First Ask About the Sample: Who Was Surveyed, and How Many
"A survey shows 90% of people support this plan"—hold on, ask one thing first: who was surveyed, and how many?
If only 10 people were asked, or they were all asked inside one fan group, the sample is both too small and unrepresentative, and that 90% tells you almost nothing about everyone.
Correlation Isn't Causation
Two things often showing up together does not mean one caused the other. This is the point most often abused.
Example: "In summer ice cream sells more, and at the same time more people drown, so eating ice cream causes drowning"—absurd. The real cause is "hot weather": heat makes people buy ice cream and makes people go swimming. The two are merely correlated; there's a common cause behind them.
Who's Saying It & Whether It Can Be Reproduced
There are two more things to watch.
Conflict of interest: if a study "proving sugary drinks are harmless" was paid for by a beverage company, you should be extra careful—not that it must be false, but that it needs other independent evidence to back it up.
Reproducibility: a reliable finding should be one that other teams, following the same method, can get similar results from. A stunning conclusion that no one has been able to independently reproduce over time deserves to have its credibility discounted (a single result isn't necessarily wrong—it just carries limited weight until more confirmation comes in).
自测 · 学完检查一下
想真正动手做题、记进度、攒连胜？到互动课里练。
"At the company's annual party we asked 8 colleagues, and 7 of them said they love overtime, so clearly everyone loves overtime." What's the biggest problem with this conclusion?
答案：The sample is too small and unrepresentative; it can't support "everyone"
Asking only 8 people, all colleagues at the same event, is a small and skewed sample that can't represent everyone. The problem isn't the arithmetic or the wording, and certainly not whether overtime is good.
"Data found that people who use a certain brand of toothpaste earn more, so this toothpaste makes people richer." Where does this inference go wrong?
答案：It treats correlation as causation; there may be a common cause
Correlation doesn't mean the toothpaste causes higher income; more likely, higher earners are more willing to buy pricey toothpaste—there's another cause behind it. This is a classic case of mistaking correlation for causation.
True or false: if a study "proving our product is harmless" was funded by the product's manufacturer, we should be more cautious about the conclusion and look for independent evidence.
答案：True
A conflict of interest brings a risk of bias. It doesn't mean the conclusion must be false, but it does call for more caution and supporting evidence from independent sources.
True or false: as long as it's labeled "experts say" or "studies show," the conclusion must be reliable and there's no need to question further.
答案：False
Experts can be wrong and studies vary in quality. You still need to ask about the sample, correlation vs. causation, who funded it, and whether it can be reproduced. An authority label isn't a get-out-of-jail-free card.
Two things often appearing together doesn't mean one caused the other; this "appearing together" is called ____, and it isn't the same as causation.
答案：correlation
Two things changing together is called correlation; correlation isn't causation—there may be a common cause behind it, or it may be pure coincidence.
When you hear a "stunning research conclusion," which point best helps you judge whether it's reliable?
答案：Whether other independent teams, using the same method, reproduce similar results
Reproducibility is an important mark of a reliable scientific conclusion: it's more credible when others can reproduce similar results. The headline, website, and share count don't determine whether a conclusion is true.