Capítulo 1
The Truth-Seeking Compass in a Post-Truth World
In a world where falsehoods spread six times faster than truth on social media, Alex Edmans offers a crucial navigation tool. "May Contain Lies" has emerged as one of the most important books of our time, earning praise from Nobel laureates and business leaders alike. Bill Gates included it in his "5 books to help you think more clearly" list, while Daniel Kahneman called it "essential reading for anyone who wants to make better decisions." The book's cultural impact extends beyond academia-it's become required reading in corporate boardrooms and government policy circles, with its frameworks adopted by organizations ranging from McKinsey to the World Economic Forum.
What makes this book particularly compelling is how Edmans transforms complex epistemological concepts into accessible wisdom. Drawing from his unique position as both a finance professor at London Business School and former Morgan Stanley investment banker, Edmans doesn't just tell us to think critically-he shows us precisely how our minds get tricked and what practical steps we can take to see through the fog of misinformation. Have you ever wondered why smart people believe ridiculous things? Or why presenting facts often makes people more entrenched in their positions? The answers lie within these pages.
Capítulo 2
The Seductive Power of Confirmation Bias
Belle Gibson's story seems almost impossible to believe in hindsight. This young Australian woman claimed to have cured her terminal brain cancer through natural methods after conventional medicine failed her. Her wellness app "The Whole Pantry" rocketed to #1 in Apple's App Store with 200,000 downloads, was featured prominently on the Apple Watch, and spawned a bestselling cookbook. She earned A$420,000 in just eighteen months.
There was just one problem: Belle never had cancer. Not even a mild case. Her entire story was fabricated.
Why did so many people-including sophisticated technology companies, publishers, and media outlets-believe her without verification? The answer lies in confirmation bias, our tendency to uncritically accept claims that align with what we want to believe. Belle's story resonated because it suggested anyone could overcome deadly illness through determination and lifestyle changes-a comforting narrative in a world where cancer remains frightening and often uncontrollable.
The consequences were devastating. Several cancer patients abandoned chemotherapy to follow Belle's methods, with at least one dying within months. But confirmation bias doesn't just affect desperate individuals. Even billionaires like Rupert Murdoch and Larry Ellison fell victim when investing in Theranos, Elizabeth Holmes's fraudulent blood-testing company. Despite implausible claims, they poured in over $700 million without proper due diligence, captivated by Holmes's compelling narrative as a young visionary saving lives.
Confirmation bias operates through two mechanisms: "naive acceptance" (believing claims we like without verification) and "blinkered skepticism" (rejecting inconvenient truths through motivated reasoning). The Deepwater Horizon disaster exemplifies the latter-engineers ignored failed pressure tests and invented a non-existent "bladder effect" to justify proceeding, resulting in America's worst oil spill that killed eleven workers and damaged 400 species' habitats.
Neuroscience reveals why this bias is so powerful. When our beliefs are challenged, our amygdala activates-the same region triggered during physical threats. This "fight-or-flight" response overrides our rational prefrontal cortex. Even more telling, when we successfully dismiss contradictory evidence, our striatum releases dopamine-the same pleasure chemical triggered by food, exercise, or sex. No wonder motivated reasoning feels so satisfying!
Capítulo 3
Black-and-White Thinking: The Cognitive Shortcut That Leads Us Astray
When Robert Coleman Atkins introduced his revolutionary diet in 1972, it became the bestselling weight-loss book in history, selling 15 million copies despite containing no scientific references or evidence of long-term safety. What made it so compelling wasn't scientific rigor but rather its simplicity: avoid all carbohydrates. This single, straightforward rule eliminated the complexity of distinguishing between different types of carbs or calculating percentages of daily calories.
The Atkins diet perfectly exploited our natural tendency toward black-and-white thinking-seeing carbs as entirely sinful and protein/fat as entirely virtuous. Unlike confirmation bias which reinforces existing beliefs, black-and-white thinking makes us susceptible to extreme claims regardless of our prior views. Atkins didn't need to be scientifically correct; he simply needed to be extreme.
