Start free trial
Trending New Genres Authors
EnglishEnglish
EspañolSpanish
简体中文Chinese
繁體中文Chinese (Traditional)
FrançaisFrench
DeutschGerman
日本語Japanese
PortuguêsPortuguese
ItalianoItalian
한국어Korean
РусскийRussian
NederlandsDutch
العربيةArabic
PolskiPolish
हिन्दीHindi
Tiếng ViệtVietnamese
SvenskaSwedish
ΕλληνικάGreek
TürkçeTurkish
ไทยThai
ČeštinaCzech
RomânăRomanian
MagyarHungarian
УкраїнськаUkrainian
IndonesiaIndonesian
DanskDanish
SuomiFinnish
БългарскиBulgarian
עבריתHebrew
NorskNorwegian
HrvatskiCroatian
CatalàCatalan
SlovenčinaSlovak
LietuviųLithuanian
SlovenščinaSlovenian
СрпскиSerbian
EestiEstonian
LatviešuLatvian
فارسیPersian
മലയാളംMalayalam
தமிழ்Tamil
اردوUrdu
Searching...
SoBrief
May Contain Lies

May Contain Lies

How Stories, Statistics and Studies Exploit Our Biases - And What We Can Do About It
Learn to tell solid evidence from persuasive stories, and spot where a claim outruns its proof.
by Alex Edmans 2024 322 pages
4.11
1k+ ratings
Listen
Immersive
V2.3
Amazon Kindle Audible
Try Full Access for 3 Days
Unlock listening & more!
Continue

Key Takeaways

Audit the claims you want to believe before they become convictions.

Desire bends judgment. Confirmation bias makes people accept agreeable claims too readily, dismiss unwelcome findings, or search only for supportive information. At Deepwater Horizon, engineers reinterpreted repeated failed safety tests as faulty tests because they wanted the well to be secure. The resulting explosion killed eleven workers. The opposite shortcut is black-and-white thinking: treating something as wholly good or bad when its effects depend on amount, type, or context. Carbohydrates, for instance, are not all alike, and excessive water can be dangerous.

Look for the missing shades. The world may be moderate, with benefits that turn harmful beyond a point; granular, with different cases behaving differently; or marbled, with good and bad features mixed together. Before accepting an extreme claim, ask whether it suits a preferred belief and whether it treats a complicated subject as a simple binary.

Analysis

What's striking is that bias is presented as a normal failure of judgment, not a defect confined to gullible people. The engineers had expertise, and the public figures backing Theranos had status and experience; neither protected them when a desired story took hold. That makes self-auditing more useful than sorting sources into trustworthy and untrustworthy camps. The framework also cautions against replacing one dogma with another: nuance is not automatic proof that every position is equally sound. It is a prompt to ask what conditions, degrees, and exceptions a claim leaves out.

Move carefully from statement to fact, data, evidence, and proof.

Each rung adds a different test. The Ladder of Misinference describes four common leaps: treating a statement as fact, a fact as representative data, data as evidence for a cause, and evidence as universal proof. The book's opening example shows how a UK committee repeated a union claim based on an early research draft, even though the published study reached the opposite conclusion. A citation or official report can still mislead through an outdated source, selective wording, or an unsupported inference.

Accuracy is only the beginning. A fact may be cherry-picked; data may fit several explanations; evidence may apply only in a particular setting. The book uses “lie” broadly for misleading claims, including careless omissions and invalid conclusions, not only deliberate falsehoods. Checking the source can catch a false statement, but it cannot by itself establish that a conclusion follows from accurate information.

Analysis

The ladder is useful because it separates mistakes that often get bundled under the label misinformation. A source can be accurately quoted while the surrounding conclusion is misleading; a dataset can be genuine while the interpretation is weak. That distinction directs scrutiny to the exact point where reasoning fails. It also discourages a tempting but shallow response: treating verification as a binary badge. A claim may clear one rung and fail the next, so confidence should be calibrated to the strongest step actually established.

A vivid success story cannot tell you what usually works.

Stories select their evidence. Apple has been explained through Steve Jobs's childhood, his focus on a few products, and a company purpose. These accounts cannot all be a complete explanation, and each can be shaped around the details that fit. The narrative fallacy is the tendency to connect events into a causal story even when other causes, or luck, could explain the outcome. A memorable anecdote may be true and still say little about what happens in general.

