16161616 psychology findings that survived the replication crisis

Judgment and choice · 1 of 16

The finding

A number you were just exposed to pulls your estimate toward it, even when it is obviously irrelevant and even if you are an expert.

Classic study

Tversky & Kahneman (1974) spun a wheel of fortune rigged to stop at 10 or 65, then asked people to estimate the percentage of African countries in the UN. Median guess: 25% after seeing 10, 45% after seeing 65 — and everyone had watched the wheel spin. Ariely, Loewenstein & Prelec (2003): MIT students who wrote the last two digits of their social security number before bidding on wine and keyboards bid up to three times more if the digits were high. Englich et al. (2006): German judges rolled dice loaded to 3 or 9 as the "prosecutor's demand" in a shoplifting case; those who rolled 9 gave sentences averaging about 8 months versus about 5.

Why it happens

People adjust away from the anchor and stop as soon as they reach a plausible value, so they under-adjust. And asking "higher or lower than 65?" makes you retrieve facts consistent with 65 — the anchor chooses which memories get to vote.

How solid

Many Labs 1 tested four anchoring questions in 36 samples across a dozen countries; every sample replicated, with effects among the largest in the whole project. Survives warnings, accuracy incentives and expertise. One of the few effects you can demonstrate reliably on a single friend.

Where it stops

The anchor has to be on the same scale. Extreme anchors work with diminishing returns. Leaning on an informative anchor (a list price, last year's number) is sensible — the bias is that it cannot be switched off when the anchor is noise.

Everyday version

The first number in a negotiation sets the range the rest of it lives in. "Was $199, now $99" works even if nobody ever paid $199.

Judgment and choice · 2 of 16

The finding

Describing the same outcome as a gain or a loss changes what people choose. Facing gains they take the sure thing; facing losses they take the gamble.

Classic study

Tversky & Kahneman (1981), the "Asian disease" problem. An outbreak is expected to kill 600 people. Program A saves 200 for certain; Program B has a one-third chance of saving all 600 and a two-thirds chance of saving nobody — 72% chose A. Reworded as losses: Program C, 400 die for certain; Program D, a one-third chance nobody dies and a two-thirds chance all 600 die — 78% chose D. A and C are the same outcome; so are B and D.

Why it happens

The description sets the reference point and people judge relative to it. From "600 will die," saving 200 is a gain, so you don't risk it. From "nobody has died yet," 400 deaths is a loss and 600 doesn't feel much worse, so you gamble on zero. Prospect theory in one problem: sensitivity diminishes in both directions, and the frame picks the reference point.

How solid

Many Labs 1 replicated it in all 36 samples. Ruggeri et al. (2020) replicated prospect theory's core patterns across 19 countries. McNeil et al. (1982): doctors and patients were much more likely to choose lung-cancer surgery when told "90% survive the operation" than "10% die during it."

Where it stops

Risky-choice framing (above) and attribute framing (beef labelled 75% lean beats 25% fat) are rock solid. Goal framing — stressing what you gain by acting versus lose by not acting — is inconsistent. Show one person both frames side by side and many become consistent.

Everyday version

Credit-card companies lobbied for stores to call the price difference a "cash discount" rather than a "credit surcharge." Same numbers; the second sounds like a loss.

Judgment and choice · 3 of 16

The finding

People keep investing in something because of what they have already spent on it, even when the spending is unrecoverable and the future case is bad.

Classic study

Arkes & Blumer (1985) sold theatre season tickets at random at full price, $2 off or $7 off; full-price buyers attended more plays in the first half of the season. In a scenario, people who had bought a $100 ski trip and, later, a $50 trip they expected to enjoy more, both for the same weekend, mostly chose the $100 trip. In the "radar-blank plane" case, 85% said a company should finish a $10 million project after $9 million was spent, even though a competitor had just launched a cheaper, better version; with the prior spend removed from the story, only 17% said go ahead.

Why it happens

Quitting turns a paper loss into a certain one, and people are loss-averse. Add a waste norm — abandoning spent money feels like admitting it was wasted — and self-justification: the person who made the original call escalates most (Staw 1976, "escalation of commitment").

How solid

Many Labs 1 replicated it. Roth, Robbert & Straus (2015) found it across domains in a meta-analysis. Sweis et al. (2018, Science): mice, rats and humans making foraging decisions all show it — the longer you have already waited, the less likely you are to give up.