This binary thinking stems from our evolutionary past. Our hunter-gatherer ancestors needed quick, decisive rules to survive-"berries are good" or "run from carnivores." Simple heuristics enabled rapid decisions in life-or-death situations. While occasionally leading to mistakes, these black-and-white rules were evolutionarily advantageous. Today, though we face fewer immediate threats, binary thinking persists because it helps us process information efficiently.
The problem is that reality is rarely binary. Most relationships in life are moderate (beneficial up to a point before becoming harmful), granular (categories contain diverse elements with different properties), or marbled (things contain both positive and negative aspects simultaneously). Studies show the Atkins diet's near-zero carb approach can reduce life expectancy by four years. Similarly, the belief that "more water is always better" led marathon runner David Rogers to fatal water intoxication.
Our categorical thinking affects decisions in surprising ways. Studies show people refuse to drink from clean bedpans or eat sterilized cockroach-dipped juice because our brains rigidly categorize items as "clean" or "dirty." Even experts like weatherman Willard Scott exhibit this thinking, expecting sudden temperature changes between months rather than gradual transitions.
Together, confirmation bias and black-and-white thinking create a perfect storm of cognitive vulnerability. When we have prior views, confirmation bias leads us to accept evidence that supports them; when we don't, black-and-white thinking makes us susceptible to extreme claims that offer simple, absolute answers.
Capítulo 4
The Ladder of Misinference: When Statements Aren't Facts
I first encountered Malcolm Gladwell's famous 10,000-hours rule through a Cirque du Soleil performer who confidently told me anyone could master any skill with sufficient practice. This resonated deeply-it aligned with encouraging messages I'd heard since childhood about hard work trumping natural talent. The rule seemed validated by reputable sources: a medical journal, Fortune magazine, and Gladwell's bestseller "Outliers," which cited research on violinists at Berlin's elite Academy of Music showing the best performers had accumulated 10,000 hours of practice.
For years I taught this rule to my Wharton students until I decided to scrutinize the original sources. What I discovered was shocking: Gladwell had only claimed practice was necessary, not sufficient for success-talent still mattered. Worse, when I examined the actual Ericsson study, I found the "best" violinists had only accumulated 7,410 hours by age 18, and the 10,000 figure wasn't even mentioned. The study relied on participants' questionable recollections of practice from age five, and even the "future teachers" would eventually reach 10,000 hours. Ericsson himself later clarified that international competition winners typically practiced 20,000-25,000 hours.
I had spread misinformation because I wanted the rule to be true, failing to recognize that a statement is not necessarily fact. This happens constantly in public discourse. When the UK Select Committee on Business investigated executive pay, they cited Edmans as evidence that "the impact of individual CEOs on company performance" was "at best ambiguous"-the exact opposite of what his submission actually stated. This wasn't likely deliberate deception but confirmation bias: the Committee read what aligned with their preexisting belief.
Misrepresentations can often be spotted without reading entire papers. A Forbes piece claimed Jensen and Meckling's famous paper advocated "unbridled self-interest" and maximizing shareholder value, but the author couldn't even get their names in the right order (writing "Meckling and Jensen"), revealing he likely never read the paper. The actual paper's first graph shows companies should balance shareholder value with "non-pecuniary benefits" like charitable contributions-hardly advocating unbridled self-interest.
Sometimes the problem isn't bad data but no data at all. In 2019, organizations claimed "CEO remuneration packages actively discourage innovation" without a single test supporting this conclusion-they merely gathered pay data and assumed bonuses discourage innovation. Even worse are non-existent studies, like when Reuters published "Boardrooms with more women deliver more on climate" citing an Arabesque study that couldn't be found anywhere.
Capítulo 5
Cherry-Picked Examples: Why Facts Aren't Data
Walter Isaacson's biography of Steve Jobs emphasizes Jobs' adoption, design sensibility, and focus on the "how" of product creation as keys to Apple's success. Meanwhile, Simon Sinek's popular "Golden Circle" theory claims Apple succeeded because it started with "why" rather than "what" or "how." Despite their fundamental contradictions, both explanations gained widespread acceptance because they exploit black-and-white thinking, feed confirmation bias, and present compelling narratives that reverse-engineer Apple's success into seemingly logical, predestined outcomes.
