In 1996 engineers in the Kathmandu Valley went through its 643 public schools and asked what a strong earthquake would do to them. The answer was hard to read. In very strong shaking, intensity IX on the scale that measures how hard the ground shakes, they estimated that about two in every three would collapse. Another one in ten would be damaged past using.
Nobody could tell them when that shaking would come. Strong earthquakes had hit the valley several times in two centuries, among them the great one of 1934, and the people who studied it expected another. But it could come next year, or after everyone in those classrooms had grown old. Nobody could say which, then or now.
The engineers, at the National Society for Earthquake Technology–Nepal (NSET), started anyway. From 1997 they strengthened school buildings one at a time, and they did it in a way meant to spread: local masons learned the work by strengthening their own village’s school, so that the skill stayed behind when the engineers left. For ten years the point was mostly to show that it could be done, and could be afforded, to a government, to donors and to parents who had to agree to it. Then the government took the plan up as its own, and between 2012 and 2015 about 300 schools were strengthened.
A week before the earthquake, about fifty earthquake scientists and social scientists from around the world met in Kathmandu to talk about how the city could get ready. One of them, James Jackson of Cambridge University, called the situation “a nightmare waiting to happen.” He said afterward that he hadn’t expected it so soon.
It came at four minutes to noon on Saturday, 25 April 2015, magnitude 7.8. Nearly 9,000 people died in Nepal. Across the country tens of thousands of classrooms were destroyed or left unsafe to use. Saturday is the one day of the week with no school, and a Nepali Times reporter, looking back, wrote that if the earthquake had come on any other day, the death toll would have been unimaginable. In the three districts of the valley, NSET later counted how the 160 school buildings it had strengthened there had done. In 125 the earthquake had left no mark. In the other 35 it had cracked the plaster. Not one needed major repair.
Nobody tossed a coin to decide which schools were strengthened, so this isn’t a fair test in the strict sense. But nobody needed one to see what had happened, and the people who started in 1997 had never had the one thing everyone wants before a big decision. They never knew when. In any one year the dial for “this year” sat low. They acted anyway, and they were right to.
Three questions before acting
The dial never reaches 100 for anything that matters much, and the world doesn’t wait for it to. A school committee, a farmer choosing a seed, a family deciding whether to take a sick child to the hospital tonight or wait for morning: each has to choose with the evidence it has. Not choosing is a choice too. Leaving the school as it was would have been a decision, just one nobody had to sign.
So how sure is sure enough? It depends less on the belief than on what you’re about to do with it. Three questions do most of the work.
How sure are you, honestly? Not certain, since you won’t be, but roughly where on the dial.
What would each kind of mistake cost? There are always two ways to be wrong. You can act when you didn’t need to, or fail to act when you did. Strengthening a school that never feels a big earthquake costs some money and a few months of building work. Leaving one weak when the ground does shake can cost the children inside. When one mistake is far worse than the other, a small chance of the worse one can be enough to act on.
Can it be undone? A choice you can take back is cheap to get wrong. A choice you can’t take back deserves much more care, and so does a harm that can’t be repaired once it has happened.
Then one more question, about the checking itself. Would another look change what you do? Checking is worth doing for as long as it might change the choice and costs less than the mistake it could save you from. After that it is only delay, and delay has its own price.
Doors that only go one way
Jeff Bezos, who built Amazon, put the third question in a picture. In a letter to the company’s shareholders in 2016, he wrote that some decisions are “consequential and irreversible or nearly irreversible”, one-way doors: once you walk through, you can’t get back to where you were. Those deserve to be made slowly and carefully, after talking to people. “But most decisions aren’t like that,” he went on. They are “two-way doors”. If you choose badly, you can open the door again and come back through it, so they should be made quickly, by a person or a small group with good judgment. What goes wrong in big organizations, he thought, is that they treat most doors as one-way, and get slow.
A year later he added a number. Most decisions, he wrote, “should probably be made with somewhere around 70% of the information you wish you had. If you wait for 90%, in most cases, you’re probably being slow.” Being wrong can cost less than you fear, if you notice it quickly and change course. Being slow costs something for certain. He was writing about running a company, but the shape of the idea fits almost anywhere.
It fits a phone, for instance. Forwarding a message is a one-way door. Once it is in a hundred phones, nobody can call it back, and no correction will reach every one of them. Not forwarding it is a two-way door: if it turns out to be true and important, you can send it tomorrow, and very little has been lost. So with a doubtful forward, the two mistakes aren’t the same size, and “not sure yet” is a good reason to wait. Going to the doctor about a pain that worries you is the opposite. A wasted visit costs an afternoon, while waiting too long on something serious may cost far more and can’t always be put right. There “not sure yet” is a good reason to go.