Compare with what did not happen. Investors Brad Barber and Terry Odean examined 13,293 frequent traders, not just people who bragged about winning. Their average annual return was 11.4%, while a buy-and-hold market investment returned 17.9%. The relevant comparison was not success versus zero, but success versus the alternative. A useful study therefore needs representative cases with and without the proposed cause, including failures and unexpected successes.

Analysis

Anecdotes are excellent for generating hypotheses and poor substitutes for testing them. Their emotional detail helps readers remember a mechanism, but it also makes an outcome feel inevitable after the fact. The counterfactual question, what would have happened otherwise, is a powerful antidote because it turns a story into a comparison. Still, representative averages do not erase individual variation. A strategy that performs worse on average may suit a particular person, but a single personal win cannot establish that it is generally wise.

Treat a striking result as suspect if researchers searched many routes.

Repeated tests manufacture hits. Data mining occurs when researchers try many inputs, measures, samples, or analytical choices, then highlight only the result that supports their preferred conclusion. In one diversity project, the author and a student tested 24 gender-related measures; most did not show the desired positive relationship with company performance. A separate fund launch relied on a different measure and outcome that produced a more appealing result. If 100 unrelated tests use a 5% significance threshold, about five may appear significant by chance.

Inspect the path, not just the destination. Ask whether the measures are natural choices, whether results hold with reasonable alternatives, and whether they survive tests in another sample. Be cautious when continuous information is split into convenient groups, such as comparing companies with at least three women directors only with firms having none. The discarded middle may be where the inconvenient evidence lives.

Analysis

A compelling result is more informative when the reader can see the choices that produced it. The problem is not that researchers make analytical decisions; some are necessary. It is that flexibility can quietly turn exploration into apparent confirmation. Robustness checks and out-of-sample tests reduce that risk, though they cannot reveal every analysis that was tried and hidden. The practical lesson is proportional skepticism: the more freedom a study had to choose its measures and sample, the less a single clean-looking result should carry the argument.

Ask what else could cause the pattern before calling it causal.

Correlation has rival explanations. Breastfed children in one study scored 4.69 IQ points higher than bottle-fed children, but breastfeeding mothers also differed in education, income, and IQ. After researchers adjusted for those factors, the estimated gap fell to 0.52 points and was no longer statistically significant. Those shared influences are common causes: factors related to both the supposed cause and the outcome. Reverse causation is another possibility, as when a serious health warning prompts someone to quit smoking, making quitting appear associated with worse health.

Seek a source of genuine variation. A randomized control trial assigns a treatment by chance, helping isolate its effect. When direct experiments are impossible, researchers may use an instrument or a natural experiment, but each depends on the input varying for reasons unrelated to the outcome. James Lind's 1747 scurvy trial randomly assigned sailors different remedies; the citrus group recovered quickly. The design, not merely the observed difference, made the finding persuasive.

Analysis

The chapter's central discipline is to treat explanation as a contest among plausible causes. That is more demanding than asking whether two measures move together, but it prevents descriptive patterns from becoming careless advice. Randomization is powerful precisely because it breaks the ordinary links between choices and prior traits. Yet experiments answer only the question their design permits. Ethical limits, imperfect implementation, and narrow samples mean that causal certainty is often partial rather than absolute.

Evidence travels only as far as its setting and sample.

A sound result can have narrow reach. Evidence may establish a cause in the studied context without proving a universal rule. Frederick Taylor's factory methods increased output in tasks such as pig-iron handling, where the work had measurable results and efficient procedures. Applying the same approach to schools encouraged test preparation, scripted lessons, and cuts to untested subjects. Teaching has multiple goals and no single method that fits every teacher and pupil.

Check the range, too. Angela Duckworth's research found that grit predicted outcomes among West Point cadets, spelling competitors, and high-scoring university students. Those groups were already unusually fit or academically strong, so the findings do not establish that grit outweighs ability for everyone. A satirical parachute trial found no benefit when volunteers jumped two feet from aircraft parked on the ground. Before generalizing, ask who was studied, what levels were represented, and why the relationship might change elsewhere.

Analysis

External validity is often treated as a technical footnote, but it is where research meets real decisions. The same intervention can work differently across populations, institutions, or levels of intensity. The examples also expose a subtle trap: an unusually capable sample can make one trait look decisive because other traits are already high. The limit is that narrow evidence remains useful; it just supports a narrower claim. Good judgment neither universalizes a result nor discards it because it is not universal.

