The Best Books I Read in 2022

This is the shortest list in the series, and it comes with an asterisk I should hand over before anything else.

2022 was the year I switched from buying books to borrowing them. And a library card changes what you read. A good chunk of what I checked out was reference material rather than the kind of book you write up in December. A stats primer. A hiking guide for a trip that summer. Useful books. Not write-up books. I also reread Atomic Habits, which already had its entry back in 2020, so I am not writing it up twice.

So three books earned the shelf. And here’s what I noticed only when I put them side by side in December. All three are arguing about the same question, and they disagree with each other.

The question is how fast you should decide. Gladwell says two seconds is often plenty. Harford says stop and count. Pollan says the speed was the wrong variable to be looking at.

So which of them is right? All three, and the rest of this post is me working out when.

Three books. Three speeds. And one of them is right for whatever you are looking at this morning.

Blink by Malcolm Gladwell

Gladwell published this in 2005, and his term for it is thin-slicing. And your mind takes a very narrow slice of experience and reaches a conclusion in about two seconds, running on pattern recognition that cannot show you its working.

And the opening case is the Getty kouros. The J. Paul Getty Museum spent fourteen months on scientific analysis before buying a Greek statue for around nine million dollars. But then the art historians started arriving.

One of them said the first word into his head was fresh. That is a strange word for an object supposed to be two and a half thousand years old. Another felt something wrong in the time it took to look up from the floor. A third got stuck on the fingernails.

So the lab said authentic and the two-second reactions said otherwise. The two-second reactions were right.

And the other case he leans on is John Gottman. He claimed he could watch a couple argue for fifteen minutes and then call whether the marriage would survive fifteen years. Reportedly right about nine times in ten. Fifteen minutes.

So does thin-slicing ever go wrong? Constantly, and Gladwell gives the failure its own name. The Warren Harding error, after a man who looked so exactly like a president that enough people stopped checking whether he could do the job.

But Gladwell refuses to sell it as magic, and that is why the book holds up. He spends as much time on thin-slicing curdling into bias as on thin-slicing working. The Amadou Diallo shooting sits in the same volume as the kouros, doing the opposite job.

Now here’s the part that stings for anyone who works in data. An expert’s snap judgement is compressed experience, and it is close to impossible to audit. You cannot put it in a design doc. You cannot hand it to the next engineer. And I have watched a good architect say this will not scale, in about four seconds, and be right, with nothing to point at afterward when somebody asked why.

So can you teach it? Barely. That is the honest answer, and it is why apprenticeship survives in fields that have every reason to have replaced it by now. You sit next to somebody for two years and absorb the pattern by proximity, because there was never a document.

That’s the first speed. Two seconds of pattern will beat two hours of reasoning, as long as you actually earned the pattern.

How to Change Your Mind by Michael Pollan

Pollan published this in 2018, and it traces an arc most of us only know a third of. Serious psychiatric research through the 1950s and 60s. Then prohibition, and roughly thirty years with the work more or less stopped. Then the resurgence, with Roland Griffiths running psilocybin studies at Johns Hopkins from 2006 and producing results that were hard to wave away.

And the Griffiths numbers are the ones that made the field hard to ignore. Volunteers were followed up more than a year later, and a majority put the session among the most personally meaningful experiences of their lives, in the bracket people normally reserve for the birth of a first child.

So he reports it the way he reported food, which means going and trying the thing rather than watching from the gallery.

But the part I kept turning over is the mechanism, and it is why this book belongs next to the other two. And the brain has a default mode network, the set of regions that hums along when you are doing nothing in particular. Self-narrative. Rumination. The running story of you.

So what does the imaging suggest? That psilocybin quiets that network down, and that connections then form between regions which normally keep to themselves. Robin Carhart-Harris’s group described a brain temporarily talking to itself along routes it usually leaves alone, and the therapeutic value seems to live inside that window.

Here’s the image I have kept. A hillside after years of rain, with deep channels cut into it. And water runs down the same grooves every time, because that is what grooves are for. What the research describes is a temporary flattening, where the water gets to find new lines before the grooves set again.

And that reframed the other two books for me. Your two-second judgement runs down a channel. The channel took years to cut, and you were mostly not present for the cutting.

So how settled is any of this? It is a young field, with small samples and a lot of enthusiasm, and Pollan stays careful in exactly the places the enthusiasts do not. Read it for the mechanism. Hold the conclusions loosely.

That’s the second speed. Speed is downstream of the channel, and the channel is the thing worth arguing about.

The Data Detective by Tim Harford

Harford published this in 2021, under a better title in the UK, and it is ten rules for reading statistics.

And rule one is this. Notice your emotional reaction to a number before you do anything else with it. There is no arithmetic in that rule at all, and it comes first, because Harford’s claim is that the mistake almost always happens upstream of the maths.

So who is he actually worried about? Good-faith people, mostly. He spends far less time on the ones cooking numbers than on the ones misreading them honestly. That is the more common failure. It is also much the harder one to fix. Being lied to is somebody else’s fault. Being convinced too easily is yours, which is exactly why nobody wants to look at it.

So I came to this one sideways. I had a stats reference out of the library that year for something unrelated, and this turned out to be far the more useful of the two, which is a mildly embarrassing thing for somebody who works with data all day to admit in public.

And he has a name for the specific failure, which is premature enumeration. Counting a thing before you have agreed what the thing is. It sounds like a joke and it is the reason half the dashboards I have ever reviewed were arguing past each other.

So the rule I use most is the one about asking what is being counted. A headline number is a measurement. A measurement had a definition. A definition was written by a person. For a purpose. On a particular Tuesday, under a deadline, by somebody who had to pick one edge case over another and then went to lunch. Unemployment counts people looking for work. Engagement counts whatever the team shipping the dashboard decided engagement would mean.

So how long does that actually take? About ten seconds, most of the time. And the ten seconds are the whole discipline, because the number has already produced a feeling in you by then and the feeling is what you are really pausing.

That’s the third speed. Slow down long enough to ask what a number is made of, and about half of them stop being frightening.

And here’s where the three of them part company, which is the reason I have kept them together. Gladwell’s expert should trust the two seconds. Harford’s reader should distrust the two seconds. Pollan’s patient should ask how the two seconds got built in the first place.

But they are all right. So how do you know which one applies to you this morning? It comes down to whether you earned the pattern or inherited it, and most of us have never stopped to ask which.

That’s my 2022 list. Three books, one short year, and a library card that turned out to change more than my spending.

A fast judgement is cheap to make and expensive to check. The work is knowing which of the two you are doing.