本周（及每周）人们常犯的错误：混淆相关性与因果关系
内容总结：
【生活技巧】你知道吗？谷歌搜索可以这样设置，过滤掉无用信息！只需几个步骤，就能优化搜索结果，比如将Lifehacker等可靠来源设为优先站点。不过，今天我们要聊一个更普遍的问题：人们总是混淆相关性与因果关系。
尽管大卫·休谟早在1739年就提出“相关不等于因果”，但这个认知误区至今仍在影响我们。例如，近20年来关于“肠道菌群”的研究可能存在问题，部分原因正是科学家和媒体混淆了相关与因果。同样，许多人曾相信适度饮酒有益健康，但实际上，饮酒只是与健康指标相关，并非健康的原因。
这种误区还体现在疫苗与自闭症的争论中。数据显示自闭症诊断率与疫苗接种率存在相关性，但这主要是因为自闭症确诊年龄与疫苗接种时间重合，而非疫苗导致自闭症。
更典型的例子是冰淇淋销量与鲨鱼袭击事件的相关性：两者同时上升的真正原因是天气变热，而非冰淇淋直接引发鲨鱼袭击。事实上，许多相关性纯属偶然，比如情景喜剧《好汉两个半》的收视率与塞尔维亚航空燃油消耗量“高度相关”，或者“我的猫抓伤我”的谷歌搜索量与美国水果消费量的“精准同步”。
如今，人工智能甚至能生成虚假“研究论文”来解释这些随机相关性，例如称“健康意识高的家庭更常吃水果，也更可能搜索猫抓伤的处理方法”。尽管这些解释看似合理，实则毫无根据。
混淆相关与因果可能导致盲目跟风减肥、误信饮酒健康论，甚至影响公共政策制定。面对“X导致Y”类结论，我们应保持警惕：除非有确凿证据，否则先假设这可能是又一场“猫抓伤与水果消费”的误导。
中文翻译：
你是否知道可以对谷歌进行自定义设置，从而过滤垃圾信息？只需几个步骤就能获得更优质的搜索结果，比如将我在Lifehacker发布的内容设为优先信源。
通常我会聚焦于某些群体普遍存在的认知误区，让其他人显得更聪明些。但本周我要探讨一个更宏大、更广泛的话题——一个你、我以及我们认识的每个人都曾误解、正在误解且将继续误解的概念：将相关性误作因果性。
自1739年大卫·休谟在《人性论》中阐述这一概念以来，人们就以各种形式重复着"相关不等于因果"的箴言。简言之：两件事同时发生并不意味存在因果关系。每个聪明人都深知这个道理，尽管这句话被反复强调，我们却依然会犯这个错误。
来看几个实例：
过去二十年关于"肠道菌群"的研究可能存在谬误，部分原因正是科学家和媒体混淆了相关性与因果性（我一直对此类研究存有直觉上的怀疑——明白这个双关吗？）
多年来人们坚信适度饮酒有益健康，但事实并非如此。在某些群体中，饮酒确实与健康指标存在相关性，但并非健康改善的成因。
那些基于相关/因果谬误的错误信念捍卫者，往往会堆砌大量数据来证明某种高度吻合的相关性——比如自闭症发病率与疫苗接种率的同步增长。这种数据呈现会让人自然认为存在因果关系，但实际上并无证据表明疫苗会导致自闭症。这种相关性很可能源于：儿童接种疫苗的年龄段正值自闭症通常被确诊的时期，且我们在疫苗接种和自闭症诊断两方面都取得了进步。但请注意这个"很可能"的表述——尽管疫苗与自闭症从未被证实存在因果联系，两者数据吻合可能存在无数种解释。
诸如冰淇淋销量与鲨鱼袭击关联这类经典辟谣案例，往往会给因果联系提供简单解释：看似冰淇淋引发鲨鱼袭击，实则因为天气炎热时冰淇淋销量和海边游泳人数同时增加。但即便这种解释也可能陷入相关性误判——听起来合理，但我们实际上并不清楚这两个数据为何吻合。有时，毫不相干的信息之间根本不存在任何关联。
请看下方图表：这充分证明《好汉两个半》的收视率与塞尔维亚航空燃油消耗量存在直接关联。
再来看看"我的猫抓伤了我"的谷歌搜索量与美国水果消费量的精确对应关系。
第二个例子是我通过泰勒·维根的"伪相关性"网站工具生成的，这个工具能让你整天制造随机关联。不仅如此，该网站还利用人工智能生成虚假"研究报告"来解释这些关联。
关于猫抓伤的案例，ChatGPT提供了这样一种解释："注重健康的家庭（关注饮食、购买水果等）更可能对轻微伤势保持谨慎。爱吃水果的人并非更易被猫抓伤，但他们更倾向于搜索'我的猫抓伤了我'以查询感染风险或处理措施。"
即便明知是胡扯，这种解释依然显得自洽。这正是我们永远无法彻底摆脱此类认知误区的原因——我们的大脑渴望相信事物。一个措辞精巧的解释、一张整洁的图表、一套听起来可信的理论，带来的满足感远胜于简单的"我们不得而知"。
将相关性误作因果性会让我们盲目跟风流行饮食法，误以为晚餐喝葡萄酒就能保持健康。这种认知偏差以可能真正伤害人们的方式影响着健康建议、公共政策和个人决策。我们最多只能保持警觉：当看到"X导致Y"的标题时，先假定这是"猫抓伤导致水果消费"式伪关联，直到有证据证明并非如此。
英文来源：
Did you know you can customize Google to filter out garbage? Take these steps for better search results, including adding my work at Lifehacker as a preferred source.