The problem is that selected examples don't constitute data. While YouTube personalities like "Reyes the Entrepreneur" boast of massive stock market profits, finance professors Terry Odean and Brad Barber's rigorous study of 78,000 brokerage accounts revealed that frequent traders earned only 11.4% annually compared to 17.9% for buy-and-hold investors. Individual success stories aren't representative of typical outcomes.
The scientific method requires forming a hypothesis about how an input affects an output, gathering representative samples rather than selected ones, establishing control groups for comparison, calculating average outcomes, and testing for statistical significance. This contrasts sharply with how authors like Isaacson draw broad conclusions from single cases. Belle Gibson's cancer story and Anders Ericsson's 10,000-hour rule both suffer from selection bias-they highlight outlier cases while ignoring countless contrary examples.
We readily accept compelling stories that exploit our biases, even when they lack scientific rigor. Business books typically present a single big idea illustrated with cherry-picked examples that fit their narrative, while ignoring contradictory cases. Business school case studies similarly reverse-engineer explanations for company successes without testing whether other companies with the same traits had similar outcomes.
Even when researchers attempt to avoid bias by starting without preconceptions, they often fall into the trap of selected samples. Jim Collins and Jerry Porras identified "visionary" companies and compared them to less successful competitors in "Built to Last," but this approach doesn't constitute a proper control group. A true control would examine hundreds of companies without their prescribed principles, not just companies without success. This methodological flaw explains why many "visionary" companies highlighted in such books subsequently failed-the supposed success formula was never scientifically validated.
Capítulo 6
Data Mining: Finding Patterns That Don't Exist
Data mining occurs when researchers conduct biased searches for particular conclusions by running numerous tests and only reporting those that yield desired results. Even with no true link between variables, running enough tests will eventually produce statistically significant results by pure chance - a phenomenon known as "p-hacking." This fundamentally undermines the scientific method, which requires forming hypotheses before testing them and reporting all results, not just favorable ones. The probability of finding at least one "significant" result increases dramatically with each additional test, reaching nearly 100% after enough iterations.
I witnessed this firsthand as an investment banking analyst creating "league tables" for a client pitch. When my initial analysis showed Morgan Stanley ranked third among competitors, my boss pushed me to manipulate the sample criteria-adjusting time periods, deal sizes, and geographic scope-until we achieved first place. We tried excluding deals below $500 million, then focused only on European transactions, then limited the timeframe to the past 18 months - continuing until we found the perfect combination to showcase our bank at the top. Similarly, Thomson Reuters selectively used data from 2007 onward to show companies with women on boards outperformed others, while the full dataset from 2002 showed no significant relationship. Their ironically titled study "Mining the metrics of board diversity" demonstrated blatant sample mining to produce PR-friendly results. This selective use of data points is particularly problematic because it creates misleading narratives that can influence important business and policy decisions.
Researchers can also manipulate data through regression analysis versus grouping techniques. Using a dataset with companies having different numbers of female directors and varying profit levels, proper regression analysis might find a slight negative relationship (-0.1) between female directors and profits, though statistically insignificant. In contrast, one could manipulate results by "grouping"-comparing only companies with three or more female directors to those with none, ignoring inconvenient middle cases that don't fit the desired narrative. This grouping technique allows researchers to claim diversity improves performance by several percentage points while concealing that within diverse companies, more diversity doesn't necessarily correlate with better performance. The choice between regression and grouping analysis often reveals more about the researcher's intentions than about the underlying data relationships.
To guard against data mining, we should implement several critical checks. First, examine whether inputs and outputs are measured in natural, unforced ways rather than artificially constructed categories. Second, verify whether the study checks for robustness across alternative measures and time periods. Third, evaluate whether there's a plausible causal relationship between the variables based on common sense rather than post-hoc rationalization. Fourth, look for pre-registered study designs where researchers commit to their methodology before collecting data. The best defense is having a strong hypothesis based on the most logical input for the question being explored, like the author's research on how sports outcomes affect stock markets, which used football results as a perfect measure that strongly affects mood without impacting the economy. This approach provides a clean test of the relationship between investor sentiment and market performance without the confounding variables present in many other studies.
