Hasty generalization

Two rude waiters on one visit to a city of a million people tell you about two waiters, not about the city.

From a few cases to all of them

A generalization is a claim about a whole group: all, most, or typically. We make them all the time, and we need to, because nobody can check every case. A hasty generalization draws such a conclusion from too few cases, or from cases that are not typical of the group. “I bought two phones from this brand and both broke within a year, so the brand makes bad phones” may turn out to be right, but two phones out of millions sold cannot show it.

Two things go wrong, often together. The first is size: a small sample can easily come out lopsided by chance. Toss a fair coin four times and getting three or four heads is not unusual at all. The second is selection: even a large sample misleads if it was gathered in a way that favours one answer. Reviews written by people angry enough to post, or a survey of a gym’s members about how often people exercise, tell you about those groups, not about everyone.

How to spot one

• Count the cases. One story, however vivid, is one case.
• Ask how the cases were chosen. Who was asked, and who was left out?
• Look at the size of the claim. “Some” needs one example; “all” or “most” needs much more.
• Look for a comparison. “Five people got sick after eating there” means little without knowing how many ate there.

Advertising uses the same move in reverse. A made-up ad says “9 out of 10 customers recommend us!” in large print and “based on 10 customers surveyed at our opening event” in small print. Ten people chosen at a party the company hosted is both a small sample and a selected one. Stereotypes about groups of people are hasty generalizations too: a handful of encounters, often remembered because they were unusual, stretched into a claim about millions.

The cure is not to stop generalising but to match the strength of the claim to the evidence. “Both phones I bought broke, so I will read reviews before buying another” is reasonable. “In a survey of 2,000 randomly chosen adults, 63% said ...” gives you both a size and a method you can judge.

Vivid stories carry more weight in memory than numbers do. A friend’s account of one bad night at a hotel can outweigh a summary of thousands of stays, because the story has a face, a place and a feeling. Logic does not say the story is false. It says the story is one case, and a single case can show that something is possible but cannot show how often it happens.