price of a used car. Right before she answers, you casually mention that your phone battery is at 88 percent. She knows the battery has nothing to do with cars. What's the most likely effect on her guess?
The second half: adjustment, and where it stops
Here’s the piece that makes anchoring feel almost reasonable from the inside. When you get an anchor, you don’t just parrot it back. You adjust — you start from that number and move toward what you actually believe. Saw 65, but you suspect Africa isn’t that dominant? You slide the number down. Saw 10, but that feels too low? You nudge it up. So far, so sensible.
The problem is where the adjusting stops. People adjust away from the anchor, and then they stop too early — the moment their estimate reaches the near edge of the range that feels plausible, they quit and lock it in. Tversky and Kahneman called this insufficient adjustment: you travel toward the truth, but you run out of momentum long before you get there, and you land closer to the anchor than to reality.
Picture the whole thing as tug-of-war between the anchor and the truth, where the anchor is holding a rope tied to your final answer. You pull toward the truth. You make progress. And then — for reasons we’ll dissect in lesson 2 — you decide “close enough” and let go while the anchor is still winning.
Now watch what happens with two people. Give one the anchor 10 and the other the anchor 65. Both adjust inward, toward the same truth in the middle. Both stop short. The first lands a little above 10; the second a little below 65. They end up far apart — and the entire gap between their two answers is manufactured out of two numbers that meant nothing. Two rigged spins of a wheel become a real, measurable disagreement between two thoughtful people.
The one-sentence version
The first number you meet sets a starting line, you drift toward the truth but quit too soon, and you finish nearer to a number that was never worth listening to.
What this course covers
Four teaching lessons, then a graded final exam. Each lesson adds one layer to the lens.
1 — The Anchor and the Adjustment
The precise definition, the wheel-of-fortune experiment up close, and the two-part machine: the anchor that grabs and the insufficient adjustment that leaves you stranded near it. We’ll also meet coherent arbitrariness — the unsettling finding that people’s answers can be internally consistent (a bigger thing costs more than a smaller thing) while the whole price ladder is nailed to an anchor pulled from thin air.
2 — Why the Anchor Won’t Let Go
The mechanism, or rather the two competing mechanisms: classic anchor-and-adjust versus selective accessibility (the anchor makes anchor-consistent facts spring to mind, so the evidence itself gets biased before you even reason about it). This is why even an absurd anchor — was Gandhi older or younger than 140 when he died? — still tugs your number.
3 — Anchors in the Wild
Where this costs real money and real years: the negotiator’s first offer, the price tag ending in 9, the appraiser who “independently” values a house near its asking price, and the courtroom where a prosecutor’s demanded sentence — or a number rolled on dice — moves the judge’s ruling.
4 — Defusing the Anchor
The countermeasures that actually help: consider the opposite, drop your own anchor before the other side drops theirs, interrogate whether the number is even relevant, and think in ranges instead of a single point. Not a cure — anchoring has no clean cure — but real friction against the pull.
Final Exam
A graded run through the whole course, one question at a time. Once you answer a question it locks — no going back, no retries — and you see your score only at the end. Treat it like the real thing.
Two analysts estimate the same company's value. One is first shown a random figure of 20 million; the other, a random figure of 200 million. Both know the figures are random and both adjust away from them. Why do their final estimates still end up far apart?
How to use this course: guess before you peek
One rule will make everything here stick harder: make your own guess before you read the answer. Every experiment in this course has a punchline, and the punchline only lands if you’ve committed a number first. So when a lesson says “how old was Gandhi?” or “what would you offer for this house?” — actually answer, out loud or on paper, before your eyes drift to the reveal. Feeling your own estimate get yanked is the fastest way to learn to feel the yank in the wild.
Next up: lesson 2 pries open the machinery. If a random wheel can move a number, what exactly is happening between the anchor landing in your mind and your answer coming out — and why can’t you just decide to ignore it? Turns out the anchor may have already rigged the evidence before you started reasoning. Let’s find out why the anchor won’t let go.