THE REFERENCE LIBRARY
A growing collection of ideas for understanding the world — and your own thinking.
Exploration means trying options to learn about them; exploitation means choosing what seems best from the evidence already available. The balance depends on the cost of experimenting, the uncertainty about alternatives, and how many future decisions could benefit from what you learn.
The explore/exploit tradeoff describes the balance between trying new things (exploration) to gather information and sticking with known good options (exploitation) to maximize reward. Formalized as the multi-armed bandit problem, it applies to decisions in everyday life, technology, medicine, and more, with optimal strategies depending on one's time horizon.
When faced with a business problem like improving government sales, the available options may be unclear. This challenges decision-makers to identify and evaluate potential actions despite ambiguity.
Exponential backoff increases the waiting interval after repeated failures, often by doubling it and choosing a random delay within the interval. It reduces repeated contention and retry traffic. Whether attempts eventually stop is a separate policy; applications to perseverance or forgiveness are analogies rather than consequences of the algorithm.
Methods that worked in the past may become less useful as conditions change. Reassess their assumptions and consider alternatives alongside the approaches you retain.
Extended Analogy is a logical fallacy that occurs when two things that are each analogous to a third are incorrectly concluded to be analogous to each other. A common example is the ad Hitlerum fallacy, which associates someone with Hitler based on a benign similarity.
Thaler and Sunstein recommend paying smaller, affordable losses yourself rather than buying extended warranties. Their On My Own Account sets aside the premiums you save to cover occasional replacements; the choice depends on what a loss would cost you and what the policy covers.
Musashi urges a fighter to coordinate the whole body in an attack, using the sword as an extension of the body and committing fully to the movement.
Extensional cues are manipulations that reduce the conjunction fallacy by prompting individuals to think in terms of class inclusion and relative frequencies rather than representativeness. Examples include asking for constituent category estimates first or using frequency instead of percentage formats.
Paterson advises investigators to keep asking questions after identifying an external factor. Review how the organization assessed a dependency, assigned responsibility, monitored performance, and prepared for a failure.
When an external event contributes to a problem, root cause analysis should also examine how the organization anticipated and responded to it. Supplier oversight, changing circumstances and dependence on consultants can reveal weaknesses that the organization can address.
When an external event causes a problem, also examine the vulnerability that allowed it to do so. Root-cause analysis can identify reasonable precautions, contingency plans, or changes in goals without assuming that every external threat was foreseeable or preventable.
Organizations face external pressures from market demand, politics and law, social and ethical expectations, technology, and competitors. Banasiewicz examines how shared purpose, organizational memory, and limits on information processing shape their responses.
Externalities are costs or benefits of an activity that spill over to others and are not reflected in its price. Taxes, subsidies, or other policies can bring private incentives closer to those wider effects.
Skepticism should not halt at religion's front door; it is appropriate and vital for religious people to apply critical thinking to avoid scams, bad beliefs, and to identify pseudoscientific claims within their faith.
Even extreme or implausible anchors can produce anchoring effects, as demonstrated in studies on estimation tasks and pricing decisions.
A mental model warning that rigid ideological commitment impairs thinking, and that one should avoid holding strong opinions without first being able to articulate the opposing arguments better than their advocates.
Extremistan is Taleb’s name for domains where a single extreme observation can dominate a total, such as wealth, book sales, or citations. Averages and ordinary fluctuations may then give a poor picture of concentration and risk.
Taleb uses Mediocristan for quantities whose extremes have limited influence on the total, and Extremistan for quantities where a single observation can dominate it. In Extremistan, a sample average may change sharply when a rare extreme value appears.
The 80/20 rule illustrates concentration: a small share of inputs, such as 20%, may account for a large share of outputs, such as 80%. Taleb emphasizes that the proportions vary and that identifying the biggest contributors in advance is a separate problem.
Taleb contrasts Mediocristan, where individual observations have limited influence on a large aggregate, with Extremistan, where rare extreme observations can dominate it. The distinction asks how much an average can conceal and whether a thin-tailed model fits the quantity being studied.
Prioritize long-term direction over short-term speed by making decisions that increase your velocity toward your ultimate goal, as exemplified by Mae West's career strategy.
Having enough savings to act independently without financial pressure, allowing one to choose work based on passion rather than necessity. It frees you from outside authority and prostituting your mind.
In the described experiment, facial expressions during recall affected self-judgments even though recall demands were held constant. Smiling was associated with higher assertiveness ratings after recalling assertive behavior and lower ratings after recalling unassertive behavior, consistent with using felt ease as information.
Taleb argues that demonstrated capability can influence an opponent more strongly than verbal threats. He develops the argument through a fictional Mafia warning, legends about the Assassins, and his own accounts of photographing abusive strangers. He also argues that public web-shaming carries risks beyond the immediate confrontation.
The application of statistical heuristics can be hindered by biases such as base rate neglect and sample size neglect.
Kahneman’s team learned that comparable curriculum projects took far longer than they had expected. They recognized the problem but kept going without discussing a new forecast, partly because of sunk costs and the embarrassment of stopping.
