What if the Hokey Pokey IS what it's all About???

I implore you: think in Correlations!

In high school biology classes, my teacher urged us to identify our "A-ha" moments; those times when a concept would suddenly click and forever changed the way you saw the world.

For me, one such moment was during my first dissection; putting my forcept down the frog's throat, into the espohagus and through the stomach, I realized for the first time that bodies do not contain bundles of random organs floating around; they're all connected, it makes sense, and furthermore the understanding of it's workings lies within my own grasp.

During my college education, one of my top "A-ha" moments occurred in a Psychology Statistics course, as well as philosophy and other psychology courses). The idea: "Correlation is not causation."

Let's define our terms. Causation is much like it sounds: the phenomenon where one item brings forth another. It seems simple, right? Eating causes satiety; kicking the ball causes the ball to move away from you; etc etc (assuming, of course, you don't take Kant's stance, which would mean you believe we NEVER know with certainty that anything causes another thing; for you people, my whole discussion will be useless!)."

Correlation," in contrast, simply refers to the phoenomenon where two things tend to occur at the same time, or in the same place. Lightbulbs are correlated with lamps, intaking more calories than you output is correlated with weight gain, umbrellas are correlated with rain, etc.

It sounds simple as well, right? No one would ever argue that umbrellas cause rain, or that a lack of a cast causes a foot to be broken, and yet...Causation and correlation are chronically confused, well, everywhere in life, from the editorial page of the newspaper to the sports locker room to the corporate board meeting.

How can this be, when it seems so obvious in the above examples? It's because when we move away from descriptions of physical items, it's very hard to identify which situations involve causality and which involve correlation.

The "null hypothesis," or the general assumption one should make, is that phenomena are correlated rather than caused. Yet, I often observe the exact opposite when I'm reading/hearing the news or even listening to people describe events in their own lives: for whatever reason, our minds seem to want to attribute causes rather than seeing correlations.

For example: a well-known correlation in the USA today is that of race and violent crime. Minority status is positively correlated with violent crime (meaning violent crimes are perpetrated more by minorities than their non-minority counterparts), but from that statistic should/can we conclude that minority status causes one to commit violent crimes?

Absolutely not, because the two items are linked by a third, more important factor: socioeconomic status. The phenomenon where two items are said to be caused by each other but are really linked by a third cause is called the "Neglect of Common Cause" or "Joint Effect."

The cool thing is, with statistical methodology cause and effect can often be teased apart (although I'll refrain from going into the math right now because, frankly, it's a Sunday morning and doing math on a Sunday morning is just wrong!).

However, it's important to remember that in many cases we can't tease apart causality from correlation because the only way to do so would be to take two identical things (let's use people, because that's the most interesting for me) and expose them to identical environments except for one distinguishing factor.

For example, the most perfect way to test whether sugar intake really does make kids more hyperactive would, in the ideal world, take two identical people who had been exposed to the same environmental forces their whole lives (which really isn't possible because conditions starting in the womb are never the same for both twin) and one day decide to solely change whether one would be exposed to sugar and the other not, and then measure hyperactivity (which you'd need an operational definition of).

Of course, this isn't possible, so researchers get around this by "random sampling." The idea is that you minimize the effects of potential environmental differences (such as whether a child in a sugar-hyperactivity study had eaten sugary cereal the morning of the test) by randomly choose a group of subjects (your "sample") chosen entirely by chance. By random sampling, you're equally likely to have a child who's eaten sugary cereal (which could be a "confounding factor," or something that gets in the way of establishing a link) in each group.

So, why am I babbling so much about this? Because when we're going about our daily lives, making causal attributions all day (anything from "I'm in a bad mood because I'm pre-menstrual" to "the Chinese economy is doing well because a certain politician is in power") we simply can't establish causation.

I wouldn't be writing about it if I thought it was simply an academic problem. It's an enormous social problem because in explaining things in simple terms of causation, we can be lulled into a sense of complacency and simplicity about our fellow humans, particularly those of groups different to us. I won't go into it now, but psych studies have found that people think more deeply about alternate causes for negative behavior about people from an "in-group" than an "out-group."

So, if I'm reacting to a recent finding that women in their 20s are (and i'm just making this up) less likely to cooperate with peers than men of the same age, I will probably consider alternate explanations almost reflexively. However, if the opposite claim is made, I might simply shrug and say, "Men...what can you do?" (As a fun side note, the opposite also happens, in that we often blindly accept positive claims about our "in-group" but think of alternate explanations for positive claims about an "out-group.")

This particular example doesn't necessarily have huge reprocussions, but what about when we take examples of ethnic groups and religions? Assuming the above principle is true, if a friend or trusted source explains someone's negative trait by their membership in, say, the Islamic faith, we are generally less likely to persue alternate explanations (such as that the trait is just part of their individual personality, or comes from their job, or their association with another friend).

The basic idea I'm getting at is that when we're explaining phenomena in the world around us, whether it's as personal as a sibling's sudden warmth ("do they want something from me? do they have an illness which is making them suddenly see the world in perspective? is it because they are happy in their relationship?") or as broad as a nation's obesidy epidemic ("is it from lack of exercise? fast food? changing ideal body types?"), it's not only more useful to think in correlations, it's simply more true.

To end, I'd like to provide an example of when this is done particularly well; not surprisingly, it's from "The Economist." I was very skeptical when I started the article, but it provides a good example of when considering the correlation v causation problem is done well: http://www.economist.com/finance/displaystory.cfm?story_id=5327652&tranMode=none | posted by Cheryl, 1/22/2006 10:57:00 AM