Judging probability by how closely something matches a mental stereotype rather than by actual statistics.
Imagine you have a bag full of 90 red marbles and 10 blue marbles. Someone pulls out a marble and tells you it 'feels smooth and cool.' You might think, 'That sounds like a blue marble!' and guess blue — even though there are way more red marbles in the bag. You're matching the description to what you think a blue marble is like, instead of remembering that almost all the marbles are red. That's the representativeness heuristic: you go by what something looks like, not by what's actually most likely.
The representativeness heuristic leads people to estimate the probability that something belongs to a particular category by assessing how similar it looks to their mental image of that category, while systematically ignoring relevant statistical information such as base rates, sample sizes, and prior probabilities. This creates a cascade of predictable errors: people neglect how common a category actually is in the population, assume small samples mirror the larger population perfectly, fall prey to the conjunction fallacy by rating detailed scenarios as more likely than simple ones, and expect random sequences to 'look' random rather than understanding true statistical randomness. The heuristic is deeply tied to pattern-matching and categorization processes, making it one of the most pervasive and consequential mental shortcuts in human judgment under uncertainty.
The same glitch looks different depending on the terrain. Finance, medicine, a relationship, a team — same mechanism, different costume.
Investors tend to categorize stocks based on how closely a company's narrative resembles past successes or failures (e.g., 'the next Amazon'), leading them to ignore base rates of company survival, overweight vivid growth stories, and chase patterns in random market fluctuations as though they signal meaningful trends.
Clinicians may diagnose conditions based on how well a patient's symptoms match a textbook prototype of a disease, while underweighting the actual prevalence (base rate) of that disease in the patient's demographic — leading to overdiagnosis of rare but 'classic-looking' conditions and underdiagnosis of common but atypical presentations.
Teachers may form expectations about a student's academic potential based on how closely the student matches their prototype of a 'gifted student' (articulate, well-dressed, eager), leading them to overlook high-potential students who don't fit the stereotype and to overestimate students who do.
People judge potential romantic partners by how well they match an idealized prototype — someone who 'looks like' the kind of person who would be a good partner — while ignoring statistical predictors of compatibility such as shared values, communication style, and life goals.
Product teams may assume a new feature will succeed because the user scenario 'looks like' a past success case, ignoring that the base rate of feature adoption is low and that surface-level similarity between scenarios does not guarantee comparable outcomes.
Interviewers frequently judge candidates based on how closely they resemble the prototype of a successful employee in appearance, demeanor, and background, rather than weighting objective performance predictors — perpetuating homogeneity and overlooking high performers who don't fit the mold.
Voters and media consumers judge political candidates or policy proposals based on how closely they resemble prototypes of success or failure ('this feels like the lead-up to another recession'), rather than analyzing the actual statistical indicators and base rates of different outcomes.
Daniel Kahneman and Amos Tversky, 1972. First formally described in 'Subjective Probability: A Judgment of Representativeness' (Cognitive Psychology, 1972), and further elaborated in 'Judgment under Uncertainty: Heuristics and Biases' (Science, 1974).
In ancestral environments, rapid categorization based on surface features was essential for survival. Identifying whether a rustling bush contained a predator or prey, or whether a stranger belonged to a friendly or hostile tribe, required fast pattern-matching against stored prototypes. Since base rates were often stable and learned implicitly through experience, the cost of occasionally misjudging was outweighed by the speed advantage of quick classification in dangerous situations.
Machine learning classifiers can replicate the representativeness heuristic when trained on biased datasets: they learn to classify inputs based on surface-level feature similarity to training prototypes rather than accounting for true base rates. This leads to stereotyped predictions — for example, resume screening algorithms that favor candidates whose profiles 'look like' past hires, reinforcing demographic homogeneity. AI systems can also overfit to representative-looking patterns in small samples, mirroring human insensitivity to sample size.
50 cards are free to preview. Buyers unlock the rest of the deck plus the interactive training — Spot-the-Bias Quiz unlimited, Swipe Deck with spaced repetition, My Blindspots, Decision Pre-Flight, the Printable Deck + Cheat Sheets, and the Field Guide e-book. $29.50$59.