Failure of imagination leaves planners unprepared for scenarios they have not seriously considered. The source uses the Titanic’s design and the 9/11 attacks to discuss the limits of a reasonable worst-case scenario.
The experiential system's focus on immediate affect and sensitivity to small changes leads to poor long-term decisions, as illustrated by smoking addiction where young smokers underestimate future cravings and risks.
The Failure to Elucidate fallacy occurs when a definition is more difficult to understand than the term it is meant to clarify, often relying on obscure language, circular reasoning, or unhelpful synonyms.
A logical fallacy where implausibly precise statistics are used to give an appearance of truth and certainty, or where negligible differences in data are used to draw incorrect inferences.
Build a learning process that includes manageable attempts, feedback, and revision. Fear of failure need not stop disciplined practice, but mistakes help only when you examine them and change your approach.
A category of logical fallacies where arguments introduce unwarranted or irrelevant elements, violating the principle of not claiming more than is known.
A fallacy is an error in reasoning where a conclusion does not properly follow from its premises, often occurring unintentionally or with deceptive intent.
A categorical syllogism commits the undistributed-middle fallacy when its middle term is not distributed in either premise. Sharing a category does not establish the relationship asserted in the conclusion.
The fallacy of composition occurs when one incorrectly infers that the characteristics, attributes, or features of individuals comprising some group will also be found in the group as a whole.
The fallacy of division occurs when it is incorrectly inferred that characteristics of a group apply to its individual members or parts.
A categorical syllogism links two terms through a shared middle term. The fallacy of four terms occurs when an argument presented in this form contains a fourth term—sometimes because the supposed middle term changes meaning between premises—so the intended link fails.
The fallacy of necessity confuses a conclusion that follows from the premises with a claim that the conclusion must hold in every possible circumstance. Ordinary facts can support a valid deduction without becoming necessary truths.
The fallacy of the false cause occurs when an argument assumes a causal relationship between two events without sufficient evidence, often mistaking correlation or coincidence for causation.
This fallacy assumes that an event has only one cause, ignoring the possibility of multiple contributing factors. It oversimplifies complex situations by attributing a single reason.
A prediction that an event will happen leaves out how much it could matter. In fat-tailed settings, estimating the probability of a war or financial loss does not tell you the scale of its consequences.
The fallacy treats speculation about the future or an unobserved alternative as an established fact. A forecast or counterfactual is not automatically fallacious; its uncertainty and supporting evidence matter.
The false attribution fallacy occurs when an argument appeals to an irrelevant, unqualified, unidentified, biased, or fabricated source to support a claim, thereby weakening the argument's credibility.
The false cause fallacy occurs when a real or perceived relationship between variables is taken as evidence that one causes the other, often confusing correlation with causation.
The fallacy of ignoring common cause occurs when a correlation between two variables is assumed to be causal, while a third variable that causes both is overlooked. This leads to mistaken causal inferences and is a form of the false cause fallacy.
The false consensus effect is a cognitive bias where people overestimate how much others share their own beliefs, attitudes, and behaviors. In social situations, this leads to assuming that one's own reactions are universal, as illustrated by the author's surprise that many people feel less awkward when awkwardness is acknowledged.
False context presents real material, such as a video or quotation, with a misleading account of what it shows. A shortened clip and a false caption can make ordinary events appear to support a claim that the full context contradicts.
False Conversion is a formal fallacy where the subject and predicate of a proposition are incorrectly switched in the conclusion, specifically for propositions of type 'A' (all) or type 'O' (some...not).
A false dichotomy presents two options as the only possibilities when there are actually more. It is a logical fallacy that often arises from a false equivalence between A and not-B, when in reality neither or both might be true.
A false dilemma presents a limited set of alternatives as exhaustive when other relevant possibilities exist. It often takes an either-or form, but the list can contain more options. A genuinely restricted choice or a stated boundary is not, by itself, a false dilemma.
The false effect fallacy occurs when an effect is incorrectly assumed from a cause, often based on a mistaken belief about what the cause produces, without proper evidence or justification.
False equivalence treats statements, positions, or situations as interchangeable while overlooking differences that matter. Cheng’s examples show how this can distort a nuanced position and make it easier to attack.
An analogy can mislead when it treats two situations as alike in a respect that does not support the argument. Identify the principle behind the comparison and ask whether the relevant similarities hold.
Accusing someone of false equivalence when their analogy is actually valid is itself a fallacy. The burden is on the critic to show that the analogy breaks down on a crucial point, not merely that it is not identical.
The fallacy of assuming a linear relationship between variables when the true relationship is nonlinear, often leading to flawed reasoning that more or less of something is always better, ignoring optimal points.
A mental model that examines decisions through the lens of false positives (undeserved outcomes) and false negatives (missed deserved outcomes). By abstracting arguments to these two types of errors, it clarifies disagreements in areas such as social policy, personal risk-taking, and even jetlag management.
A logical fallacy where old footage or information is presented as if it were current, misleading audiences about its true context and meaning.
The principle that a scientific theory must be able to be proven wrong by conceivable observations; if it cannot be falsified, it is not empirical science.
A falsifiable hypothesis can conflict with possible observations. Seeking such observations helps test an idea more seriously than collecting only examples that fit it.