Where it stops

Weaker when people are prompted to think about future value, when the decider did not make the original commitment, and when there is a face-saving exit. Past investment sometimes carries real information, and reputational costs of quitting are real.

Everyday version

Sitting through a bad film because you paid. Shipping a feature nobody wants because eight engineers spent a quarter on it. Concorde — hence "the Concorde fallacy."

Judgment and choice · 4 of 16

The finding

People demand more to give something up than they would pay to acquire it. Ownership itself raises the price.

Classic study

Kahneman, Knetsch & Thaler (1990) gave coffee mugs to half the students in a class and let a market run. Sellers wanted a median of roughly $5–7; buyers would pay roughly $2–3. Theory predicts about half the mugs should change hands; actual trading was about a quarter of that. Knetsch (1989) gave students a mug or a chocolate bar and offered a swap; about 90% kept whatever they had been handed, in both directions.

Why it happens

Loss aversion — giving up the mug is a loss, receiving it a gain, and losses weigh roughly twice as much. Sellers focus on the mug they are losing, buyers on the money they are losing (Carmon & Ariely 2000), and ownership pulls the object slightly into the self.

How solid

Replicated hundreds of times in labs and real markets, with mugs, lottery tickets and hunting permits. Shows up in chimpanzees and capuchin monkeys.

Where it stops

Vanishes for things held purely to exchange — tokens, cash, inventory. Market experience shrinks it: List (2003) found professional sports-card dealers barely showed it. Weaker when people expect to trade.

Everyday version

"Try it free for 30 days" — returning it now feels like losing something. Homeowners overpricing the house they have lived in. Easy to add a subscription, painful to cancel one.

Cognitive bread and butter · 5 of 16

The finding

Naming the ink colour of a word is slow and error-prone when the word spells a different colour (RED printed in blue). Reading the word is unaffected by the ink.

Classic study

Stroop (1935). Participants named the colours of 100 items: coloured squares took about 63 seconds; colour words in conflicting inks took about 110 seconds — 74% slower. Reading the words aloud was just as fast whatever the ink colour.

Why it happens

For a literate adult, reading is automatic — it happens whether you want it to or not — and it is faster than colour naming. The word's answer arrives first and has to be suppressed before the colour's answer can be spoken; the delay is the cost of resolving that conflict. The cleanest demonstration that some mental processes cannot be switched off, and the standard task for studying cognitive control.

How solid

MacLeod's 1991 review covered fifty years and found the effect in essentially every participant ever tested. Many Labs 3 included it and got the same enormous effect. Measured within each person, across dozens of trials, in milliseconds; effect sizes several times larger than anything in social psychology. Shaky cousin: the "emotional Stroop" clinical variant is less consistent.

Where it stops

Smaller in young children who read slowly and in second-language versions; reducible with practice, not removable.

Everyday version

Write GREEN in red ink and name the ink — you will feel the tug. A red button labelled "Go" causes errors: two automatic readings fighting.

Cognitive bread and butter · 6 of 16

The finding

For the same total study time, spreading practice across sessions produces far better long-term retention than bunching it together.

Classic study

Ebbinghaus (1885) noticed it memorising nonsense syllables: repetitions spread over three days needed fewer total repetitions than cramming into one. Cepeda et al. (2006) reviewed 254 studies; spaced practice won almost every time. Cepeda et al. (2008): the best gap depends on how long you need to remember — for a test a week away, a gap of about a day; a year away, about three weeks. Bahrick et al. (1993) taught foreign vocabulary in 13 sessions spaced two, four or eight weeks apart; eight-week spacing was worst during training and best when tested up to five years later.

Why it happens

An immediate repeat feels familiar, so you process it less deeply. Spaced repeats occur in slightly different mental contexts, giving the memory more retrieval routes. And each spaced repeat forces you to retrieve the earlier one, which strengthens it — the effort is the point.

How solid

Works for facts, vocabulary, motor skills and maths procedures; in children and adults; in labs, classrooms and animals. Dunlosky et al. (2013) rated it "high utility." Close to a law.

Where it stops

Cramming is fine if the test is tomorrow and you never need the material again; the advantage grows with the retention interval. Massed practice produces a fluency illusion, so learners prefer the strategy that works worse.

Everyday version

Five 20-minute sessions across two weeks beat one 100-minute session the night before — badly — when tested a month later. Every spaced-repetition app is built on this.

