DEEP READING · READ 918
Think Again: The Power of Knowing What You Don't Know · Adam Grant · 2021
We spend our lives practicing how to think and almost no time practicing how to rethink — and in a world that changes faster than we learn, knowing what you don't know, and being willing to be wrong, is worth more than knowing a little more. The book isn't offering you a new set of opinions; it's offering a skill — holding your opinions more loosely — and three uses for it: changing your own mind, helping other people change theirs, and building a place where changing your mind is allowed.
Adam Grant is an organizational psychologist at Wharton and one of the most fluent translators of academic research into public language working today (his earlier books: Give and Take, on givers, takers and matchers, and Originals). Think Again, published in 2021, is the most "meta" of them — where his earlier books taught you a particular conclusion, this one is about how to hold all of your conclusions. It sits squarely in the behavioral-science trade-book tradition (Kahneman, the Heath brothers, Duhigg), with that tradition's characteristic strengths and weaknesses: readable, densely exampled, actionable — and also selective in its cases and inclined to smooth out complicated research, which the criticism section below takes seriously.
This is the book's skeleton. When we think and argue, Grant says, we tend to slip into one of three professional modes. Preacher: your convictions are challenged, so you defend and proselytize the sacred. Prosecutor: you spot a flaw in someone else's reasoning, so you marshal evidence and win the case. Politician: you want the approval of an audience, so you campaign, court and tailor. All three are familiar, and all three share one hidden premise: you already have the answer, and the only remaining question is how to get it accepted.
His fourth mode is the scientist — not a job but a default move: treat what you believe as a hypothesis to be tested, go looking for the evidence that would kill it, and let the result change you. The difference isn't whether you hold opinions; it's the ownership status of the opinions in your head. Preachers own their views. Scientists rent them.
| Trigger | What you're doing | The cost | |
| Preacher | A sacred belief is offended | Defending, proselytizing | The belief becomes part of you |
| Prosecutor | Someone else is wrong | Building a case, refuting | You win the argument, lose the person |
| Politician | You want approval | Campaigning, accommodating | Views drift with the audience until you don't know your own |
| Scientist | You hit uncertainty | Forming hypotheses, hunting disconfirmation | Slow, uncomfortable, reads as indecisive |
The first three assume the conclusion is settled; only the fourth treats it as a variable.
This isn't only a metaphor. Grant cites a randomized trial with Italian entrepreneurs (Camuffo and colleagues): one group was trained to think like scientists — treat the business plan as a hypothesis, design cheap experiments to falsify it, let the data decide whether to persevere or pivot — while a control group got conventional training. A year later the trained founders had substantially higher revenue and were more likely to pivot decisively when the original idea was wrong. The mechanism is plain: people who treat a plan as a creed explain bad data away; people who treat it as a hypothesis treat bad data as a result.
What it changes: in a meeting or an argument you can now ask yourself, mid-sentence, which one am I being right now? Simply being able to name the mode takes heat out of the system. And you start noticing that the hardest one to catch is the politician, because it disguises itself as open-mindedness — you appear to be listening to everyone while actually searching for the formulation nobody will object to.
Most advice about confidence slides along a single line: more of it, or less of it. Grant splits it into two independent quantities — confidence in your ability, and confidence in your current answer. Low on both is one version of impostor syndrome; high on both is the armchair quarterback. The cell he wants is confident humility: believing you're capable of figuring it out while doubting the answer you currently hold. These don't contradict each other — it's precisely because you trust yourself to learn that you can afford to admit this draft is wrong.
The companion habit is taking pleasure in being wrong. Grant describes visiting Daniel Kahneman and pointing out that one of his arguments didn't hold; Kahneman's response wasn't defense but visible delight — in paraphrase, wonderful, I was wrong, now I'm less wrong than I was. Which yields a genuinely useful psychological move: detach your opinions from your identity. Define yourself as "someone who takes evidence seriously and updates," not as "someone who believes X." The first survives being refuted intact; the second experiences refutation as a slap.
Grant borrows the psychologist Anthony Greenwald's name for the opponent: the totalitarian ego — the dictator inside the head whose job is to screen out any information that threatens your self-image. It isn't a pathology; it's the factory setting. Rethinking is therefore not an attitude but a movement made against physiological resistance.
One more useful split: detach your present self from your past self. Most people who won't change their position aren't defending today's view; they're defending the person they were three years ago. Reframe the change as evidence of growth rather than betrayal — if you hold exactly the same opinions you held three years ago, the one thing that's certain is that you haven't learned anything in three years.
Everyone has a support network — the people who comfort you when things go badly. Grant argues you also need a challenge network: a small group of people you trust to take your work apart. The two rarely overlap, and the challenge network has to be recruited deliberately, by name. Wait passively for criticism and the only criticism you'll get is from the people who don't care how you feel.