Seek opposing arguments, then judge sources by methods, not prestige.

Deliberately seek disconfirmation. The book recommends asking how a study would look if it produced the opposite result, searching for the strongest opposing case, and considering whether the authors would publish an unwelcome finding. Peer review adds scrutiny, but publication is not a guarantee: journals can favor surprising positive results, and even reputable studies can later be challenged. A systematic review is often more informative than one paper because it surveys the broader research and distinguishes agreement from unresolved questions.

Share with care. For books, reports, and articles, examine the authors' expertise, incentives, balance, and access to evidence. A prestigious endorsement is not a method. Before reposting a claim, check whether the study exists and supports the wording. Research cited in the book found that a simple prompt to consider accuracy reduced misinformation sharing, suggesting that people may know how to judge claims but fail to use that ability when sharing.

Analysis

This approach replaces blanket trust with layered trust. Peer review, credentials, and institutional reputation are useful signals, but none should substitute for the underlying argument. Seeking dissent is especially valuable because it expands the information available before judgment, rather than merely polishing a conclusion already chosen. There is a real cost: constant investigation is impossible. The answer is selective attention, with more scrutiny for high-stakes claims, extreme promises, and conclusions that strongly flatter an existing belief.

Build teams where dissent is safe, structured, and genuinely heard.

Difference alone is not inclusion. After the failed Bay of Pigs invasion, President Kennedy changed how his advisers handled the Cuban Missile Crisis. EXCOMM, a smaller committee with varied perspectives, considered several options before narrowing the choice to an air strike or a blockade. Kennedy sometimes stayed out of discussions, reducing pressure to echo his preference. The blockade helped avert a direct military confrontation with the Soviet Union.

Design for independent thought. Organizations can use silent starts, private votes, anonymous idea collection, and red teams that actively challenge a proposal. They can reduce status barriers, invite junior staff to speak first, and reward useful failures. Demographic diversity may broaden experience, but people also need the safety and opportunity to contribute. Research described in the book found that cognitive diversity aided team performance only up to a point; too much difference can make collaboration harder.

Analysis

The organizational lesson is procedural: leaders cannot simply hire different people and expect better decisions to emerge. Meeting order, defaults, titles, and voting methods shape whose information becomes visible. Structured dissent makes disagreement less dependent on personal courage, which is important where hierarchy raises the cost of speaking. At the same time, dissent is not automatically insight. Strong teams need mechanisms for testing objections and integrating useful differences, rather than treating conflict itself as evidence of intelligence.

Teach people to test ideas, not merely memorize facts.

Train the thinking moves. Telling people to be unbiased had little effect in one study, while asking them to consider how they would react to the opposite result reduced biased interpretation. Statistical literacy helps people distinguish a fact from representative data, and data from evidence. Curiosity matters as well: research discussed in the book found that scientific curiosity was associated with less political polarization, while greater scientific knowledge alone could intensify it.

Keep identity out of the way. Climate messages that ridicule political opponents can make the issue a test of group loyalty instead of a question about evidence. Communication may work better when messengers or proposed solutions are credible across political identities. Fact-checking prompts can reduce sharing of false claims, but labeling only some stories as false can make unlabelled stories seem more credible. Balanced checks, including labels for accurate material, help avoid that implied-truth effect.

Analysis

The distinction between knowledge and thinking skill is central: information does not interpret itself, and added expertise can supply more sophisticated defenses for a settled position. Teaching a repeatable question, such as what evidence would change the conclusion, may therefore be more durable than exhorting students to be objective. Public communication faces a related challenge: facts arrive attached to messengers and identities. Evidence-based messaging must account for that social context without tailoring the truth to whatever an audience prefers.

Use evidence to clarify trade-offs, not pretend values have vanished.

Research informs choices, not every reason for choosing. Evidence can estimate likely outcomes, but it cannot decide how much a society should value profit, fairness, health, or personal freedom. A diversity policy, for example, need not depend entirely on proving that diversity raises company returns. Ethical or social goals may matter even when the financial effect is uncertain or negative.

Make disagreement more honest. Two people can accept the same evidence and still choose differently because their priorities differ. Evidence also describes averages, not guarantees for each person. The book's closing counsel is to lower the temperature of disputes, identify shared aims, and listen for useful points in opposing arguments. A policy or personal choice can rest partly on judgment or experience, provided it is not presented as an irrefutable scientific rule.