Cognitive bread and butter · 7 of 16

The finding

Trying to recall information strengthens memory more than re-reading it — even without feedback, and even though it feels less effective.

Classic study

Roediger & Karpicke (2006). Students read short science passages, then either re-read them several times or took repeated free-recall tests with no feedback. Five minutes later the re-readers remembered slightly more. A week later the tested group remembered about 60% versus about 40% for the re-readers. The students had predicted the opposite. Karpicke & Blunt (2011, Science): retrieval practice beat concept mapping — the poster child of "active learning" — even on inference questions a week later.

Why it happens

Retrieval is not a passive readout; it rewrites the memory, strengthening the route you used and attaching new cues. Bjork's "desirable difficulty": effort that lowers immediate performance and raises long-term retention. Re-reading mostly produces recognition fluency — the feeling of knowing without the ability to produce.

How solid

Meta-analyses (Rowland 2014; Adesope et al. 2017) put the effect around half a standard deviation, across ages from primary school to medical school and in real classrooms.

Where it stops

Retrieval has to succeed at least partly or be followed by feedback; failed attempts in the dark do not help much. The advantage appears after a delay; on an immediate test re-reading can look equal or better, which is why learners misjudge it.

Everyday version

Close the book, write down what you remember, then check. Attempt the flashcard before flipping. Practice exams over highlighting.

Cognitive bread and butter · 8 of 16

The finding

Recalling a list, people remember the first few items (primacy) and the last few (recency) better than the middle.

Classic study

Murdock (1962) had people freely recall word lists of different lengths and found the same U-shaped curve every time. Glanzer & Cunitz (1966) showed the two ends have different causes: 30 seconds of counting backwards wiped out recency but left primacy intact, while slower presentation boosted primacy but not recency.

Why it happens

Early items get extra rehearsal because nothing else is competing yet, so they reach long-term memory. Late items are still in short-term memory when recall begins — until a distracting task flushes them. That dissociation was a main pillar for the idea of separate short- and long-term stores; later theories reframe recency as temporal distinctiveness, but the curve is untouched.

How solid

Immediate and delayed recall; words, pictures, faces and odours; children and adults; monkeys and rats. Present in nearly every individual, with huge effect sizes.

Where it stops

Order effects in persuasion and impression formation (Asch's 1946 "intelligent, industrious… envious" lists) are related but separate findings — do not collapse them into this one.

Everyday version

Fifteen spoken groceries — you get the first two and the last two. The middle candidate in a day of interviews. The middle of a long presentation.

Cognitive bread and butter · 9 of 16

The finding

People fail to notice large changes to a scene when the change coincides with a brief interruption — a blink, an eye movement, a flicker, something passing in front — or even in plain view when attention is elsewhere.

Classic study

Rensink, O'Regan & Clark (1997) alternated a photograph with a slightly altered version with a blank screen in between; observers often took tens of seconds to notice changes as large as an aircraft engine disappearing. Simons & Levin (1998): an experimenter asked pedestrians for directions, two workmen carried a door between them, and a different experimenter walked away holding the map — about half the pedestrians did not notice. Sister effect: the "invisible gorilla" (Simons & Chabris 1999), where half the viewers counting basketball passes missed a person in a gorilla suit — that is inattentional blindness, failing to see something new rather than something changed.

Why it happens

Vision does not build the detailed, stable picture it feels like it builds. You hold a few attended objects in detail and use the world itself as memory for the rest. Spotting a change needs a "before" to compare, and if you were not attending there is none. Normally a change produces a motion signal that grabs attention; the flicker or the door masks it.

How solid

Replicates in every lab that has tried it — photographs, films, live interactions, across cultures. Film editors have exploited it for a century; continuity errors mostly go unseen. "Change blindness blindness" (Levin et al. 2000): people are confident they would notice changes they in fact miss.

Where it stops

Changes to attended, central objects are caught fast. Experts notice domain-relevant changes better. The eyes often do land on the changed object — the failure is in comparison, not looking.

Everyday version

You check the mirror, glance back, and the motorcycle that arrived during the glance is not "there." Users miss an inline error message that appears during a page reload because nothing moved to draw the eye.

Personality and intelligence · 10 of 16

The finding

The ways people differ, as described in ordinary language, boil down to five broad dimensions — openness, conscientiousness, extraversion, agreeableness, neuroticism. Scores are stable over years, agreed on by people who know you, partly heritable, and predictive of important outcomes.

