Hacker News Reader

Comment by edw519 | Hacker News Reader

This story is a perfect example of flawed logic that we see so often here on hn that whenever someone points it out, they immediately get downvoted.

So, one more time, here goes:

correlation != causation

It goes something like this...

"We have observed a and b together very often. Therefore, a must have caused b."

Kinda reminds me of the garbage strike in New York City years ago. Garbage piled up on the streets and rats appeared everywhere. One of these "experts" probably would have said, "Look at all the garbage these rats brought."

Maybe we need a new "correlation != causation" flag.

View on HN

Where does the article claim causation?

"It would be easy to test the theory, said Steven Pinker, a Harvard cognition researcher: "See if carriers of the Ashkenazi-typical genetic mutations score higher on IQ tests than their noncarrier siblings."

Cochran and Harpending readily acknowledge the need for such experiments. But they have no plans to do them. They say their role as theorists is to generate hypotheses that others can test."

   > Kinda reminds me of the garbage strike in New
   > York City years ago. Garbage piled up on the streets 
   > and rats appeared everywhere. One of these "experts" 
   > probably would have said, "Look at all the garbage 
   > these rats brought."

:-) I also like: "Most traffic accidents are caused by people who have a driver's licence. Therefore, the driver's licence is responsible for traffic accidents."

Uh, actually the driver's license causation seems to make sense (in an environment where drivers have to have a driver's license). If yu took away all driver's licenses, accidents might go down. If you took away the rats, the garbage would stay.

well, if you take away all driver's licences, I guess that includes the driver's licences of police officers. Which would make it relatively difficult to enforce the rule that "drivers have to have a driver's licence"... [Porsche vs police bicycle...] :-)

roadblocks?

It's called a theory. Then you try and prove it using the scientific method. Then we move on to other problems.

So is causation impossible to prove?

It seems like every single study ( at least those reported in the media) suffers from the correlation != causation problem.

We can never be sure of anything (because everything could just be a simulation in a computer etc.). We can only be certain to a degree. I suppose just finding some causation is not sufficient to make a new scientific discovery. It can give pointers for directions of research, but then somehow you need to make the new hypothesis testable.

http://xkcd.com/552/

Reply on news.ycombinator.com