Capítulo 7
Correlation Is Not Causation: The Breastfeeding Example
When my son Caspar was born underweight, we faced a difficult decision about whether to supplement breastfeeding with formula. Conventional wisdom and initial Google searches suggested breastfeeding was clearly superior for outcomes like IQ, with multiple studies appearing to support this conclusion. When I called an NCT breastfeeding counselor for advice, she adamantly advised against formula despite not being medically qualified.
This prompted me to examine the research more carefully, where I discovered that many studies failed to account for "common causes"-factors like mother's IQ, education level, and socioeconomic status that influence both breastfeeding choice and child outcomes. After reading a British Medical Journal study that controlled for these factors, I found the supposed 4.69 IQ point advantage of breastfeeding dropped to an insignificant 0.52 points. This revelation led us to adopt combination feeding, supported by further research from Emily Oster's book "Cribsheet" that debunked many claimed benefits of exclusive breastfeeding.
Correlation isn't causation because common causes may be driving both variables. Even with robust methodology-starting with a hypothesis, using random samples, and avoiding data mining-studies can fail to show causation. Data alone is just a collection of facts; evidence distinguishes between hypotheses by ruling out alternatives.
The critical distinction between correlation and causation matters because people often leap from description to prediction. While it's accurate to say "breast-fed babies have higher IQs," it's dangerous to conclude that switching to breastfeeding will boost a child's intelligence. Even prestigious organizations make this error-McKinsey's study claimed companies acting long-term deliver better performance, extrapolating to predict $3 trillion in economic growth if all companies followed suit. But industry factors likely drive both investment levels and performance outcomes.
To properly control for common causes, researchers use statistical techniques like regression analysis. Regression allows you to control for multiple factors simultaneously by including them as "controls" in your equation. The regression then shows how much the output changes when you increase a single input without changing any other inputs. However, regressions can only control for observable factors-like searching for keys under a lamppost rather than in dark bushes where they might actually be.
Capítulo 8
From Data to Evidence: When Correlation Does Imply Causation
To establish causation rather than mere correlation, researchers developed randomized control trials (RCTs). This approach was pioneered in 1747 by James Lind, who randomly assigned different remedies to twelve sailors with scurvy. By making treatment assignment exogenous (random) rather than endogenous (chosen based on factors related to recovery), Lind eliminated common causes as explanations. The sailors given citrus fruits recovered dramatically faster, providing the first evidence that citrus cured scurvy.
RCTs evolved further when Austin Flint discovered the placebo effect in 1863. Flint gave patients with rheumatism a diluted quassia extract with no medicinal properties and found they improved just as much as those given conventional drugs. This led to modern "blind" RCTs where the control group receives a placebo, ensuring subjects don't know whether they're receiving the treatment.
When RCTs are impractical due to cost or ethical concerns-you can't force people to smoke to test cancer causation-researchers use instruments: real-world factors that cause input changes for reasons unrelated to outputs. Caroline Hoxby ingeniously used rivers as an instrument to study school choice effects. Since 18th-century school districts rarely crossed rivers, metropolitan areas with more rivers naturally had more school districts. By isolating this exogenous variation in school choice (caused by rivers) from endogenous factors (like parental engagement), Hoxby showed that areas with more school districts had better educational outcomes.
Natural experiments provide another path to evidence when events naturally create test and control groups. Card and Krueger's Nobel Prize-winning research on minimum wage used this approach, comparing New Jersey fast-food employment (where minimum wage increased) with Pennsylvania's (where it remained unchanged). Their difference-in-differences calculation showed employment surprisingly increased after the wage hike.
When valid instruments and natural experiments aren't available, researchers can use common sense tests to strengthen causal claims. These tests either support your theory or rebut rival explanations. In Edmans' football study, he supported the investor sentiment theory by showing losses affected small stocks (held by local investors) more than large ones. He also rebutted the rational explanation by showing pre-game odds didn't affect market declines.
Capítulo 9
Evidence Is Not Proof: The Scientific Management Fallacy
Frederick Winslow Taylor's scientific management revolution in manufacturing promised a universal approach to efficiency. Taylor meticulously analyzed factory tasks, breaking them into components to find the "one best way" to perform each operation. His methods dramatically improved productivity-quadrupling pig-iron handler Schmidt's output while increasing his wages by 61%. Taylor's 1911 book "The Principles of Scientific Management" became the most influential management book of the 20th century.
