Think Twice Today https://thinktwicetoday.com/ How people decide — the biases, the arithmetic of risk, the quality of evidence, and the habits that make a good decision more likely than a lucky one. en Base rate neglect: why a vivid detail beats a background numberhttps://thinktwicetoday.com/biases/base-rate-neglect-why-a-vivid-detail-beats-a-background-number/https://thinktwicetoday.com/biases/base-rate-neglect-why-a-vivid-detail-beats-a-background-number/Wed, 26 Aug 2026 09:00:00 GMTBiasesSamar BhatiaWhen a description feels like a match, people stop asking how many of that kind exist in the first place, and that omission drives a surprising share of everyday misjudgement. Survivorship bias is why advice from people who succeeded is mostly noisehttps://thinktwicetoday.com/biases/survivorship-bias-is-why-advice-from-people-who-succeeded-is-mostly-noise/https://thinktwicetoday.com/biases/survivorship-bias-is-why-advice-from-people-who-succeeded-is-mostly-noise/Sun, 23 Aug 2026 09:00:00 GMTBiasesAdrian NovakThe people who failed doing exactly the same thing are not available to be interviewed, and their absence quietly rewrites every lesson drawn from the ones who did not. Hindsight makes a coin flip look like a planhttps://thinktwicetoday.com/biases/hindsight-makes-a-coin-flip-look-like-a-plan/https://thinktwicetoday.com/biases/hindsight-makes-a-coin-flip-look-like-a-plan/Wed, 19 Aug 2026 09:00:00 GMTBiasesSamar BhatiaOnce an outcome is known, the mind quietly revises what it believed beforehand, which destroys the record you would need to learn anything from experience. More options can make a choice worse, but the evidence is messier than the storyhttps://thinktwicetoday.com/choices/more-options-can-make-a-choice-worse-but-the-evidence-is-messier-than-the-story/https://thinktwicetoday.com/choices/more-options-can-make-a-choice-worse-but-the-evidence-is-messier-than-the-story/Sat, 15 Aug 2026 09:00:00 GMTChoicesSamar BhatiaThe idea that too much choice paralyses people became famous faster than it was verified, and what survives careful testing is narrower and more useful than the popular version. Opportunity cost is the price that never appears on the taghttps://thinktwicetoday.com/choices/opportunity-cost-is-the-price-that-never-appears-on-the-tag/https://thinktwicetoday.com/choices/opportunity-cost-is-the-price-that-never-appears-on-the-tag/Wed, 12 Aug 2026 09:00:00 GMTChoicesSamar BhatiaEvery choice quietly spends the best alternative you did not take, and because that cost is never invoiced it drops out of most reasoning entirely. Absolute and relative risk describe the same event and feel nothing alikehttps://thinktwicetoday.com/risk/absolute-and-relative-risk-describe-the-same-event-and-feel-nothing-alike/https://thinktwicetoday.com/risk/absolute-and-relative-risk-describe-the-same-event-and-feel-nothing-alike/Tue, 11 Aug 2026 09:00:00 GMTRiskVarun KrishnanA change expressed as a percentage of a percentage can sound alarming or trivial without either version being false, which makes knowing the difference a basic reading skill. Anchoring survived the replication wars, but its explanation did nothttps://thinktwicetoday.com/biases/anchoring-survived-the-replication-wars-but-its-explanation-did-not/https://thinktwicetoday.com/biases/anchoring-survived-the-replication-wars-but-its-explanation-did-not/Sat, 08 Aug 2026 09:00:00 GMTBiasesTara MukherjeeThe first number in the room pulls every later number towards it, and this is one of the few classic effects that large multi-laboratory tests reproduced rather than demolished. Confirmation bias is a search problem before it is a belief problemhttps://thinktwicetoday.com/biases/confirmation-bias-is-a-search-problem-before-it-is-a-belief-problem/https://thinktwicetoday.com/biases/confirmation-bias-is-a-search-problem-before-it-is-a-belief-problem/Tue, 04 Aug 2026 09:00:00 GMTBiasesTara MukherjeePeople rarely refuse contrary evidence outright; they simply never go looking for it, and that quiet asymmetry in searching does most of the damage. Why money already spent keeps getting a votehttps://thinktwicetoday.com/choices/why-money-already-spent-keeps-getting-a-vote/https://thinktwicetoday.com/choices/why-money-already-spent-keeps-getting-a-vote/Mon, 03 Aug 2026 09:00:00 GMTChoicesTara MukherjeeCosts that cannot be recovered are irrelevant to what you should do next, and yet they reliably distort the decision anyway. A decision can be good and still turn out badlyhttps://thinktwicetoday.com/choices/a-decision-can-be-good-and-still-turn-out-badly/https://thinktwicetoday.com/choices/a-decision-can-be-good-and-still-turn-out-badly/Thu, 30 Jul 2026 09:00:00 GMTChoicesRohan D’SouzaJudging choices by their results confuses two different