Analysis

This ending keeps critical thinking from becoming a new form of technocracy. Evidence narrows what can responsibly be claimed, but values determine which outcomes deserve weight and whose costs count. That distinction can make disagreement less threatening: some disputes concern facts, while others concern priorities that no dataset can settle. It also requires candor. A decision based partly on principle or preference is not intellectually weak; disguising that basis as conclusive research is the more serious mistake.

Analysis

The book's deepest contribution is not a checklist of statistical errors but a model of how confidence accumulates faster than justification. A claim becomes a fact through repetition, a fact becomes a pattern through selective examples, and a pattern becomes a causal story through interpretation. Each transition can feel natural, especially when a conclusion is emotionally or politically useful. The Ladder of Misinference makes those transitions visible and gives readers a way to ask what kind of support a claim actually has.

The account is strongest when it connects individual psychology to institutional incentives. Confirmation bias does not operate only inside a person's mind. Committees preserve preferred narratives, companies select favorable metrics, journals reward splashy findings, and social platforms reward shareable certainty. Better reasoning therefore requires both personal habits and designs that make challenge easier: representative samples, independent review, private votes, and public access to sources. A lone instruction to think critically is weak protection when the surrounding system rewards overstatement.

The framework also usefully limits its own ambition. Evidence can show average effects, not dictate values or guarantee individual outcomes. Still, its emphasis on causal testing fits empirical claims better than moral, aesthetic, or interpretive questions. A checklist cannot convert every disagreement into a testable hypothesis, and an imperfect study is not worthless merely because it lacks experimental certainty. The most responsible use of the book is therefore not suspicion toward all claims, but calibrated belief: ask how strong the evidence is, how far it travels, and what remains a matter of judgment. That posture makes skepticism constructive rather than cynical, and turns uncertainty into a reason for precision rather than paralysis.

Summary in 30 Seconds
A true anecdote cannot establish what usually works: compare outcomes across representative cases, including failures and alternatives. Correlation can reflect shared causes or reverse causation; in one breastfeeding study, an initial 4.69-point IQ gap fell to 0.52 after adjustment. Test results are also vulnerable to selective analysis: among 100 tests using a 5% threshold, about five may look significant by chance. Check methods, sample, context, and competing explanations. Evidence informs choices, but values still shape them.
Contains spoilers

Last updated:

Report Issue
Want to read the full book?

Glossary

Confirmation bias

Belief-shaped interpretation

The tendency to search for, interpret, and remember information in ways that support an existing belief or desired conclusion. It includes accepting agreeable claims without sufficient scrutiny, rejecting unwelcome claims through motivated reasoning, and seeking information that confirms an initial hunch while avoiding evidence that could disprove it.

Black-and-white thinking

Binary view of complexity

The habit of treating a thing as wholly good or wholly bad, or as applying in every case. It overlooks moderation, where effects change with amount; granularity, where different forms have different effects; and marbling, where benefits and harms coexist within the same subject.

Ladder of Misinference

Four escalating reasoning errors

A framework for tracing how people overstate what information establishes. It describes four leaps: accepting a statement as fact, treating a fact as representative data, treating data as evidence for a cause, and treating evidence as universal proof. Each step requires additional checks that the preceding step alone cannot provide.

Narrative fallacy

Stories mistaken for causes

The tendency to connect events into a coherent explanation, even when the sequence may reflect other causes or luck. A story about a successful person or company can make one factor seem decisive while excluding failures, alternative explanations, and cases that do not fit the narrative.

Data mining

Searching for convenient results

The practice of trying many analytical choices and reporting the results that support a preferred conclusion while obscuring those that do not. Researchers may vary inputs, outcomes, measures, samples, or group definitions. Even statistically significant findings can arise by chance when enough tests are attempted.

Endogeneity

Input shaped by other factors

A condition in which the input being studied is not independent of other factors that affect the outcome. It can arise because people choose an action, because a measured trait is linked to other traits, or because the input results from another process. Endogeneity makes a correlation insufficient to establish causation.

Randomized control trial

Random assignment test

An intervention study in which participants are assigned by chance to a treatment group or a control group. Random assignment helps make the groups comparable by breaking links between the treatment and other causes of the outcome. Differences in results can then be attributed more confidently to the treatment.