Taylor's principles were eagerly applied to education in the early 20th century. Professor John Franklin Bobbitt advocated treating teachers like factory workers and students like metals to be shaped. This approach culminated in the 2001 No Child Left Behind Act, which linked standardized test scores to teacher pay, school funding and even school survival. The result was devastating-schools narrowed curricula to focus on tested subjects, teachers followed scripted lessons rather than adapting to student needs, and education became mechanistic rather than inspiring. Experienced teachers quit in droves, frustrated by the loss of autonomy and creativity.
Scientific management failed in education despite succeeding in manufacturing because of three critical differences. First, unlike iron tonnage, educational output can't be standardized across different classrooms with different student populations. Second, teaching produces multifaceted outcomes beyond test scores-critical thinking, love of learning, and respect for different viewpoints. Finally, unlike metal cutting, effective teaching depends on the specific teacher-student relationship, not a universal "one best way."
Evidence is not proof because it lacks universality. While mathematical proofs like Archimedes' circle formula apply everywhere forever, evidence may only be valid in the specific context where it was gathered. Even if evidence has internal validity (uncovers causation), it may not have external validity (apply in different settings).
Angela Duckworth's research at West Point revealed that "grit"-a combination of passion and perseverance-predicted which cadets would complete the grueling Beast Barracks training better than the Whole Candidate Score. She extended this finding to National Spelling Bee contestants and University of Pennsylvania undergrads, claiming grit "beats the pants off IQ" across contexts. However, these studies suffer from restriction of range-they only examined people who were already exceptional (West Point cadets, spelling bee finalists, Ivy League students). This creates two problems: the control variables (fitness, IQ) were already so high they became irrelevant, artificially elevating grit's importance; and grit may only matter when combined with high ability.
Capítulo 10
Building a Smarter Society: From Individual to Collective Wisdom
Thinking smarter as individuals requires actively seeking dissenting viewpoints. I almost skipped a Brexit talk by Roger Bootle because he was a Leave supporter, contrary to my Remain position. Despite initial reluctance, attending proved eye-opening. Becoming more knowledgeable isn't just about defending against misinformation but actively gathering diverse perspectives.
The peer-review process serves as a crucial certification mechanism for academic studies, similar to how regulatory bodies approve medicines. When papers are submitted to scientific journals, editors ask leading scholars to evaluate their quality, with elite journals rejecting up to 95% of submissions. This rigorous process helps readers have confidence in published results. The Theranos example demonstrates the real-world importance of proper scientific vetting-Elizabeth Holmes's fraudulent claims went unchallenged because none were peer-reviewed.
Organizations need more than individuals reading scientific research; they require diverse thinking to ensure the whole exceeds the sum of its parts. The Cuban Missile Crisis of October 1962 demonstrates this principle. When President Kennedy learned of Soviet missiles in Cuba, he assembled EXCOMM, a diverse fourteen-member committee to address the crisis. Unlike the disastrous Bay of Pigs invasion where groupthink prevailed, Kennedy ensured EXCOMM included diverse viewpoints. The committee ultimately chose a naval blockade and diplomatic ultimatum rather than military strikes, averting potential nuclear war.
Cognitive diversity-variety in backgrounds, experiences, beliefs, and problem-solving approaches-proves crucial for effective decision-making. Research by Aggarwal and colleagues confirms that teams with diverse cognitive styles demonstrate higher collective intelligence. Building cognitively diverse groups requires looking beyond simple demographics to consider age differences, varied career paths, and socioeconomic backgrounds.
Diversity alone isn't enough-inclusion is essential. Many companies take an "add diversity and stir" approach, expecting performance improvements simply from recruiting diverse individuals. What matters is creating conditions where diverse colleagues feel comfortable sharing different viewpoints. Studies found that DEI is associated with higher future performance, while demographic diversity alone is not.
Creating smarter societies requires teaching critical thinking skills in schools through specific techniques like "consider the opposite"-training students to be skeptical of weak studies, embrace different viewpoints, and challenge their own theories. Research shows teaching elementary statistics improves judgment across various everyday problems. While statistical literacy provides the means to avoid confirmation bias, cultivating curiosity provides the motivation to overcome it.
