Baby David Hume (Fun with Correlation, Causation and TV)
Posted by Peter Armstrong Tue, 17 Oct 2006 10:45:00 GMT
About a month ago, Gregg Easterbrook speculated about a link between TV watching by young children and autism.
Now, my son Evan is two years old.
So, I did a bit of math:
probability Easterbrook is right * potential negative consequences for my son if he is = X
enjoyment Evan receives from TV + enjoyment Caroline receives from TV + enjoyment I receive from TV = Y
Is X > Y?
Can I look in the mirror? Of course it is.
(I like looking in the mirror.)
The "potential negative consequences for my son" multiplier is SO HUGE that even if the probability Easterbrook is right is small (say 10%, hell, even .001%) it's no question.
(Note to math geeks: yes, I know this can be done as an expected value question.)
So, I cancelled cable, unplugged the TV, moved it into my office and covered it with a duvet. I blogged about it here.
That was 8 days ago.
Now, yesterday Gregg Easterbrook provided a link to a study showing empirical evidence of a correlation between TV and autism.
The only reason I might be able to sleep tonight (it's 4 AM so it's not looking good at the moment) is that I have already cancelled cable.
Of course, any damage that has been done has already been done.
And it was something I could have prevented.
Okay, maybe I don't like looking in the mirror very much right now.
"But it was just a bit of Baby Einstein. Oh, and Treehouse (a Canadian network with kids shows). "Max and Ruby" and "Mighty Machines" can't be too bad, can they? Oh, and Franklin. And Little Bear. And 'This is Daniel Cook'."
Oops.
[And at the end of a night of debugging, our intrepid blog writer heads off to Slashdot.]
Ooh, look, here we go kids: Correlation Doesn't Mean Causation.
Woo hoo, it's like Philosophy 100 without attractive classmates.
Oh, what the hell.
The following is a copy-paste-modify from my Slashdot post here:
In Canada we have warning labels on cigarette packs. Big warning labels. Cigarettes cause cancer, etc. So, naturally, some dollar store entrepreneur created fake warning labels.
Anyway, when I was a stereotypical angry young philosophy student, I thought it would be fun to make my own fake warning labels to put on my cigarette packs. I could surely come up with a better joke. So, who did I turn to? Hume, of course.
So, my cigarette packs had a big warning: "Correlation Does Not Imply Causation" on them. I thought it was a good joke, by philosophy joke standards anyway.
Now, I knew perfectly well that in this case even though it did not imply it, it was in fact true. Of course cigarettes caused cancer. In many cases correlation is, um, correlated with causation. But I was 18 so I didn't care; I thought it was funny.
What a joke.
So, the point is: correlation is a start. If there is a correlation, you should look for ways to establish whether causation exists or not.
Now, you cannot do perfect experiments on the effect of smoking on humans--you would have to start with a large random set of non-smokers, make a random half of them smoke their entire lives, and then in about 80 years check to see how many of the smokers died of cancer compared to how many of the non-smokers. It would be completely morally reprehensible (not to mention impossible: some of the non-smokers would start smoking and some of the smokers would quit). Not only would the ethics boards at the university not approve, it would take too long to run the study for it to help you to get tenure! ;-)
[Note for the sarcasm-impaired: of course I'm joking.]
So, do you just give up and say "thank you for smoking" or "well, we'll never prove anything according to David Hume then"?
No, you don't.
There are statistical tools like factor analysis which let people smarter than me figure out how much of A is (probably) caused by B, etc.
Anyway, since I am older, fatter and wiser, stuff I thought was funny at 18 is certainly not funny any more. I quit smoking over 5 years ago. I certainly wouldn't give my son a cigarette, ever.
However, if there is a strong correlation between TV and autism, I have to wonder whether I am in effect doing something similar.
What if further analysis proves (as much as you can scientifically prove anything regarding humans) that it is indeed a cause?
What would I have done???
[Yeah, yeah, I know that in the case of smoking and cancer there's a mountain of evidence, whereas in the case of TV and autism there is one study. It's obviously not a perfect analogy, but this is a blog post and I've been debugging way too long to care.]
"Correlation Does Not Imply Causation" does not mean act insanely. You have only ever seen gravity by correlation, but you still believe in it. (Yes you do. Wipe that smirk off your face.)
Do I think there is conclusive, Hume-would-be-proud proof that TV causes autism? No, not yet.
Do I think that TV is good for young children?
Would I give my 2 year old son a cigarette?