When the harm itself is the one-way door, people call acting early precaution. In 1992 the world’s governments, meeting in Rio de Janeiro, wrote one version into a declaration on the environment. Its Principle 15 says that “where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation.” Each word there is doing work. Serious or irreversible means not for every small worry. Cost-effective means the step taken has to be worth what it costs. It doesn’t say “do nothing until you’re sure there’s no risk”, which would be a rule nobody could live by, since nothing is free of risk and doing nothing has dangers of its own.
Precaution can point toward acting as easily as toward holding back. In 2011 Nepal decided to give chlorhexidine for newborns’ cords across the whole country, the first country to do so, at a time when the main evidence was the trial in Sarlahi, whose range still reached “no effect”. The antiseptic was cheap and simple to use, and a newborn’s death can’t be undone. Later trials in Africa found no fall in deaths there, and the World Health Organization’s advice is now narrower. That doesn’t make 2011 a mistake. It was a reasonable bet on the evidence of the time, made in a way that could be adjusted, and it was adjusted when the evidence moved.
Firmly and lightly
Once a choice is made, it deserves to be carried out properly. Half a strengthened wall holds up nothing. That is the firm part. The light part is keeping your eyes open for the evidence that would tell you to change course, and the best time to decide what that evidence would be is before you start, while you can still think about it calmly.
In May 1996 the New Zealand guide Rob Hall set a turnaround time for the clients he was leading up Everest: anyone not on the summit by then would turn back, wherever they were. It was one in the afternoon by the account of Jon Krakauer, a writer climbing with the team (others say two), and over the weeks before, Krakauer wrote, Hall had told them again and again that sticking to it was the most important rule on the mountain. On 10 May the rule wasn’t kept. Many climbers, from several teams, were still going up well after it. One of Hall’s clients, Doug Hansen, had climbed with Hall’s team the year before and been turned back about 100 meters below the summit by deep snow and the late hour. This time he reached the top after three in the afternoon, and Hall stayed with him. A storm came in. Eight climbers died in it, and Hall and Hansen were among them. What was said up there, and why the rule gave way, nobody can now know.
The rule wasn’t foolish. That is the point of a rule like it: it is made down below, with a clear head, for a moment high on a mountain, tired and very near the top, when wanting the summit is strongest and every reason to keep going sounds good. A turnaround time is a promise from your calm self to your hopeful self, about the evidence that will mean stop.
The same habit works on the ground. Before you start something you aren’t sure of, say what would make you stop: “If I haven’t found the original source in ten minutes, I won’t share it.” “If the scheme can’t show me its registration, I won’t pay.” “If the fever hasn’t come down by morning, we go to the clinic.” Then, when the moment comes, you don’t have to argue with yourself. You only have to look.
People have been reaching for this for a long time. In 1650, shortly before his army fought the Scots at Dunbar, Oliver Cromwell wrote to the leaders of their church: “I beseech you, in the bowels of Christ, think it possible you may be mistaken.” Every person who has ever been sure has needed that sentence, the man who wrote it included. More than three centuries later the statistician Dennis Lindley named a rule after him. Keep certainty, 0 or 100 on the dial, for things that follow from logic alone, such as two and two making four. Everything else, however unlikely, keeps a sliver of a chance, because a belief held at 100 can never be moved by anything. He was willing to give even the moon being made of green cheese a very small chance.
So the two halves go together. Act on your best estimate as if you meant it, because you do. Hold the estimate as a number, not a badge, so that when the evidence moves, the number can move with it.
What to keep
Underneath every page so far has been one question, asked in many different ways: what is this resting on? Here it is, cut down to the questions worth keeping on a phone, each with the chapter where it’s explained.
- What is it resting on? One claim repeated a thousand times is still one claim.
- How sure am I, as a number? And will I keep score?
- What exactly is claimed, and what holds it up?
- Who is saying it, and what do others say about them?
- Out of how many?
- Compared with what?
- What would show it wrong?
- Is it true, before I pass it on?
- Do I want it to be true?
- What would being wrong cost, and can it be undone?
You won’t need most of them most days. Most of what you hear is roughly true, nearly everything you know you were told, and a person who stops to check every sentence never gets anything done. That is no better than believing everything, only slower. The questions are for the moments that count: a forward about to go to every phone in the family, money about to leave a pocket, a remedy about to replace a medicine, a crowd about to go into the street.
That last one is where this book began. In December 2000 people in Kathmandu were completely sure of a sentence nobody had heard, and they acted on it in a way that couldn’t be taken back. Four years earlier, in the same valley, the engineers looking at its schools could never be sure of the one thing they most wanted to know, which year the earthquake would come. They acted anyway, slowly and carefully, in a way that could only make things safer, and in 2015 their buildings stood.
So how sure should a person be? Exactly as sure as the evidence allows, and no surer. And then sure enough for what you are about to do. That is a different amount every time, and finding it is most of what thinking clearly is for.
Found a mistake? Please tell us.