Where it came from

Statistics, not theory. Allport & Odbert (1936) pulled roughly 18,000 trait words from the dictionary. Cluster the ratings people give on those words and five groups keep emerging (Fiske 1949; Tupes & Christal 1961; Goldberg 1990; Costa & McCrae's NEO inventory). The same structure appears across many languages with modest variation; HEXACO adds a sixth factor, honesty–humility.

The evidence

Roberts & DelVecchio (2000) pooled 152 longitudinal studies — rank-order consistency rises from about .3 in childhood to about .5 in college to about .7 after 50; the person more conscientious than you at 30 is very probably still more conscientious at 60. Self-ratings correlate about .4–.6 with ratings from friends and spouses. Heritability from twin studies is around 40–50%. Roberts et al. (2007): traits forecast mortality, divorce and occupational success about as well as IQ or socioeconomic status. Conscientiousness is the best personality predictor of job performance across occupations (Barrick & Mount 1991) and of longevity; neuroticism predicts anxiety, depression and relationship dissatisfaction; extraversion predicts happiness and leadership emergence. People mature predictably: more conscientious, agreeable and emotionally stable through adulthood.

Why it happens

Fleeson's model — a trait is a distribution of momentary states; an extravert is not always extraverted, they spend more hours in extraverted states. Stability comes from genes, from environments people choose for themselves, and from habit.

How solid

Personality psychology mostly sidestepped the replication crisis — correlational methods, large samples, instruments validated for decades. Contrast: the Myers-Briggs sorts people into types on dimensions that are actually continuous, as many as half of test-takers get a different type when retested weeks later, and it adds nothing predictive beyond the Big Five.

Where it stops

Correlations are modest (.2–.3), so traits predict patterns across many situations far better than any single act. Mischel's 1968 critique that situations matter as much as traits was largely right — both matter, each around .3. "Five" is a convenient level of abstraction; each factor splits into facets that often predict better.

Everyday version

Hiring on conscientiousness is a valid, modest bet. Reading a colleague from a four-letter type code is not.

Personality and intelligence · 11 of 16

The finding

Scores on any diverse set of mental tests — vocabulary, arithmetic, spatial puzzles, memory span — correlate positively with each other. That "positive manifold" yields a general factor, g. IQ scores built on it are highly reliable, stable across a lifetime, and predictive of education, work, health and longevity.

Classic evidence

Spearman spotted the positive manifold in 1904; nobody has found a dataset of diverse tests without it. Stability: the Scottish Mental Survey tested nearly every 11-year-old in Scotland in 1932; Deary and colleagues retested survivors at 77–80 and found a correlation of around .6–.7 across 66 years. Prediction: Strenze (2007) — IQ correlates about .56 with educational attainment, .45 with occupational level and .23 with income. Health: Calvin et al. (2017, BMJ) followed nearly 66,000 Scots tested at 11 in 1947 to age 79; each standard deviation of childhood IQ was associated with roughly 20% lower risk of death, strongest for respiratory disease, heart disease and stroke. Work: Schmidt & Hunter (1998) put general mental ability as the best single predictor of job performance; Sackett et al. (2022) lowered the estimate substantially but kept it useful, with structured interviews predicting at least as well.

Why it happens

Less settled than the phenomenon. g correlates with processing speed, working-memory capacity and various brain measures; heritability rises from about .2 in infancy to .6–.8 in adulthood as people increasingly shape their own environments. Van der Maas et al. (2006) argue g is not one underlying thing but emerges because abilities feed each other during development — better memory helps vocabulary, which helps reasoning. The data pattern is the same either way.

How solid

Arguably the most replicated finding in psychology. Full-scale IQ test-retest reliability above .9. Predictive relationships replicate across countries and cohorts.

Where it stops

The tests measure a specific family of abilities — not creativity, judgment or social skill. A correlation of .5 leaves three-quarters of the variance unexplained, so individual predictions are loose. The Flynn effect — scores rose about three points a decade through the 20th century — shows environment moves whole populations. Debates about group differences are about causes and remain contested; these findings do not settle them.

Everyday version

Two children in the same school, one at the 85th percentile and one at the 15th at age 11 — decades later the first is more likely to hold a degree and a complex job, and somewhat more likely to be alive, with plenty of exceptions both ways.

Social perception · 12 of 16

The finding

Faced with a unanimous group giving an obviously wrong answer to a simple question, a large fraction of people go along at least some of the time — and most of them know the group is wrong.

