Reversing the Arrow of Causation

Mijara T’ran:

You’ve likely heard the skeptic’s cautionary advice that correlation is not causation. We should be wary whenever someone observes that two things occur together, therefore one causes the other. Unfortunately, this kind of shaky logic is extraordinarily common these days. Rigorous experimentation, where you hold as many variables constant as possible and then change the factor you’re studying to observe its effects, is difficult, expensive, and time-consuming. This holds especially true in the area of health, where the subjects are people with their myriad habits and lifestyles (lots of variables to control for!), and the effects under study might not change in measurable ways for years or decades.

To get around these difficulties, researchers turn to an easier, cheaper, faster alternative: data mining! We have accumulated a wealth of data on innumerable facts of people’s lives, especially their health. By comparison to clinical trial and experiment, it’s easy to sift that data and narrow down to the variables we’re interested in and see how they are interrelated. The drawback, as you might have guessed, is that the output is all correlational data. We can see which lines on the graph climb at the same pace, but without isolating them and seeing what happens to one when we change the slope of the other, it’s very difficult to be sure which one of them is causing the other. Indeed, they might not be related by cause and effect at all, but could both be effects of some third variable.

Here’s a trick I’ve adopted to help spot bad correlation-causation claims, a critical-thinking shortcut that’s quick to apply and often insightful. Ask yourself: “Is it plausible that what they’re claiming is the cause might actually be the effect, and vice versa?” People doing correlational studies frequently don’t consider the possibility, simple though it is! Here are a few examples to get you thinking.

The correlation: People who are overweight eat a lot and don’t exercise much.
The claim: Overeating and lack of exercise cause weight gain.
The flip: Gaining weight causes you to eat more and exercise less.

If you’ve read any of this blog up to now, you’re familiar with this one. It looks weird at first, since we’re so inundated with the message that fat people are fat because of their bad behavior. But it turns out this flip brings immense insight when investigated. In fact, there’s a different cause to weight gain: the hormone insulin, whose action ramps up when we eat carbohydrates, the type of food energy found in staples like bread and pasta. The metabolic effects of weight gain in turn slow down our desire to exercise and prompt us to eat more. Stop eating carbs, insulin quiets down, you lose weight, and your appetite decreases and your energy for exercise returns!

The correlation: People who sit a lot during their days have higher incidence of obesity, diabetes, and heart disease.
The claim: Sitting makes you more prone to get sick.
The flip: Getting sick makes you sit more.

A while ago, a scary infographic made the rounds of the Internet, saying how sitting is killing you. Its message prompted me to get a standing desk at work and to set one up at home. But after learning about the causation-flip trick, I came back to the topic and tried it out: great Primes, it’s so obvious! People who are obese, diabetic, or have weak hearts are going to have more trouble getting up and staying active than people not suffering from those conditions. So of course you’re going to see sitting time climb along with the incidence of those conditions. Now, there are valid points in that article, notably the pieces that come from experimental observation rather than correlational leap. But the graphic’s biggest punch comes from something that, having engaged this trick, seems sketchy.

Try it out on your favorite topics! What if we’re not getting stupider because of our cell phones, but we’re more enthralled by our cell phones because some third factor is weakening our brains? What if your shiftless cousin didn’t stop looking for work because he was lazy, but got lazy (depressed, lethargic) because of his inability to find work? It won’t always yield some new insight, but the occasions when it does may surprise you!


4 thoughts on “Reversing the Arrow of Causation

  1. Abram says:

    Flipping correlations does not automatically lead you to the truth either, though. A doesn’t necessarily cause B because they’re correlated, but B doesn’t necessarily cause A either, just because it supports the rest of what you believe.

    • SabreCat says:

      If I somehow claimed it did, I would be extremely remiss. (I’m in fact a little stung that you somehow came to that conclusion, it’s so blatantly wrong.) It’s a technique, not a magic bullet.

      • Abram says:

        I wasn’t saying that you said that it was a magic bullet, I was saying it has a very easy trap to fall into, although perhaps not spelling it out as much as I should have.

        This technique, if used on it’s own, is good against new propaganda/researcher biases, bad against already accepted beliefs, because you can choose which causation fits your personal narrative without challenging yourself.

        It’s a good stepping point for further research, though.

      • SabreCat says:

        Ah, sure. The crucial thing to remember is that flipping the posited cause and effect gets you an alternative hypothesis to what you’ve been given, but it remains a hypothesis about correlational data. It doesn’t take the place of more rigorous work to demonstrate a causal link.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s