Starting With a Strange Statistic
Someone noticed that the more sunglasses a city sells, the more people get sunburned. So do sunglasses cause sunburn? That just doesn't sound right.
The truth is: on sunny days, people both buy more sunglasses and get sunburned more easily. Sunglasses and sunburn merely "go up together"—neither one caused the other.
The Hidden "Common Cause"
Many cases of "fake causation" actually have a common cause behind them, driving both things at once. Find it, and the mystery is solved.
For example: "Kids with bigger shoe sizes know more words." Do big feet make you smarter? Really it's age—as a child grows up, their feet get bigger and they learn more words.
Coincidence, and Causation "in Reverse"
Besides a common cause, there are two other common situations. One is pure coincidence—the data happened to move together this time, but pick a different stretch of time and it falls apart.
The other is reverse causation—you think A causes B, but actually B causes A. Take "hospitals are full of sick people, so hospitals make people sick." In fact people go to the hospital because they're already sick—the direction is backwards.
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。
A city's data shows that fires which draw more fire trucks also end up causing greater property loss. Someone concludes: "Stop sending so many trucks—the more trucks, the worse it burns." What's wrong with this reasoning?
答案：The bigger the fire itself, the more trucks it draws and the more damage it causes—the blaze pushed up both at once
Right: the real driver is the common cause "scale of the fire," which raises both the number of trucks and the losses; the two merely grow together. Among the wrong options, "water feeds the flames" is a made-up fake mechanism, "no connection" overshoots (they are correlated, just not causal), and "big dataset means causation" is exactly the trap this lesson aims to break.
A survey finds that primary-school kids with more books at home also have higher rates of myopia. To judge whether "reading actually causes nearsightedness," which step should you take first?
答案：Check whether factors like household income or parents' education bring both more books and more close-up eye strain
Right: more books and more myopia are likely both pushed up by the common cause "family socioeconomic conditions," so checking that first keeps you from blaming reading unfairly. Note that "draw a causal conclusion straight away" and "accept it as settled fact" both mean believing without checking; "hide the books" looks like an experiment, but with no other variables controlled, that single step can't establish causation.
"Neighborhoods with more cafes also have higher housing prices." Rather than cafes raising prices, it's better to say there's a background factor pushing up both at once (e.g., the neighborhood was already bustling and busy). In this lesson, that factor affecting both is called their "______ cause."
答案：common
Right: the "how bustling the neighborhood is" that makes both cafes and prices rise is the common cause—the real driver behind the scenes, not the cafes themselves. The first move to see through fake causation is to look for this C.
"People who go to the gym a lot actually get more injuries than those who don't—so does the gym make people get hurt?" The most likely flaw here is:
答案：It's probably the reverse: people who love exercise and train a lot are the ones who go to the gym often, and more exercise naturally means more injuries
Right: at its core this is a **common cause**—the single thing "loving exercise / training a lot" makes people both go to the gym often and get hurt more easily; the gym and the injuries are merely pushed up together by that one cause, not the gym manufacturing injuries. Note it isn't "reverse causation": the injuries don't loop back to make people go to the gym—the real driver behind the scenes is the C, "loving exercise." Among the wrong options, "no connection" denies the correlation, "equipment must be dangerous" is an unfounded claim, and "too large a sample distorts it" treats a big sample as a bad thing—it's the opposite.
Someone online posts a chart: over the past decade, a country's "per-capita cheese consumption" and the "number of people who died tangled in their bedsheets" track almost identically. Which judgment is most sound?
答案：It's most likely coincidence—two unrelated things happen to wiggle together, and over a different range of years they'd probably stop matching
Right: cheese and bedsheet accidents have nothing to do with each other; however alike the curves look, they're just moving together by chance—that's "coincidence," and it usually collapses over a different time period. Be wary: you shouldn't force a common cause or causation just because curves line up—"forcing a common cause" and "matching curves mean causation" are both over-reading, while "cheese makes people thrash" is an ad-hoc fake mechanism.
Of the four bits of reasoning below, which one actually holds up (closest to a sound causal judgment)?
答案：With people randomly split into two groups and all else equal, the group that drank a certain beverage was more prone to insomnia—that looks more like the beverage causing insomnia
Right: randomly splitting people into groups balances the various confounding factors between the two groups **on average in a large sample** (and also weakens reverse causation), so the remaining difference is more likely real causation—this is what moving from correlation to causation should look like. Note that randomization doesn't eliminate every confounder with 100% certainty; it makes the two groups broadly comparable across a large sample. The other three are, respectively, reverse causation (umbrellas), a common cause (the gold chain reflects underlying wealth/status), and a common cause plus a tacked-on causal claim (both ice cream and drowning are driven by "hot weather / summer break" and don't cause each other).