Classic study

Asch (1951, 1956). One real participant sat with seven confederates judging which of three lines matched a standard — a task people get right over 99% of the time alone. On 12 of 18 trials the confederates unanimously gave the wrong answer. About 37% of participants' answers on those trials matched the group; three-quarters conformed at least once; a quarter never did. Afterwards, most said they knew the answer was wrong but did not want to stand out; a minority had begun to doubt their own eyes. Variations: a single confederate answering correctly cut conformity to around 5%; the effect plateaued at a group of three or four; answering privately in writing largely removed it.

Why it happens

Deutsch & Gerard (1955) split it into two forces. Normative influence — going along to avoid disapproval while privately disagreeing — drives most Asch conformity, since the task is unambiguous. Informational influence — believing others might see something you don't — dominates when the task is ambiguous, as in Sherif's earlier studies of a stationary light that seemed to move in the dark.

How solid

Bond & Smith (1996) meta-analysed 133 studies in 17 countries: found everywhere, larger in collectivist cultures, declining in the US since the 1950s but never gone. Replicated with modern samples and online groups. Berns et al. (2005): conforming on a perceptual task changed activity in visual areas — some people's perception really shifts.

Where it stops

Needs unanimity; one ally breaks it. Most answers were still correct. Asch himself thought the study showed independence as much as conformity and disliked the "sheep" reading.

Everyday version

A meeting where everyone nods and the one sceptic goes quiet. The countermeasure follows from the variations: collect independent written estimates before anyone speaks.

Social perception · 13 of 16

The finding

Explaining other people's behaviour, we lean on their character and underweight their situation, even when the situational pressure is obvious.

Classic study

Jones & Harris (1967) had people read essays for or against Castro. Told the writer had been assigned the position by a debate instructor, readers still concluded the pro-Castro writer was more pro-Castro — although the assignment made the essay uninformative. Ross, Amabile & Steinmetz (1977) ran a quiz game: a coin toss made one participant the questioner, who wrote hard questions from their own store of knowledge, and one the contestant, who mostly failed. Observers, and the contestants themselves, rated questioners far more knowledgeable, ignoring that the role produced the gap.

Why it happens

Gilbert's account — we categorise the behaviour and characterise the person automatically, then correct for the situation only as an effortful afterthought that fails when we are distracted. Perceptually the person is the figure and the situation the background. And we underestimate how strong social pressures are.

How solid

The attitude-attribution paradigm was included in Many Labs 2 and replicated across sites. Culture adds texture: East Asian participants show less of the bias when situational cues are made salient (Choi, Nisbett & Norenzayan 1999); Morris & Peng (1994) found Chinese-language newspapers explained the same murders situationally where English-language ones explained them dispositionally. The baseline bias appears everywhere.

Where it stops

The related "actor–observer asymmetry" — that we explain our own behaviour situationally — is weak and depends on whether the behaviour is good or bad (Malle 2006). "Fundamental" is contested; "correspondence bias" is the neutral term, and it shrinks when people are motivated to be accurate.

Everyday version

The driver who cuts you off is a jerk, not someone racing to a hospital. The support agent is rude, not working a 15-minute call quota. The colleague who missed the deadline is lazy, not blocked on a dependency you never saw.

Social perception · 14 of 16

The finding

Once you know how something turned out, it seems more predictable than it was, and you misremember your own earlier expectations as closer to the outcome. "I knew it all along."

Classic study

Fischhoff (1975) gave people accounts of an obscure 19th-century war between the British and the Gurkhas with four possible outcomes. Those told the actual outcome rated it far more likely — roughly 57% versus about 33% among those not told — even when instructed to answer as if they didn't know. Fischhoff & Beyth (1975) had students predict outcomes of Nixon's trips to China and the USSR, then recall their predictions afterwards; people remembered having predicted what occurred. Caplan et al. (1991) showed anaesthesiologists identical case files; the same care was judged substandard far more often when reviewers were told the patient had died.

Why it happens

Roese & Vohs (2012) separate three layers: memory distortion (you cannot recover your prior state of knowledge once the outcome is integrated), inevitability (you generate reasons it happened, and reasons make it feel determined), and foreseeability (having reasons makes you feel you could have seen it). Plus a motivational strand — the world feels more controllable if outcomes were predictable.

How solid

Guilbault et al. (2004) meta-analysed 122 studies: a consistent moderate effect in children and adults, novices and experts, across cultures. It persists when people are warned about it.