I usually focus on something a subset of the population gets wrong, so the rest of us can feel smart, but this week, I’m going bigger and broader, and describing something that you, me, and everyone we have ever met has been wrong about in the past, is currently wrong about, and will be wrong about in the future: ,istaking correlation for causation.
People have been repeating some variation of "correlation is not causation" since at least 1739, when David Hume articulated the concept in A Treatise on Human Nature. To paraphrase Hume: Just because two things happen at the same time doesn’t mean one is causing the other. Every smart person already already knows this, and it's repeated constantly, but we all still get it wrong.
Here are some examples:
The last 20 years of research on "gut biomes" could be wrong, partly as a result of both scientists and the media mistaking correlation and causation. (I've always had a gut feeling—get it?—this research was suspect.)
For years people believed that drinking alcohol in moderation is good for your health. But it isn't. It's correlated with better health in some populations, but it doesn't cause better health.
Defenders of mistaken beliefs derived from the correlation/causation fallacy will often compile volumes of data that shows a nearly exact correlations between, say, higher rates of autism and higher rates of vaccination, which makes it totally understandable to believe one causes the other. But there's no evidence that vaccines cause autism, and the correlation is probably because children received vaccines around the age autism is generally diagnosed, and we've gotten better at both vaccinating children and diagnosing autism. But "probably" is doing some lifting in that sentence. While no causal link between vaccines and autism has ever been demonstrated, there could be any number of reasons the rates line up.
Classic debunking examples of correlation and causation, like the link between ice cream sales and shark attacks, tend to offer a pat explanation for the causal link—it only seems like ice cream causes shark attacks because both ice cream sales and swimming in the ocean rise when the weather is warmer—but even that is potentially mistaking correlation for causation. It makes sense, but we don’t actually know why those two numbers line up. And sometimes there just isn't any reason for connection between two pieces of disparate information.
Check out the chart below. It's proof that the ratings of Two and Half Men directly correlates with the amount of jet fuel used in Serbia.
Or check out the exact connection between people googling "my cat just scratched me" and U.S. fruit consumption.
I made the second example on Tyler Vigen's Spurious Correlations, with a tool that will lets you make random connections all day. Not only that, the site uses AI to generate bogus "research papers" to explain the connection.
In the case of the cats, ChatGPT offers this as a possible explanation:
"Health-conscious households (those that track diet, buy fruit, etc.) are more likely to treat even minor injuries cautiously. A person who eats more fruit is not more likely to be scratched, but they are more likely to Google 'my cat scratched me' to check for infection risks or treatment steps."
Even though I know it's bullshit, it still tracks. That's why we can never really stop being wrong in this specific way. Our brains want to believe. A neatly phrased explanation, a tidy chart, a plausible-sounding theory—it's so satisfying. A simple "we don't know" can't hold a candle to that certainty.
Mistaking correlation for causation makes us go on fad diets and believe we're being healthy by drinking wine at dinner. It shapes health advice, public policy, and personal decisions in ways that can actually hurt people. The best we can do is try to be aware of it—when we read a headline that says "X causes Y," to assume this is a "cat scratches cause fruit consumption" situation until there's evidence that it isn't.