Contents
Cognitive Biases & Decision-making
Section 1
Core Idea
Typicality — Kahneman and Tversky's representativeness heuristic — is the mental shortcut where people estimate the probability of a case belonging to a category by how well it resembles a mental prototype of that category, rather than by the actual base rate of the category in the world. It sounds abstract; in practice, it is the single most consequential cognitive bias in hiring, investing, and product strategy, because it silently substitutes the question "how likely is this?" with the much easier question "how much does this look like the kind of thing I'm imagining?"
The founding demonstration is the Linda problem, published by Tversky and Kahneman in 1983. Linda is described as thirty-one, single, outspoken, bright, a philosophy major concerned with discrimination and social justice, active in the anti-nuclear movement. Subjects are asked which is more probable: (a) Linda is a bank teller, or (b) Linda is a bank teller and active in the feminist movement. Roughly eighty percent choose (b), even though (b) is a subset of (a) — every feminist bank teller is a bank teller, so (a) must be at least as probable. The description of Linda is a better representative match to the feminist bank teller stereotype, and typicality overrides the logic of conjunction. This is the "conjunction fallacy," and it is a stable finding across expert populations, including physicians estimating disease probability and forecasters estimating geopolitical events.
The reason typicality is so hard to root out is that it usually works. Prototypes are compressed summaries of real correlations. A person in a suit at 3pm in Midtown Manhattan is more likely to work in finance than a randomly-selected New Yorker, and pattern-matching to the "banker" prototype gets you a serviceable estimate in a fraction of the time proper Bayesian reasoning would take. The problem arises in the tail: when base rates are extreme (rare diseases, rare successes, rare hires), the prototype match dominates the estimate even when the actual population contains almost none of the category. That is when the bias produces expensive errors — bad hires, bad investments, bad diagnoses — and it is precisely when smart people are most confident in their intuition.
Section 2
How to See It
Hiring
You're seeing typicality when a candidate is being evaluated against a mental template of "the kind of person who succeeds here" rather than against the base rate of that person's demographic actually succeeding in the role. The strongest predictor of the mistake is that the interviewer can articulate the template but not the base rate.
Investing
You're seeing it when a company is described as "the next Google" or "the Uber of X." Both phrases are prototype-matching in disguise. The correct question — "what percentage of companies that looked like Google at this stage actually became Google?" — is almost never asked, and the answer is a very small number.
Product & Strategy
You're seeing it when a small feature request from a vivid, high-status customer gets prioritised because it "sounds like what an enterprise buyer would ask for," while the aggregate data on what enterprise buyers actually pay for goes unread. Vividness beats frequency; the prototype beats the population.
Section 3
How to Use It
The practical fix is to force the base rate onto the table before you allow the prototype match to speak. This is the single most reliable calibration technique in cognitive science, and it is barely used outside dedicated forecasting communities. Kahneman calls it taking the outside view in Thinking, Fast and Slow: instead of asking "how likely is this specific case to succeed given its specific features?", ask "what fraction of cases in the reference class of things that look broadly similar to this one actually succeeded?" The outside view is almost always more accurate and almost always more pessimistic than the inside view, which is why people resist it.
Decision filter
"What percentage of the reference class of things that look like this actually turn out the way I'm predicting? If I don't know the number, I don't know the probability — I know a story."
The reference class must be constructed carefully. If you're evaluating a Series A SaaS company, the reference class is not "great software companies" — it is "Series A SaaS companies with roughly this revenue profile and roughly this market at roughly this moment." The correct base rate for a Series A SaaS company reaching a billion-dollar outcome is on the order of one to three percent depending on vintage. Any pitch that implicitly assumes it should be evaluated against the base rate of "the next Google" — which itself has a base rate near zero — is asking you to double up on typicality error.
As a founder or hiring manager
Build a reference-class file for every high-leverage decision you make often. For hires: the last twenty people you interviewed for this role, their offer status, and whether the ones you hired worked out. For product bets: the last ten features you shipped and the base rate at which each type of feature moved retention. Once the reference class is written down, prototype-matching still happens — but it now runs against a live base rate rather than an imagined one.
As an investor
When a founder pitches you a comparison — "this is the next X" — treat it as a data request, not a data point. What is the base rate of companies that looked like X at this stage becoming X? What is the base rate of companies that pitched themselves as "the next X" becoming X? Both numbers are tiny, and both are informative. The pitch has told you the founder's mental prototype; the base rate tells you what to actually pay.
Section 4
Common Misapplications
Three failure modes recur.
First, reference-class laziness. The most common mistake in applying the base-rate correction is to choose a reference class that is either too broad or too narrow, and then treat the resulting number as authoritative. "SaaS companies" is too broad. "SaaS companies exactly like ours" is too narrow — the reference class collapses to one. The right reference class is usually somewhere in the middle: broad enough to have a meaningful sample, narrow enough to share the relevant structural features. Constructing it is a judgment call, and the number produced is a range, not a point estimate.
Second, using base rates to dismiss genuine outliers. Base rates are the correct anchor, but outliers exist by definition. The mistake is to conclude that because the base rate for founders under twenty-five reaching a billion-dollar outcome is very low, no such founder is worth backing. The correct posture is that the prior probability is low and the pitch must carry evidence sufficient to overcome that prior. That is Bayes's rule, not typicality's rule. When strong evidence exists — technical proof, customer traction, unusual founder track record — the posterior can and should update far above the base rate. What must not happen is the inverse: updating the base rate itself because the prototype match feels strong.
