THE PSYCHOLOGY OF HUMAN MISJUDGMENT
Why smart people make terrible decisions
The Bridge That Learned to Dance
My first idea: somebody was careless. Cheap steel, or a sum done wrong. That is what I expect when something new goes wrong on day one.
The bridge was strong enough the whole time. Nothing about it was weak. The problem was a question nobody had asked: what do people do when the floor moves?
A competition is held for a new footbridge over the River Thames in London.
The Millennium Bridge opens. Around 90,000 people cross on the first day.
It is closed after two days because of the sideways sway.
Tests with volunteers identify synchronous lateral excitation.
The bridge reopens with dampers fitted underneath. It has been steady since.
Nobody was lazy. Nobody was stupid. They tested the bridge for weight and for wind, because those are the things bridges are tested for. So: what makes a question invisible?
The Puzzle With One Rule
Siri has had seven yeses in a row and she is about to be told she is wrong. Before you read on: what is the one kind of set she has not tried even once?
The rule was any three numbers that go up. One, two, ninety. Three, four, five. Siri’s rule fitted every test she ran — because she only ever ran tests it would pass.
A hundred years later this has a name: confirmation bias. Looking for what would agree with me, and calling it evidence. Next question: if it feels the same from the inside, how would I ever catch myself doing it?
The Perfect Wrong Mirror
A rocket launch is violent. Things shake loose. That is the boring answer and boring answers are usually right — so I looked for what had come loose.
The instrument used to check the mirror was itself out of true by about 1.3 millimetres. So the mirror was polished, very carefully and very accurately, to the wrong shape.
The primary mirror is ground and polished.
The telescope is launched.
The images are blurred. The mirror is found to be the wrong shape.
An investigation reports that two other tests had shown the error and were discounted.
A servicing crew installs corrective optics. The telescope works as intended.
The Seam Nobody Owned
One team wrote the small nudges from the thrusters in pound-force seconds. The other read them as newton-seconds. One pound-force is about 4.45 newtons. If every nudge is counted as about a quarter of its real size, where does the spacecraft think it is?
The planned closest approach was about 226 km above the surface. The spacecraft came in at roughly 57 km — far below the level anything could survive.
🧭 WHAT THE FIRST TEAM BELIEVED
The file we send out is in the units our own software has always used.
🧮 WHAT THE SECOND TEAM BELIEVED
The file we receive is in metric units, because the mission standard says so.
⚠️ WHAT WAS ACTUALLY TRUE
Both were right about their own half. Nobody owned the sentence between them.
🔧 WHAT CHANGED AFTERWARD
Checks that deliberately cross team boundaries, and a route for a small worry to travel.
The Line That Wasn't
The others in the room were in on it. Only the last person was really being asked. Alone, people got it right nearly every time. In the group, about three quarters went along with the wrong answer at least once.
With one other person disagreeing, going along with the crowd collapsed to a small fraction of what it had been. Not a big campaign. One voice.
This one has a name too: social proof. The useful bit is not that people copy. It is how cheap the cure was. One other voice. So: who in a room is allowed to be that voice?
The Wheel That Changed the Answer
You are selling a bicycle. A stranger asks what you want for it. You could name a price first, ask them to name one first, or say you have not decided. Which do you choose, and what does each one give away? There is no obvious right answer here.
They knew the wheel was random. They watched it spin. It moved them anyway. Knowing about an anchor does not cut the rope.
Name for the file: anchoring. And a rule I can actually use — decide what a thing is worth to me before anybody tells me what they want for it.
Everyone Is in the Better Half
Write down a number you are 98 percent sure the truth falls inside — how many kilometres it is from London to Sydney, say. Give a lowest and a highest. Now: how often do you think a range like that actually contains the answer? Write your guess down first.
Being sure is a feeling. Being right is a measurement. They come from completely different places, and only one of them can be checked.
Nobody in that survey was lying. Almost everyone can think of a moment they drove carefully. Almost nobody can watch themselves being careless — you are busy driving.
You cannot feel your way to knowing how good your judgment is. You have to keep score — and the score has to be written down before the answer arrives.
The Ward Next Door
In one bad year about one mother in ten died in the doctors’ clinic. Next door, with midwives, it was closer to one in twenty-five. My first guess was that the doctors there were simply worse. They were the best-trained people in the building.
The evidence was counted, written down and published. It still took the profession decades to accept it. Not because doctors were wicked. Because the fact accused them.
