Two systems: one fast, automatic and associative, one slow and effortful. The shortcuts the fast one takes produce errors that are not random but predictable, which is what makes them worth studying at all. Anchoring, availability, loss aversion, base-rate neglect. I have kept it at the top of this list and I want to be straightforward about why that took some thought. On 14 February 2017 Kahneman publicly accepted a critique showing that the social-priming studies behind chapter 4 barely replicate. He wrote that he had placed too much faith in underpowered studies. He did not withdraw the book, and the heuristics work that won him the Nobel in 2002 is untouched by any of this. So: read it, read chapter 4 with your guard up, and know that ego depletion, money priming and the disfluency effect have gone the same way.

I put this on the list because I could not find it on anybody else's, and a reading list where Kahneman goes unchallenged is not a reading list, it is a press release. Gigerenzer's case is that simple rules of thumb are not defects. They are adaptations, and in the real world, where you never have all the information, they often beat the complicated models. His distinction between risk, where the odds are known, and uncertainty, where they are not, has changed how I think about my own decisions more than anything else here. He also shows that a lot of famous biases quietly disappear when you give people frequencies instead of percentages. He has been fighting this battle for thirty years and it shows in the tone.

Klein went and watched firefighters, nurses and military commanders instead of undergraduates in a lab, and found that experts under time pressure do not weigh options at all. They recognise a pattern, land on the first workable course of action, and run it through in their heads. Expertise turns out to be closer to perception than to analysis. It is interview and observation work rather than controlled experiment, which makes it harder to prove wrong and also harder to prove right. I still think it is the most under-recommended book on this subject.

Being clever and being rational are not the same thing. An IQ test measures how much processing power you have, not whether you bother to use it, which is why very intelligent people make very poor decisions with some regularity. This is the driest book on the list and I nearly left it off. It stayed because it is the academic floor underneath the popular titles, and because anyone who reads three of those and starts feeling expert should read one of these.

If somebody tells me they have a decision to make on Thursday, this is the book I name. Four things that reliably go wrong — looking at too few options, hunting for evidence that agrees with you, deciding in the grip of a short-term feeling, and being too sure — and a four-step process built against them. The single idea worth the cover price is that most bad decisions come from considering too few options rather than from analysing them badly. It synthesises other people's research rather than producing its own, and a few of the studies it leans on come from the same wobbly literature as Kahneman's priming chapter. It is still the most useful book here.

A decision is a bet on a future you cannot see, so how a decision turns out and how good it was are two separate questions. Duke calls the habit of grading one by the other “resulting”, and once you have a word for it you notice yourself doing it constantly. Her prescription is to say your beliefs as probabilities, keep score, and find people who would rather be right than agreeable. One correction, because nearly every page about her gets this wrong: Duke won a World Series of Poker bracelet in 2004 and finished her PhD in cognitive psychology at Penn in June 2023, fifteen years after she left the programme and five years after this book came out. She also works as a paid decision consultant, which you should know while reading her.

The other half of the argument, and the one I have found more use for. Nobody quits early enough, because sunk costs, identity and the fear of looking flaky all push in the same direction, and by the time you finally do the sums you have already spent the years. The tool is kill criteria: you decide in advance what would have to be true for you to walk away, so the decision gets made before you are emotionally committed to it. I find the argument more convincing than the evidence underneath it, and I have used the tool anyway.

Straight decision analysis for people who are not analysts. Define the problem, set out what you actually want, generate alternatives, work out the consequences, then face the trade-offs honestly. Raiffa more or less invented the modern discipline and this is it with the mathematics removed. It has dated. It assumes a person who can list their own objectives and it says almost nothing about the emotional and identity side of a decision. For one important, structured choice with several criteria, I still have not found anything better.

Forecasting is a skill you can learn rather than a gift you have. The people who did it best in Tetlock's tournament thought in fine-grained probabilities, broke big questions into small ones, updated in small steps, and kept score of how they did. The tournament was real: a US intelligence research programme that ran from 2010 to 2015, which Tetlock's team won outright. One update I have not seen anywhere else. A re-analysis published in 2025 found that once you control for how the tournament was designed, the effects of its teaming and training largely vanish, and in places reverse. Good forecasters exist. The book's explanation of what makes them good is shakier than it sounds on the page.

Weather, baseball, earthquakes, poker, elections, the financial crisis. Silver's argument is that most prediction fails because people are too sure of themselves and because they mistake noise for a pattern. Underneath it is Bayesian updating, taught without a single equation. The political chapters have aged, and his own 2016 forecast became the standard stick to beat him with, though he gave Trump a far higher chance than almost anyone else did.