Third, letting typicality run in reverse. Just as people over-estimate the probability of prototypical matches, they under-estimate the probability of non-prototypical successes. The founder who looks nothing like your mental picture of a founder, the candidate whose resume doesn't fit the template, the business model that resembles nothing you've seen before — these are systematically under-priced by prototype-matching organisations, which is why systematic contrarians (Buffett on unglamorous businesses, Thiel on antisocial founders) can extract durable returns from exactly the same information everyone else has access to.
There is also a domain-specific hazard worth naming. Medicine, law, and finance all suffer from professional-grade typicality bias, because expert intuition is genuinely useful and the categories being pattern-matched are usually correct. That correctness is what makes the error mode invisible in aggregate — the ninety-five cases where the prototype match is right camouflage the five cases where it produces catastrophic misdiagnosis, wrongful conviction, or a mispriced position. Calibration audits are the only reliable way to surface these.
Section 5
Founders & Leaders
Charlie MungerVice Chairman, Berkshire Hathaway
Munger's ruthless preference for base-rate thinking over story-based reasoning is a lifetime application of the outside view. His famous line — that a great business bought at a fair price is almost always a better bet than a fair business bought at a great price — is a rejection of typicality-based investing, where a specific narrative about why this fair business will become great overrides the base rate of fair businesses becoming great.
Jim SimonsFounder, Renaissance Technologies
Simons built Renaissance on the explicit premise that human pattern-matching — the discretionary trader's prototype-based judgment — is systematically inferior to statistical models trained on base rates. The Medallion Fund's returns are, at some level, the world's most expensive proof that typicality is expensive and calibration is cheap once the machinery to enforce it exists.
Section 6
Company Examples
Airbnb
Airbnb was famously turned down by investors who pattern-matched it to the wrong prototype ("strangers won't sleep in strangers' homes at scale"). The correct reference class — early-stage marketplaces with strong repeat-usage economics — would have produced a very different base rate. The story is a canonical case of typicality producing a large miss, and it is one of the reasons Y Combinator explicitly trains partners to build reference classes from within its own portfolio rather than from the broader intuition pool.
NVIDIA
For most of the 2010s, Nvidia was pattern-matched to "graphics-card company," which produced a low prototype match with the AI infrastructure category it was actually building into. The base rate of companies successfully pivoting from consumer graphics to enterprise AI was near zero — but the specific evidence (CUDA adoption, developer traction, Jensen Huang's decade-long bet on parallel computing) should have updated any Bayesian above that base rate. Most public-market investors ran on prototype, not evidence, until the update was priced in explosively in 2023.
Section 7
Connected Models
Produces
Base Rate Neglect
Typicality is the mechanism; base rate neglect is the effect. Every prototype-based estimate has substituted a resemblance judgment for a frequency judgment.
Special case of
Conjunction Fallacy
The Linda problem is typicality at its purest: a more specific story feels more probable than a less specific one, even though logic requires the opposite.
Interacts with
Availability Heuristic
Vivid prototypes come to mind faster and feel more probable. When the prototype is also the easiest example to retrieve, typicality and availability compound.
Corrected by
Bayesian Reasoning
Bayes' rule is the mathematical fix: start from the base rate, update on evidence proportional to its diagnostic value, never let the story override the prior without earning the update.
Section 8
One Key Quote
— Daniel Kahneman, Thinking, Fast and Slow, 2011"When called upon to judge probability, people actually judge something else and believe they have judged probability."
Section 11
Summary & Further Reading
Typicality is the substitution of resemblance for probability. It is the reason "this looks like X" so often becomes "this is likely to be X," even when the base rate of X is near zero. The correction is simple in principle and unnatural in practice: build the reference class, name the base rate, and let the specific evidence update the prior rather than replace it. Operators who make this correction habitual extract durable edge from a world of counterparties who don't.
01
Book
The definitive treatment of the representativeness heuristic and its cousins, from the researcher who spent forty years documenting them. The Linda problem and the outside view are both explained in his own words.
02
Book
The foundational academic collection that established the field. The original Tversky and Kahneman papers on representativeness are the source material every modern treatment builds on.
03
Book
The applied manual for correcting typicality error in real forecasting environments — including detailed instruction on reference-class construction and outside-view discipline.
Why this matters next
mental modelsLeverage
Typicality applied the Leverage mental model
mental modelsNarrative
Typicality applied the Narrative mental model
mental modelsScale
Typicality applied the Scale mental model
mental modelsIntuition
Typicality applied the Intuition mental model
mental modelsDemonstration
Typicality applied the Demonstration mental model
mental modelsEnvironment
Typicality applied the Environment mental model
Frequently asked questions
What is Typicality?
Typicality is a mental model used for better thinking and decision-making.
How do you apply Typicality?
To apply Typicality, identify situations where this framework is relevant, then use it as a lens to evaluate your options and decisions. The model is most useful when combined with other complementary mental models.
What category does Typicality fall under?
Typicality falls under the Systems & Complexity category of mental models. Other models in this category can be found on the Systems & Complexity hub page.
Why is Typicality important?
Typicality is important because it provides a structured way to think about problems that would otherwise be approached with intuition alone. Understanding this model helps you avoid common reasoning errors and make better decisions.
Continue exploring
LEANEMFLGLL(
Mental model
Lollapalooza Effects
When multiple cognitive biases or forces act in the same direction simultaneousl
Mental model
Antifragility
Beyond resilience — some systems actually gain from disorder, volatility, and st
Mental model
Emergence
Complex system-level properties that arise from simple interactions between indi
Mental model
Feedback Loops
Circular causal chains where the output of a system feeds back as input — either
Mental model
Gall's Law
A complex system that works is invariably found to have evolved from a simple sy
Mental model
Leverage (Systems)
Places within a complex system where a small shift produces large changes — Mead
More like this, in your inbox
I send a newsletter every week — free, no spam, unsubscribe anytime.