Where it stops

Forcing people to explain how the other outcomes could have happened reduces it. Truly surprising outcomes that resist explanation sometimes produce the reverse. It is less about self-flattery than the pop version implies — it happens even when people have nothing to gain.

Everyday version

After a launch flops, everyone knew it would. Post-mortems that punish a decision that was sound on the information available. The countermeasure is mechanical: write forecasts down before the outcome — decision journals, pre-mortems, prediction logs.

Social perception · 15 of 16

The finding

Simply encountering something repeatedly makes you like it more, even when you don't remember encountering it.

Classic study

Zajonc (1968) showed people nonsense words, Chinese-like characters and yearbook photos between 0 and 25 times, then had them rate each; ratings rose with exposure. Kunst-Wilson & Zajonc (1980) flashed irregular polygons for a thousandth of a second, too fast to see consciously; later, participants preferred the flashed polygons over new ones about 60% of the time while being unable to recognise which they had seen. Moreland & Beach (1992) had four similar-looking women attend a large lecture course 0, 5, 10 or 15 times without speaking to anyone; at term's end, students rated the woman they had seen most as the most attractive. Mita, Dermer & Knight (1977): people prefer their own mirror image while their friends prefer the true photograph — each likes the version they have seen most.

Why it happens

Familiar things are processed more fluently, and that ease gets misattributed as liking. Evolutionary framing: a stimulus you have met repeatedly without harm is, by definition, safe, so mild warmth toward it is a sensible default.

How solid

Bornstein (1989) meta-analysed 208 experiments and Montoya et al. (2017) 268 studies; both find it reliably. Appears in infants, chicks and rats; with music, faces, brands and abstract shapes. Subliminal exposure often produces a larger effect than conscious exposure, because recognition can trigger boredom or counter-arguing.

Where it stops

Inverted U — liking rises over the first 10–20 exposures, then plateaus or falls. It intensifies existing dislike rather than reversing it. Varied stimuli wear out more slowly than identical repetition.

Everyday version

The song you hated in week one and hum in week four. Why advertisers pay for reach they cannot tie to any sale. Why a redesign that is better on every measurable dimension still gets "I preferred the old one" for a month.

Defaults · 16 of 16

The finding

Whatever happens when a person does nothing gets chosen at far higher rates than the same option when it must be actively selected. In high-stakes, unfamiliar or tedious decisions, the default often decides.

Classic study

Johnson & Goldstein (2003, Science) compared European organ-donation registration. Where you are a donor unless you opt out — Austria, France, Hungary, Poland, Portugal — effective consent rates were 85–99%. In otherwise similar opt-in countries — Denmark, Germany, the Netherlands, the UK — roughly 4–27%. An online experiment removed the confounds: with non-donor as the default, 42% chose to donate; with donor as the default, 82% stayed in; with a forced choice and no default, 79% chose to donate. Madrian & Shea (2001): a large US employer switched its retirement plan from opt-in to automatic enrolment; participation among new hires jumped from about 37% to about 86%. Most also stayed at the default contribution rate and fund, both conservative — the default anchored savings down as well as participation up.

Why it happens

Three forces stack. Effort and procrastination: switching needs action, and "later" is always available. Implied endorsement: people read the default as what the designer recommends or what most people do. Loss aversion: the default becomes the reference point, so switching feels like giving something up — the endowment effect applied to a choice.

How solid

Jachimowicz et al. (2019) meta-analysed 58 studies: a large average effect across health, environment, finance and consumer decisions, with some publication bias but nowhere near enough to explain it away. Policy confirmed it at national scale: UK pension auto-enrolment, introduced in 2012, took workplace participation from around 55% to around 88%. When a 2022 reanalysis found the broader "nudge" literature shrank drastically after correcting for publication bias, defaults were the category that clearly held.

Where it stops

Defaults move the paperwork, not necessarily the downstream result — opt-out organ registration does not automatically raise transplants, because families are still consulted and hospital capacity binds. Weaker when people hold strong preferences; can backfire when the default is transparently self-serving. It cuts both ways: pre-ticked boxes, auto-renewals and buried cancel buttons are the same mechanism turned against people, which is why several jurisdictions now ban pre-ticked consent.

Everyday version

Nobody changes their privacy settings. A university that made double-sided the printer default reported cutting paper use by around 40%. The design rule: set the default to what most people would choose if they thought carefully, because most of them won't.