Who qualifies? His answer is counterintuitive: disagreeable givers. Two separate things are in play — agreeableness is a style (how gently you speak), while giving is a motive (whether you want to help you or beat you). Of the four combinations, the valuable one is the person who is abrasive in manner and genuinely invested in your getting it right: they're blunt because they refuse to take the easy exit of saying "looks good." The dangerous one is the agreeable taker — all warmth to your face, and then acting on their own interest.
The organizational version is another distinction worth having: task conflict is disagreement about what to do and how, while relationship conflict is personal friction. Research finds the first is often productive, particularly early in creative work, while the second is almost uniformly corrosive — and most teams' real problem is that the two get fused. The moment "your proposal has a flaw" is heard as "you are a flaw," task conflict degrades into relationship conflict and nobody raises anything again. High-quality disagreement doesn't come from being more polite; it comes from a culture that keeps "criticizing the plan" and "rejecting the person" visibly separate. Grant also cites Charlan Nemeth's work showing that authentic dissenters outperform assigned devil's advocates — when everyone knows the objector is only playing a part, nobody is actually moved.
This is the heart of part two, and the most counterintuitive section. Our default persuasion strategy is more facts, tighter reasoning, additional evidence — which produces what Grant calls the logic bully, someone who rolls over you with an airtight chain of argument. The usual result is that your opponent concedes out loud and hardens inside.
He draws on classic negotiation research by Neil Rackham, who recorded expert and average negotiators at work. The differences are sharp: experts spend far more time looking for common ground, ask many more questions, and offer fewer reasons. The last one is the least intuitive. Average negotiators pile on justifications, assuming each one adds persuasive weight — but arguments dilute each other. Give five reasons and your counterpart only has to break the weakest one for the whole stack to look broken. Fewer and stronger beats more and mixed. Experts also rarely enter the defend-attack spiral, where each rebuttal drags the conversation further from the actual question.
The constructive technique is motivational interviewing, developed in clinical psychology for addiction work, with three moves: ask open questions, listen reflectively (say back what you heard, to check that you got it), and affirm the other person's autonomy ("whether to change is entirely your call"). The logic behind it: people don't resist change; they resist being changed. Once autonomy feels threatened, defending the position becomes a way of defending the right to decide.
Grant profiles a specific practitioner: Arnaud Gagneur, a neonatologist in Quebec who works with parents hesitant about vaccinating. He doesn't recite data or scold; he sits down on the maternity ward, asks what worries them, listens, and acknowledges that the decision is theirs — and studies of the approach found meaningfully higher subsequent vaccination rates. What he wins is not the argument; it's the other person going back through their own reasoning. (Honestly: effects for this family of interventions are typically moderate rather than miraculous — see the criticism below.)
What it changes is immediate: next time you want to persuade someone, swap "I need to prove I'm right" for "I want to understand why they think that." A usable test — if the other person never once says "huh, I hadn't thought of that," you were probably performing rather than persuading.
Binary bias is our habit of compressing continuous, many-layered questions into two opposing camps: for or against, true or false, us or them. It's economical, and it empties out the middle — where nearly all the information in a real question lives.
The antidote is complexifying: deliberately showing the range of positions and qualifications inside a question. You'd expect this to confuse people and reduce engagement; the evidence points the other way. When people see there are more than two available views, defensiveness falls and curiosity rises, because they no longer have to pick a side before they're allowed to speak. Grant's illustration is climate coverage: frame it as believers versus deniers and identity activates instantly, whereas presenting the real disagreements among scientists about pace, mechanism and remedy leaves people room to step in.
He offers two concrete moves. First, keep the qualifiers in your writing and speech — the conditions under which a claim holds, the cases where it doesn't. You lose a little force and buy a lot less resistance. Second, counterfactual thinking: invite people to imagine what they would believe had they been born elsewhere and raised by a different family. That converts "my position is who I am" into "my position is a product of my circumstances," and stereotypes loosen accordingly.
Worth noting: Grant does something admirable in this material. He recounts Sherif's famous Robbers Cave experiment — two groups of boys pushed into hostility, then reconciled through shared goals — but also the researcher's failed attempt the previous year, where the boys refused to turn on each other and instead united against the researchers, wrecking the design. Reporting the study that didn't work is itself a demonstration of the book's thesis.
The book opens with the Mann Gulch fire in Montana in 1949. A crew of smokejumpers is overtaken by a wildfire, and the foreman, Wagner Dodge, does something nobody has seen before: he lights the grass in front of him, burning a patch of ground to ash, and lies down in it so the main fire passes around him. He survives. When he calls to his crew to join him, none of them do; most keep running uphill, and thirteen die. Several of the dead were still carrying their heavy tools and packs. The organizational theorist Karl Weick summed the scene up as a failure to "drop your tools" — because the tools were their professional identity. To drop them was to stop being a firefighter.