Instrument

External shock to an input

A variable used to isolate changes in an input when researchers cannot assign that input directly. A valid instrument must affect the input, and it must affect the outcome only through that input. Its apparent irrelevance to the outcome can help show that it is not simply another common cause.

Natural experiment

Real-world random division

A situation in which an external event, such as a law change, creates treatment and comparison groups without researchers assigning people to them. It can support causal conclusions when group membership is effectively outside participants' control and the groups would otherwise have followed comparable paths.

External validity

Applicability beyond the study

The extent to which a finding from one study applies to other settings, populations, or ranges of conditions. A study can establish a real effect in its own context while failing to show that the effect holds elsewhere. Differences in goals, participants, or intensity can limit generalization.

Groupthink

Conformity suppressing challenge

A group decision failure that occurs when members value acceptance and harmony enough to suppress doubts or support a presumed consensus. It can leave a team with less information and fewer alternatives than its members collectively possess. Structured dissent and inclusive procedures can reduce the risk.

Cognitive diversity

Varied ways of thinking

Differences in people's backgrounds, experiences, beliefs, information-processing styles, and approaches to solving problems. Such variation can broaden a team's understanding and improve collective reasoning when members can contribute their views. Excessive difference or weak communication can make coordination harder.

Cultural cognition hypothesis

Identity shapes evidence judgments

The idea that people often evaluate information partly through the cultural or political identity it signals, rather than only through its evidential quality. A message associated with an opposing group can provoke resistance, while a familiar messenger or solution may make the same issue easier to consider.

Scaffolding

Prompts that improve reasoning

Real-time questions or cues that help people construct more complete arguments. Prompts can ask someone to develop counterarguments, add reasons, or check whether each point is relevant. The approach supports better reasoning without supplying new facts about the issue itself.

Download PDF

To save this May Contain Lies summary for later, download the free PDF. You can print it out, or read offline at your convenience.
Download PDF
File size: 0.33 MB     Pages: 17

Download EPUB

To read this May Contain Lies summary on your e-reader device or app, download the free EPUB. The .epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.
Download EPUB
File size: 1.38 MB     Pages: 21
Want to read the full book?
Follow
Listen
Home
Swipe
Library
Get App
Try Full Access for 3 Days
Listen, bookmark, and more
Compare Features Free Pro
📖 Read Summaries
Read unlimited summaries. Free users get 3 per month
🎧 Listen to Summaries
Listen to unlimited summaries in 40 languages
—
❤️ Unlimited Bookmarks
Free users are limited to 4
—
📜 Unlimited History
Free users are limited to 4
—
📥 Unlimited Downloads
Free users are limited to 1
—
Risk-Free Timeline
Today: Get Instant Access
Listen to full summaries of 26,000+ books. That's 12,000+ hours of audio!
Day 2: Trial Reminder
We'll send you a notification that your trial is ending soon.
Day 3: Your subscription begins
You'll be charged on Nov 24,
cancel anytime before.
Consume 2.8× More Books
2.8× more books Listening Reading
Our users love us
600,000+ readers
Trustpilot Rating
TrustPilot
4.6 Excellent
This site is a total game-changer. I've been flying through book summaries like never before. Highly, highly recommend.
— Dave G
Worth my money and time, and really well made. I've never seen this quality of summaries on other websites. Very helpful!
— Em
Highly recommended!! Fantastic service. Perfect for those that want a little more than a teaser but not all the intricate details of a full audio book.
— Greg M
Save 62%
Yearly
$119.88 $44.99/year/yr
$3.75/mo
Monthly
$9.99/mo
Start a 3-Day Free Trial
3 days free, then $44.99/year. Cancel anytime.
Unlock a world of fiction & nonfiction books
26,000+ books for the price of 2 books
Read any book in 10 minutes
Discover new books like Tinder
Request any book if it's not summarized
Read more books than anyone you know
#1 app for book lovers
Lifelike & immersive summaries
30-day money-back guarantee
Download summaries in EPUBs or PDFs
Cancel anytime in a few clicks

Why one summary?

Scanner
Find a barcode to scan

⭐
⭐
⭐
We have a special gift for you
Open
38% off
DISCOUNT FOR YOU
⭐
⭐
⭐
$79.99
$49.99/year
Only $4.16 per month
Continue
2 taps to start, super easy to cancel
Settings
General
Widget
Loading...