Represent-
ativeness
heuristic.

Does it look like X? Then it probably is X. This shortcut is useful. Until it is not.

The Linda problem

In 1983, psychologists Daniel Kahneman and Amos Tversky posed a question to hundreds of students: Linda is 31 years old, single, outspoken, and deeply concerned with social issues. She studied philosophy. As a student, she was active in anti-nuclear demonstrations. Which is more likely: that Linda is a bank teller, or that Linda is a bank teller active in the feminist movement?

The majority said feminist bank teller. But this is logically impossible. Every feminist bank teller is a bank teller. The statement "Linda is a bank teller active in the feminist movement" is a subset of "Linda is a bank teller." A subset cannot be more likely than the set it belongs to. And yet the description feels like it belongs together. The pieces represent a coherent stereotype — philosophical, politically engaged, progressive — and our brains reach for the coherent story rather than the statistical fact.

This is the representativeness heuristic in action. You evaluate probability by asking: does this case represent the category? Does Linda represent "feminist"? If yes, she must fit the category. But representation is not probability. A person can match a stereotype in every visible dimension and still not belong to the category — because membership is determined by base rates, not by surface features.

Base rate neglect

The core failure in representativeness thinking is ignoring base rates. Base rate is how common something is in the population. Lawyers are more common than attorneys. Physicians are more common than cardiac surgeons. But when you meet someone who speaks precisely, dresses formally, and asks probing questions, you may think "lawyer" or "doctor" — the high-status professions that match the prototype — rather than the statistically more common alternatives.

Medical diagnosis is saturated with representativeness errors. A patient presents with fatigue, weight loss, and abdominal pain. The classic textbook presentation of pancreatic cancer fits perfectly. The doctor thinks pancreatic cancer. But pancreatic cancer is rare. The same symptoms much more likely indicate a dozen more common conditions. The representativeness of the match outweighs the base rate of the disease, and the patient gets misdiagnosed.

This is why Bayesian reasoning — explicitly incorporating base rates into judgment — produces more accurate diagnoses than pattern-matching alone. The doctor who asks "how common is this, and how well does this presentation match the common versus the rare conditions?" outperforms the doctor who relies on "does this look like what I learned in medical school?"

Conjunction fallacy

The Linda problem is also a conjunction fallacy: the error of believing that a specific combination of events is more likely than any one of those events alone. "Feminist bank teller" is a conjunction of two events: "bank teller" AND "feminist." It cannot be more likely than "bank teller" alone. But because "feminist bank teller" is more representative of Linda's description than "bank teller" alone, we assign it higher probability.

Conjunction fallacies appear in financial markets constantly. An investment narrative that combines several compelling elements — "the CEO is visionary, the market is expanding, the technology is disruptive" — feels more credible than a simpler story. But adding elements to a story reduces its probability, because each additional claim must be true for the whole story to be true. The most compelling narrative is often the least probable.

When evaluating any claim, the first question should be: what is the base rate? The second question: does this description actually distinguish this case from the base rate, or does it just feel distinctive? The feeling of distinctiveness is not evidence of probability. It is evidence of a good story.

Stereotypes as representativeness

At its deepest level, representativeness is the cognitive mechanism underlying stereotyping. A stereotype is a mental model of what a category looks like. When you encounter a member of a category, you compare them to the prototype. If they match the prototype, they are accepted as representative. If they do not, they are marked as atypical. This is automatic and below conscious awareness.

Stereotypes persist not because people are malicious, but because the representativeness heuristic is efficient and automatic. The brain needs shortcuts. The shortcut of "category X looks like Y" works often enough to be reinforced. Even when the shortcut fails — even when it produces discrimination, misdiagnosis, or bad hiring decisions — it feels accurate, because the feeling of accuracy is generated by the heuristic itself, not by external validation.

Breaking out of representativeness errors requires deliberate effort. You must learn to ask: what is the base rate? What is the probability of the conjunction versus the components? Is this judgment based on actual data, or on how well this case fits my mental prototype? These questions feel cold and mechanical, but they protect against the warm feeling of confident wrongness.

Applications in practice

In hiring, representativeness errors create homogeneity. Candidates who match the prototype of "successful employee" — shaped by whoever has historically been successful in the organization — get hired over candidates who are statistically more likely to perform well but do not match the prototype. This is not prejudice in the malicious sense. It is the natural output of a cognitive shortcut applied to an environment it was not designed for.

In investing, representativeness causes people to buy stocks of companies with stories that fit a compelling narrative — the visionary founder, the massive addressable market, the revolutionary technology. The narrative is representative of "great investment." But great investments are not defined by how representative they are of great investments. They are defined by outcomes. And the base rate of good outcomes for compelling narratives is not higher than the base rate for boring narratives. In fact, boring narratives often have better outcomes, because the excitement premium has not inflated the price.

The cure for representativeness bias is not to stop using mental models — you cannot — but to remember that your mental model of a category is a simplification, not a definition. The world is messier than the prototype. People who remember this make better predictions, better diagnoses, better hires, and better investments. Not because they are smarter. Because they know that "looks like" is not the same as "is."