things, and it is the fastest way to learn the wrong lesson from experience. Reversible and irreversible choices deserve different amounts of thoughthttps://thinktwicetoday.com/choices/reversible-and-irreversible-choices-deserve-different-amounts-of-thought/https://thinktwicetoday.com/choices/reversible-and-irreversible-choices-deserve-different-amounts-of-thought/Tue, 28 Jul 2026 09:00:00 GMTChoicesGautam PillaiThe cost of being wrong depends almost entirely on whether you can undo it, which means the same amount of deliberation is wasteful in one case and reckless in the other. Very small probabilities behave badly in the mindhttps://thinktwicetoday.com/risk/very-small-probabilities-behave-badly-in-the-mind/https://thinktwicetoday.com/risk/very-small-probabilities-behave-badly-in-the-mind/Mon, 27 Jul 2026 09:00:00 GMTRiskRohan D’SouzaBelow a certain size, people stop distinguishing between numbers that differ by a factor of a hundred, which is why rare events are simultaneously over-feared and under-prepared for. A test that is right most of the time can still be mostly wronghttps://thinktwicetoday.com/risk/a-test-that-is-right-most-of-the-time-can-still-be-mostly-wrong/https://thinktwicetoday.com/risk/a-test-that-is-right-most-of-the-time-can-still-be-mostly-wrong/Thu, 23 Jul 2026 09:00:00 GMTRiskTara MukherjeeWhen the thing being detected is rare, the great majority of positive results can be false without anything at all being defective about the test. Expected value stops being good advice when the loss is permanenthttps://thinktwicetoday.com/risk/expected-value-stops-being-good-advice-when-the-loss-is-permanent/https://thinktwicetoday.com/risk/expected-value-stops-being-good-advice-when-the-loss-is-permanent/Sun, 19 Jul 2026 09:00:00 GMTRiskVarun KrishnanAveraging outcomes assumes you get to keep playing, and a bet that ends the game is not made acceptable by having attractive odds. Risk and uncertainty are not the same problemhttps://thinktwicetoday.com/risk/risk-and-uncertainty-are-not-the-same-problem/https://thinktwicetoday.com/risk/risk-and-uncertainty-are-not-the-same-problem/Wed, 15 Jul 2026 09:00:00 GMTRiskVarun KrishnanOne has a known distribution and the other does not, and confusing the two produces confident numbers about situations that cannot support them. What a control group is actually forhttps://thinktwicetoday.com/evidence/what-a-control-group-is-actually-for/https://thinktwicetoday.com/evidence/what-a-control-group-is-actually-for/Sun, 12 Jul 2026 09:00:00 GMTEvidenceRohan D’SouzaA comparison group exists to answer one question — what would have happened anyway — and almost every weak study is weak because it cannot answer it. Regression to the mean makes useless interventions look like they workedhttps://thinktwicetoday.com/evidence/regression-to-the-mean-makes-useless-interventions-look-like-they-worked/https://thinktwicetoday.com/evidence/regression-to-the-mean-makes-useless-interventions-look-like-they-worked/Mon, 06 Jul 2026 09:00:00 GMTEvidenceRohan D’SouzaExtreme measurements are followed by less extreme ones for purely statistical reasons, and any programme aimed at the worst cases inherits that improvement for free. Before you call it causation, three other explanations have to failhttps://thinktwicetoday.com/evidence/before-you-call-it-causation-three-other-explanations-have-to-fail/https://thinktwicetoday.com/evidence/before-you-call-it-causation-three-other-explanations-have-to-fail/Fri, 03 Jul 2026 09:00:00 GMTEvidenceRohan D’SouzaCorrelation has a small number of standard alternative explanations, and running through them takes about a minute and settles most arguments. Why a shelf of famous psychology findings quietly stopped being truehttps://thinktwicetoday.com/evidence/why-a-shelf-of-famous-psychology-findings-quietly-stopped-being-true/https://thinktwicetoday.com/evidence/why-a-shelf-of-famous-psychology-findings-quietly-stopped-being-true/Thu, 02 Jul 2026 09:00:00 GMTEvidenceGautam PillaiWhen large teams began repeating well-known experiments with bigger samples and fixed analysis plans, a substantial share of the results did not come back — and understanding why is more useful than any of the original claims. Analysis has too many degrees of freedom, which is why preregistration existshttps://thinktwicetoday.com/evidence/analysis-has-too-many-degrees-of-freedom-which-is-why-preregistration-exists/https://thinktwicetoday.com/evidence/analysis-has-too-many-degrees-of-freedom-which-is-why-preregistration-exists/Wed, 01 Jul 2026 09:00:00 GMTEvidenceTara MukherjeeThe same dataset can be analysed in dozens of defensible ways, and choosing among them after seeing the results turns noise into findings without anyone intending it.