Near the end of his book, Daniel Kahneman admits something most authors would cut. Decades of studying how judgment goes wrong did not make his own judgment go right. He is still overconfident, still predicts too boldly, still underestimates how long things take. He only got better at recognizing the situations where errors happen, and best of all at recognizing them when someone else was walking in.
That confession is the whole argument. A bias is a feature of a situation, and situations can be rebuilt. The question is when to stop asking a person to concentrate harder and start changing the room they decide in.
Put the trigger in the workflow
Couples asked what share of the housework they do give answers that add up to more than 100 percent, because each remembers their own work best. "There is more than 100 percent of the credit to go around" works as a reminder because the fight is familiar and the cue is easy to spot. Kahneman's other suggestion is a shared vocabulary. A colleague at the watercooler who says "anchor" or "inside view" summons a whole diagnosis in two words. But if the same error keeps returning, the trigger belongs in the process, not in a promise to remember.
A formula that knows less and wins anyway
Orley Ashenfelter predicted the future price of Bordeaux vintages from the age of the wine, the summer temperature, the rain at harvest and the rain the previous winter. His predictions correlated at better than nine tenths with later prices and beat the experts who could actually taste the wine. The wine world was furious. People forgive a hundred scattered human errors more easily than one visible mechanical one.
The formula won by being boringly consistent, so the redesign is to collect evidence in a fixed order and combine it by rule. As a young army psychologist, Kahneman replaced a useless global interview with six separately scored traits, factual questions and a formula. The interviewers hated being turned into robots, so he let them close their eyes at the end and give one intuitive score. Prediction jumped from useless to moderately useful, and that final intuition added value precisely because it came after the structure.
The same discipline fixes forecasts. Julie read fluently at four, so people predict she will graduate near the top. Kahneman's four steps are to start from the average outcome for her group, note the number your intuition wants, estimate how well early reading really predicts college grades, and move from the average toward the intuition only by that fraction. A correlation of three tenths keeps 30 percent of the distance. It also demands a control group. Children selected for being unusually depressed improve at retest whatever you do, so only the improvement beyond an untreated group belongs to the treatment.
Ask what usually happens
Plans are written by the people who love them. The Scottish Parliament building was estimated at up to 40 million pounds and cost about 431 million. In a study of rail projects from 1969 to 1998, more than 90 percent overpredicted their passengers, by 106 percent on average. Homeowners who expected to spend 18,658 dollars on a kitchen spent 38,769. Nobody there was stupid. They could picture their own plan in detail and not the combined odds of illness, bureaucracy and bad luck that had delayed everyone before them.
Reference class forecasting reverses the order of thinking. Find the comparable projects first. Get their real costs, times and failure rates. Make that the baseline. Only then let the team argue, with evidence, for why this one is different. Gary Klein's premortem adds the missing social permission. When a plan is nearly chosen, the leader says it is a year from now, we did this, and it was a disaster. Everyone writes the history of that disaster, independently, before anyone speaks. Doubt that would have sounded disloyal becomes an assignment.
The same fix works on risk. Most people take a sure 240 dollars over a 25 percent shot at 1,000, then take a 75 percent chance of losing 1,000 over a sure loss of 750, and the popular pair combines into the worse package. A standing policy, always the high deductible, never the extended warranty, check the portfolio quarterly, decides those cases in a wide frame before a salesperson or a fresh loss can shrink it.
Heroic willpower is least reliable at the moment of maximum pressure.
Defaults are never neutral
Every choice already arrives with a wording, an order and a box that is ticked or not. Leaving that undesigned makes it accidental, not neutral. Kahneman's favored response is called libertarian paternalism. Set the default to what most people would want on reflection, and keep the exit real. Save More Tomorrow let employees commit a slice of future raises to savings, so take-home pay never fell and nobody felt a loss. Millions saved more, and anyone could leave. Meat labels allowed to say "90 percent fat-free" only if "10 percent fat" appears beside them at the same size are the same idea applied to honesty. A tiny, well-hidden opt-out does not preserve agency. Make the nudge visible, keep the exit easy, and, I would add, check who the helpful default is actually helping.
The decision factory
Kahneman asks organizations to see themselves as factories whose products include decisions. Before a big call, name the reference class, record independent estimates, hold the premortem. After it, check whether the evidence was combined as promised, whether any override was explained, and whether dissent got its turn. Then judge the process by what was knowable at the time, not by how the dice landed.
None of this takes the choice away. It puts the comparison cases and the objections in the room before intuition gets the final word. The person still decides. They just decide in a room built by someone who knew, as Kahneman knows about himself, that trying harder was never going to be enough.
Three things to keep
Name your triggers in advance: a first number, an extreme result, a fan's plan.
Before committing, have everyone write the story of how it failed.
Keep every default visible and every exit easy, then measure who it actually helps.