Rare, unpredictable, enormous events drive history; our models price them at close to nothing because they assume the world is normally distributed; and the answer is not to forecast better but to build something that survives being wrong. Putting Taleb and Silver on the same list means recommending two people who think the other's method is nonsense. Read both and pick a side. If Taleb's manner puts you off, Fooled by Randomness from 2001 is shorter and less pleased with itself.

If anything matters, it can be observed, and if it can be observed it can be measured. The point of a measurement is not to remove uncertainty but to reduce it, and Hubbard will teach you to estimate probabilities you can be held to, to work out what a measurement is worth before you pay for it, and to build a rough model. It asks more work of you than anything else on this page, and it is also an advertisement for the author's own consulting method. Both of those are true and neither stopped me including it.

Things now go wrong more often because we fail to apply what we already know than because we do not know it. A good checklist takes the load off memory and, more importantly, flattens the hierarchy in the room, so the junior nurse is allowed to stop the senior surgeon. It is also the best-written book on this page by a distance. And here is the part nobody prints. The evidence it rests on is an eight-hospital study where deaths roughly halved. A much larger study across 101 hospitals in Ontario, published in 2014, found no significant effect at all. My reading is that checklists work where a culture takes them seriously and do nothing where they are simply imposed, which is a harder lesson than the one the book teaches.

Organisations spend their attention on bias, meaning error that leans one way, and ignore noise, meaning the sheer scatter between judgements of the same case. Two judges hand down different sentences for the same crime. Two underwriters price the same risk differently. The same doctor disagrees with herself on a Friday afternoon. Usually the scatter is the bigger problem, and usually it is the easier one to fix. It is long and it repeats itself, and the distinction is really variance against bias, which statisticians have known about for a century. The examples are worth the repetition.

The book to hand somebody who runs meetings. Four ways groups go wrong — amplifying an individual error instead of correcting it, cascading, polarising, and burying the information only one person in the room holds — and then what to do: the leader speaks last, roles get assigned, someone is told to argue the other side. The finding I keep coming back to is that groups who make their judgements independently and then average them regularly beat groups who talk it through together.

Where the word comes from. The Bay of Pigs, Pearl Harbor, the escalation in Vietnam. Janis argues that a tight group under stress will suppress its own doubts to keep everyone agreeing, and end up somewhere worse than any single member would have gone alone. The case studies are superb and still read well. I have shelved it under history rather than science on purpose. Groupthink is one of the most quoted ideas in management and one of the least supported: laboratory studies have repeatedly failed to find it, and several found the opposite, with the most cohesive groups censoring themselves least. Take the vocabulary and the case studies. Leave the certainty.

If you searched for this because of one specific decision — whether to marry, whether to have a child, whether to leave the country — start here. I have looked at a great many lists on this subject and this book is on almost none of them, which I think is the single biggest thing wrong with all of them. Roberts' argument is that these choices cannot be solved with the tools the rest of this page teaches, because you cannot know in advance what the experience will be like or who it will turn you into. A pro and con list forces you to invent exchange rates between things that have none. What he offers instead is to decide by asking what kind of person you want to become. It is an essay, not a method, and some readers close it comforted but no clearer. I was not one of them.

Design thinking pointed at your own working life. Build several plausible versions of the next five years instead of hunting for the one correct path, then test them cheaply through conversations and small experiments. It came out of the authors' Stanford course and it is genuinely good at getting a stuck person to move. Two things to know before you buy it. It is a teaching method rather than a tested intervention, and I could not find a controlled evaluation of it anywhere. And it quietly assumes a reader with room to manoeuvre: money, mobility, nobody depending on them.

Too much choice makes people anxious and regretful rather than free; and the people who insist on the best available option are reliably less happy than the ones who take the first option that is good enough. Half of this survived and half did not, and I would rather tell you which. The choice-overload finding, the famous jam experiment, has largely failed to replicate, and a meta-analysis in 2010 put the average effect at close to nothing. The distinction between the two kinds of chooser has held up much better, and it comes straight out of Herbert Simon. That second half is why the book is still here.

Everything above is a footnote to this. Simon's point was that nobody can optimise, because nobody has all the information, all the time or all the processing power. So we satisfice: we search until something clears the bar, and then we stop. He won the Nobel in economics for it in 1978. It is a 1940s academic book about public administration and it reads exactly like one. Read a chapter. I am not going to pretend I read it cover to cover.

This turns up on nearly every list of decision-making books and I have labelled it rather than quietly dropped it, because the mistake is worth naming. Nudge is about designing the situation in which other people choose: defaults, what gets made obvious, how an option is framed. It is a fine book and it will not help you decide anything yourself. It also has a problem. A re-analysis published in 2022 found the overall nudge effect falls from substantial to nearly nothing once you correct for the fact that studies finding nothing tend not to get published. The original authors dispute it. I have yet to see another list mention it at all.