Hasty Generalization: Too Few Cases, Too Large a Claim

A traveller eats one bad meal in a new city and declares the whole city can't cook. The claim fails because one dinner cannot stand in for thousands of restaurants, cooks, and cuisines. This is the shape of every hasty generalization: a conclusion about a large group resting on a sample too small, too narrow, or too similar to itself to represent that group.

Two things decide whether a generalization is earned or hasty: sample size and sample bias. Size asks whether you have enough cases. Bias asks whether those cases were chosen in a way that lets them stand for the whole, rather than for just one corner of it. A sample can be large and still biased, if everyone in it was pulled from the same mailing list, neighbourhood, or walk of life. Growing the sample without changing how it was gathered doesn't fix this; a thousand students from one athletics list still only describes athletes with certain schedules, not the district.

This matters right next to its opposite error, the fallacy of composition, and next to causal fallacies, because weak samples often get mistaken for causes. It also explains why personal experience isn't automatically suspect: a doctor's patients, observed consistently over years, can be a real dataset, not an anecdote. The test is never just how it feels to generalize, but whether the sample's size and breadth earn the size of the claim being made.