Vol.331 思考，快与慢：我的大脑为什么总是欺骗我

System One is the unconscious fast thinking that runs on its own, great at intuitive association and emotional reaction, handling 95 to 96 percent of daily decisions. System Two is the effortful, energy-burning slow logic responsible for complex calculations, but it's lazy by nature and mostly just rubber-stamps System One's conclusions.

The deliberation you think you're doing has, 95 percent of the time, already been done by System One — System Two only confirms it afterwards. Free will may be far more limited than we'd like to believe.

A typical case: listening to a podcast runs on System One; the moment someone asks you what 17 times 24 equals, System Two gets awakened.

The original quote: 'System Two only steps in when it absolutely has to be engaged and awakened — and then it says, wait, this isn't right.'

Related reading note: reading this alongside The Molecule of More can trigger a massive wave of self-loathing, so the hosts suggest reading them separately.

Dayi compared System Two's laziness to a popular internet meme: a father's love is as unmoving as a mountain — because he doesn't move. Only in a 'called-to-see-the-principal' crisis does he show up to correct you; the rest of the time System One does whatever it wants.

The cause: small samples easily produce extreme results — it's just random fluctuation. But System One forces a causal story onto it and refuses to accept that there's no reason; it's simply random.

The fix: when you see a statistical conclusion from a tiny sample, first ask whether the sample size is big enough, then decide whether to buy the causal narrative.

How serious: watch out — a lot of medical anxiety and regional prejudice stems from exactly this pattern.

The guest's take: 'The doctor can't explain why you got this disease. From a scientific standpoint it's simply random — but System One won't accept that explanation.'

The mechanism: the first number that appears anchors your later judgment. In the Gandhi age experiment, people first asked whether he died before or after 140 ended up anchoring their exact-age answers near 140.

The insight: marketing gimmicks like 'limited to 12 cans' and a negotiator opening at 10 and settling at 8 are all manipulating the range of your decision via anchoring.

The mechanism: we don't evaluate gains and losses by absolute amount, but by change relative to a reference point. Give the same 50,000-yuan year-end bonus to two people — the one who expected 80,000 suffers, the one who expected 10,000 is overjoyed.

The insight: the brain struggles with absolute values — numbers only make sense when placed inside a comparison system.

The original quote: 'The human nervous system has a hard time accepting absolute values — it has to be placed in some framework, and it's far more sensitive to changes in value.'

The mechanism: the pain of losing a given amount is about 1.5 to 2.5 times the pleasure of gaining the same. In the coin-flip experiment, most people only join if the upside is around 200 dollars against a 100-dollar loss on tails.

A typical case: people sell their winning stock while holding the loser, because selling the loser means 'confirming the loss.' And a 30-day free trial creates the endowment effect, which lowers return rates.

How to apply it: before any major purchase, ask yourself — 'If I'd never owned this thing, how much would I be willing to pay for it?'

The mechanism: the experiencing self lives in every single moment — the aching back, the dry mouth — but it has no say in decisions and forgets easily. The remembering self evaluates the past, makes the decisions, and follows the peak-end rule and duration neglect.

The insight: the sunflower-seed effect — no matter how happy you were along the way, one bad seed at the end ruins your entire memory of the experience.

The quote: 'Are we living in the present or living in memory? Could it be that our over-attention to remembering happiness makes us neglect the present itself?'

Related idea: enduring a miserable trip just to check in at scenic spots for social media is a classic case of the remembering self hijacking the experiencing self.

Why it matters: a society can have high GDP and strong social status — satisfying the remembering self — while people lie awake anxious every night, the experiencing self suffering. That society may not be truly happy.

Boundary conditions: long-term planning still needs the remembering self in charge — just not at the cost of every present moment.

Xing Guang's tangent: a large model 'taking a car to the car wash' fails over 80 percent of the time, because it gets anchored by the 50-meter distance and ignores the objective function — washing the car. A model's long-winded reasoning looks like System Two, but underneath it's still heuristic fast thinking.

How to apply it: when you prompt an AI, state your preconditions and objective function explicitly, and watch out for its 'hallucinations' — perfectly logical reasoning built on missing premises.