In Vienna, the clinic staffed by doctors loses far more mothers than the one staffed by midwives. The gap is recorded monthly.
A colleague dies after a scalpel wound received during an autopsy, with symptoms like the mothers'.
Handwashing in a chlorine solution is made compulsory in the clinic.
Deaths in that clinic fall to about one in a hundred.
He publishes the full results, fifteen years after the observation.
Germ theory gives the explanation he never had, and the practice spreads.
He also could not say why washing worked, and he was not gentle with the people he needed to persuade. Being right is not the same as being convincing. Both things are true at once, and neither of them makes anybody a villain.
Ask of any uncomfortable fact: what would it cost me to accept this? The bigger the answer, the harder you should look at your reasons for rejecting it.
The Praise That Seemed to Backfire
Roll two dice and write the total down. Suppose you roll a 12 — the highest there is. Now praise the dice warmly and roll again. What will most likely happen to the total? And what if you had shouted at them instead?
The instructors were watching something real. Their explanation was the problem. The world seemed to punish praise and reward criticism — by arithmetic alone.
Name for the file: regression to the mean. Frightening thought — how many of the rules people live by are arithmetic wearing a costume? Next question: what else does my mind count badly?
The Dangers We Can Picture
In a 1978 study people were asked to judge how often different things kill. Accidents were judged about as deadly as disease. Disease actually killed something like fifteen times more people. Same people, same year, same country.
The mind does not count events. It counts how easily an example comes to mind — and then reports that feeling back to you as if it were a number.
📰 WHAT COMES TO MIND FIRST
The rare, sudden, dramatic event. It is memorable precisely because it is rare.
📊 WHAT THE COUNT ACTUALLY SAYS
The quiet, slow, ordinary cause, many times over, almost never reported.
🧠 WHY THE GAP OPENS
The mind measures how easily an example arrives, not how often the thing happens.
🔍 THE CHECK THAT CLOSES IT
Ask for the number. If nobody has the number, treat the feeling as a guess.
Ten Mysteries, One Shape
I went back through my own board. Bridge engineers who ran every test. Opticians who measured a mirror three ways. Navigators who noticed and said so. Doctors who counted deaths every month. Not one careless person on the whole board.
Every single case had clever, careful, well-equipped people in it. That is not a coincidence in the stories. It is the finding.
🧠 INTELLIGENCE
Solving the problem in front of you, quickly, and being right about it.
⚖️ JUDGMENT
Knowing which problem you are actually looking at, and what you are assuming.
🚫 WHY BEING CLEVER IS NOT ENOUGH
A quick mind builds a better argument for the wrong answer, not a slower one.
✅ WHAT ACTUALLY HELPS
A check you did not invent yourself, applied before you are sure.
The list so far: am I hunting for yeses? What would prove me wrong? Whose number is this? Who else is in the room? What would it cost me to believe this? Is this a cause, or is it arithmetic? Where is the count?
Intelligence and judgment are two different tools. Being good with the first one does nothing whatsoever to sharpen the second.
And Then What?
You have a plan that works brilliantly ninety-nine days out of a hundred. On the hundredth day it loses everything you have. You could run it, run it smaller, run it with a reserve set aside, or not run it at all. What do you choose — and what does your answer assume about how many hundred-day runs you get?
Good decisions do not guarantee good outcomes, and good outcomes do not prove good decisions. What a habit of checking does is improve the odds, and let you survive the times it does not.
A Vienna clinic starts washing hands, and the deaths fall.
A room of people call a short line long, and one dissenting voice breaks the spell.
A three-number puzzle shows people testing only what they hope will pass.
A rigged wheel of fortune moves people's answers about the world.
A survey finds people judge dangers by how easily they can picture them.
Nearly nine in ten drivers place themselves in the better half.
A telescope mirror is polished, accurately, to the wrong shape.
Two units of measurement meet at a seam nobody owned.
A new bridge sways because a crowd walks in time with it.
Something in every one of the ten cases that I have not written down yet. Somebody was paid to build the bridge. Somebody was paid to launch the telescope. Somebody was rewarded for agreeing. In not one single case was anybody paid to go and find the error.
Intelligence does not protect anybody from a predictable mistake. But a mistake that is predictable is a mistake you can go looking for on purpose — and that is a skill, not a gift.
NEXT: THE INCENTIVE TRAP
In every story Siri found, someone was paid to build, to launch, to agree. Nobody was paid to find the mistake. What happens to a number when a person's reward depends on it?