The metaphor runs through the book. Escalation of commitment is its behavioral form: the more time, money and self you've sunk into a path, the harder it becomes to concede that the path is wrong, so you raise your bet (sunk cost is only part of it; the stronger driver is not wanting to admit error). Identity foreclosure is its biographical form: locking into a self-description too early — "I'm a med student," "I'm an engineer" — after which every new piece of information gets recruited to maintain it. Which is why Grant goes after the question every child is asked: "What do you want to be when you grow up?" presupposes that a person is a noun, and that you only get one.
His remedy is plain and workable: schedule career checkups, the way you schedule dental ones — once or twice a year, sit down and ask whether the work you're doing is still aimed at the problem you actually want to solve, and whether you're persisting or just unable to let go. He also files an honest amendment to grit: perseverance on the right path is resilience, and on the wrong path it's stubbornness — and the two feel exactly the same from the inside. The only way to tell them apart is to stop periodically and re-examine the goal itself, which is precisely the move that "never give up" talks you out of.
The last layer is organizational. Grant leans on Amy Edmondson's psychological safety: whether people on a team can admit mistakes, raise problems and say "I'm not sure" without fear of humiliation. The clarification he repeats matters — psychological safety is not niceness, not comfort, and emphatically not lower standards. It is the precondition that lets high standards actually be enforced, because bad news can travel upward.
Edmondson's best-known finding is an inversion. Studying hospital teams, she found that the units rated as better reported more medication errors — which looked at first like better teams making more mistakes. Closer observation showed they weren't erring more; they were reporting. The weaker teams' errors simply disappeared into silence. Where there's no psychological safety, "error rate" isn't measuring errors at all; it's measuring fear.
The paired mechanism is process accountability: evaluating the quality of how a decision was made, not only how it turned out. The reason is straightforward — outcomes are heavily contaminated by luck, so scoring people purely on results teaches them to avoid risk and to construct flattering stories afterward. The process (what was known at the time, which options were weighed, what would have counted as evidence to abandon the plan) is the part that can actually be learned from. Grant also suggests retiring the phrase "best practices" in favor of "better practices" — "best" is a word that announces rethinking is now closed.
Three parts, which are the same move at three radii: change your own mind → help others change theirs → build a group that keeps changing.
How is any of it established? Three kinds of evidence, interleaved: experiments and meta-analyses from social and organizational psychology (psychological safety, task conflict, motivational interviewing, dissent research); randomized trials such as the Italian founders study, which is the hardest evidence in the book; and a large volume of narrative cases (Mann Gulch, BlackBerry's collapse, the vaccine conversations, teachers and coaches). The strength is a clean three-level structure where every claim converts into an action. The weakness is that the cases were selected to illustrate conclusions already reached, and not every piece of social psychology it draws on has survived the last decade's replication scrutiny.
1. We've practiced how to think all our lives and almost never how to rethink — and where change outruns learning, rethinking is worth more than knowing a little more.
2. In argument we slip into three roles — preacher (defending the sacred), prosecutor (attacking the flaw), politician (courting approval) — all of which assume the conclusion is settled; the missing fourth is the scientist, who treats belief as a hypothesis and hunts for what would kill it.
3. Intelligence doesn't guarantee rethinking and can make it worse: the better you are at reasoning, the better you are at manufacturing support for what you already believe.
4. The target is confident humility — trusting your capacity to work it out while doubting your current answer; the two are compatible precisely because trusting yourself to learn is what lets you disown this draft.
5. Detach opinion from identity: be "someone who takes evidence seriously," not "someone who believes X" — and note that if your views are identical to your views three years ago, the only certainty is that you've learned nothing in three years.
6. Alongside the support network that comforts you, build a challenge network: a few people you personally recruit to take your work apart, ideally disagreeable givers — rough in delivery, genuinely invested in your getting it right.
7. Don't be a logic bully: expert negotiators spend more time on common ground, ask more questions, and give fewer reasons — arguments dilute one another, and one strong reason beats five weak ones.
8. People don't resist change; they resist being changed — so effective persuasion is asking, reflecting back, and explicitly handing the decision back, until the conclusion comes out of their mouth.
9. Don't flatten a spectrum into two teams: showing the range of positions and their qualifications lowers defensiveness and raises curiosity, because nobody has to pick a side before speaking.
10. Drop your tools — those firefighters ran to the end still carrying their gear, because the gear was their identity. Perseverance on the right path is resilience and on the wrong path is stubbornness, and from the inside the two feel identical; the only way to tell is to stop, on a schedule, and re-examine the goal itself.