System 2 is your slow, deliberate thinking — the part that does maths, checks logic, and resists impulses.

It runs on one small pool of attention. Every hard task pays out of the same account — so spending heavily on one leaves less for everything else.

Navigating busy highway traffic is hard. So is solving 17 × 24. Try both together and each steals attention from the other.

Simultaneous hard tasks don’t queue politely — they compete for the same pool, and both get worse.

Four things raise the price of a task:

• Holding several ideas that each need a separate action
• Overriding a habitual response — executive control
• Switching between tasks
• Time pressure

Add-3 sits at the limit of what most people can do. Harder, and they give up — you can’t will yourself past the limit, just as you can’t sprint faster than a sprint.

Effort isn’t invisible. During Add-3, pupils dilated by 50% and heart rate rose by 7 bpm — mental strain is a physical event.

Strain climbs further with stress, distractions, physical exertion (even a stroll uses mental resources), and exercising self-control. All of it makes clear thinking harder.

Skill. The more skilled you get, the less energy a task consumes — a doctor diagnosing, a grandmaster picking a chess move.

Flow. With clear goals, immediate feedback, and challenge matched to skill, you concentrate without spending willpower at all.

Incentives. Given the right reward, people resist depletion far better.

At ease: good mood, you believe what you hear, trust your intuitions, feel it’s familiar — creative but superficial. System 1 hums along and System 2 accepts whatever it says.

Under strain: vigilant and suspicious, you invest more effort and make fewer errors — but you’re less intuitive and less creative.

Four inputs create ease: repeated experience, clear display, a primed idea, and a good mood.

The trap: people shown made-up names later falsely “remembered” them as celebrities, and words shown more often were rated more positive. You like what you’ve seen often — regardless of merit.

Clear, high-contrast, simple, rhyming messages simply feel more believable — “Woes unite foes” lands as truer than “Woes unite enemies.”

Run it in reverse: when the $1.10 bat-and-ball problem was printed small and hard to read, accuracy rose significantly. The strain switched System 2 on.

Put the whole book so far together and the danger moment is simple: System 1 builds a neat story that connects the dots, feels no surprise — so System 2 never wakes up to question it. It just accepts.

That shortcut is fine for everyday low-stakes calls — you don’t need to investigate how to brush your teeth. It is risky when the stakes are high and information is thin: sitting on a jury, hiring someone, or reacting in an emergency.

WYSIATI: What You See Is All There Is. System 1 builds the best story it can from the evidence in front of it. It does not stop to ask: how good is this evidence? How much is missing?

In one experiment, people heard only one side of a legal case — and they were more confident in their judgment than people who heard both sides, even though they knew the setup. Less evidence, more certainty.

If you like one thing about a person — they look good, they’re friendly — that feeling spills onto everything else. You assume they are also smart and capable, with no evidence at all.

Then the loop: you like smart, capable people, so now you like this person even more. Your feelings fill in the blanks your knowledge left empty. (It works in reverse too — one disliked trait darkens everything.)

Two men, same five words, different order: Amos — intelligent, hard-working, strategic, suspicious, selfish — sounds likable. Barry — selfish, suspicious, strategic, hard-working, intelligent — does not. The first words change how you read the later ones.

It is powerful: teachers who were told (randomly) that some students were “about to bloom” saw those students do better by year-end — the Pygmalion effect. Kahneman caught it in his own exam grading and fixed it by marking one question across all students instead of one student at a time.

Confirmation bias: we bend information so it confirms what we already believe. It shows up three ways:

• We notice data that fits, and discard data that doesn’t
• We seek out sources that agree, and avoid sources that don’t
• We remember confirming facts easily, and forget the rest

Show someone a word, face, or logo again and again, and they will like it more — just from having seen it. A memory of it makes it easier to see again, and that ease feels like liking.

Short card, because you met this in Part 1–2 with cognitive ease: repetition creates ease, and ease feels good — whether or not the thing deserves it.

People crave cause-and-effect stories, especially when things are uncertain. So System 1 builds one — stock-market moves get horoscope-style explanations (the same event “explains” both a rise and a fall), and famous people get neat origin stories, like Steve Jobs succeeding because of his childhood.

Once a story is established, it is very hard to overwrite — which is how frauds like Theranos and Enron kept going: observers believed the story they wanted to hear.

How much you like or dislike something decides what you believe about it. Like an option, and you judge its benefits bigger and its risks smaller than the alternatives. Dislike it, and everything flips.

The strange part: hear one new fact about the benefits, and your estimate of the risks drops too — even though you learned nothing new about risks. The feeling moved both dials.

Psychologist Daniel Gilbert’s model: believing is System 1’s job — it builds the best interpretation of an idea, as if it were true. Unbelieving is System 2’s job — it checks and rejects the false ones.

A taxed System 2 stops unbelieving. In experiments, people holding digits in memory believed false sentences far more easily. That is one reason late-night infomercials work — and why persuasive nonsense spreads when people are exhausted. (This is the System 2 capacity limit from Part 1–2, doing damage.)