Tom W is intelligent but unimaginative, likes science fiction, and has a need for order and clarity. Is he more likely to be a humanities major or a computer science major? Almost everyone says computer science. Almost everyone is wrong — because humanities programs enroll far more students. We judge by resemblance and ignore the base rate.
The Framework
The representativeness heuristic is System 1's method for judging probability: instead of calculating how likely something is, it asks "how much does this resemble the prototype?" Tom W resembles the computer science student prototype more than the humanities student prototype, so System 1 confidently assigns him to computer science — completely ignoring the fact that humanities students outnumber CS students by perhaps 5:1. The resemblance judgment (high) substitutes for the probability judgment (low given base rates).
This substitution produces predictable errors. Base rate neglect — ignoring population-level statistical information in favor of individual descriptions — is the most common. The conjunction fallacy is the most dramatic: Linda is described as politically active, concerned with social justice, and a philosophy major. People judge "Linda is a bank teller AND active in the feminist movement" as more probable than "Linda is a bank teller" — a logical impossibility (the conjunction of two events cannot be more probable than either event alone). But "feminist bank teller" resembles Linda's description more than "bank teller" does, and resemblance overrides logic.
Where It Comes From
Tversky and Kahneman introduced representativeness in their 1974 paper. Chapters 14-16 of Thinking, Fast and Slow provide the full treatment: the Tom W problem (Ch 14), the Linda conjunction fallacy (Ch 15), and the distinction between causal and statistical base rates (Ch 16). The key finding from Chapter 16 is that not all base rates are ignored — causal base rates (individual stories that provide a causal mechanism) are used, while statistical base rates (population-level numbers without causal context) are ignored. "85% of cabs in the city are Green" is ignored; "The Blue cab company is involved in a disproportionate number of accidents" changes judgment.
> "When people are asked a probability question, they sometimes answer a quite different question about resemblance instead." — Thinking, Fast and Slow, Ch 14
Cross-Library Connections
Cialdini's social proof in Influence exploits representativeness: when a product is endorsed by someone who resembles you, the resemblance heuristic tags the product as "for people like me" — a representativeness judgment that substitutes for a quality evaluation.
Hormozi's customer avatar work in $100M Offers implicitly manages representativeness: by designing offers that look like what the target customer expects, Hormozi ensures the offer passes System 1's resemblance test. An offer that doesn't resemble the category the prospect expects feels "off" — representativeness creates an immediate credibility check.
Berger's Social Currency principle in Contagious works through representativeness: people share content that makes them resemble the social identity they want to project. The content is judged by "does sharing this represent who I want to be?" — a representativeness judgment, not a utility calculation.
The Implementation Playbook
Hiring: The Tom W problem lives in every interview room. A candidate who looks like your image of a great engineer (quiet, detail-oriented, tech-obsessed) will be rated higher than one who doesn't — regardless of actual skill. The correction: focus on base rates (what percentage of people with this background succeed in this role?) and track record (objective performance metrics) rather than prototype matching.
Marketing: Ensure your branding, messaging, and visual identity match the representativeness prototype of your category — or deliberately violate it if you're positioning as a disruptor. A SaaS tool that looks like a toy won't be taken seriously by enterprise buyers, because it doesn't resemble the "enterprise software" prototype. The prototype match is System 1's first credibility check.
Medical Diagnosis: Doctors make representativeness errors by matching symptoms to the most vivid disease prototype rather than the most statistically likely diagnosis. A rare but dramatic disease that matches the symptom pattern perfectly will be suspected before a common but undramatic disease that's far more probable. Decision support tools that display base rates alongside symptom patterns counteract this.
Risk Assessment: When evaluating the probability of an event, ask: "Am I judging probability by resemblance or by statistics?" If your startup looks like Uber (high resemblance to the unicorn prototype), you may overestimate its chances. The base rate of startups that resemble Uber is still ~99% failure. Resemblance does not equal probability.
Persuasion: To make an argument more convincing, make it resemble the category of "true arguments" in your audience's mind. Use specific numbers, expert citations, and causal mechanisms — not because they prove your point, but because they match the prototype of a credible argument. System 1 evaluates arguments by resemblance to the "convincing argument" prototype before System 2 evaluates their logic.
Key Takeaway
The representativeness heuristic is why stereotypes feel true, why individual stories outweigh statistics, and why conjunction can feel more probable than either of its parts. It's efficient — resemblance is a fast proxy for probability in many contexts — but it systematically ignores the most important statistical variable: the base rate. The correction is brutally simple and perpetually ignored: before judging how likely something is based on how much it resembles the prototype, ask "how common is this category in the first place?"
Continue Exploring
[[Base Rate Neglect]] — The specific error produced by representativeness: ignoring population statistics
[[Conjunction Fallacy]] — The Linda problem: when representativeness overrides the most basic law of probability
[[Substitution Heuristic]] — Representativeness as a specific case of the general substitution mechanism
📚 From Thinking, Fast and Slow by Daniel Kahneman — Get the book