“Cosal” Inference

27 Apr

We often make causal claims based on fallible heuristics. Some of the heuristics that we commonly use to make causal claims are:

  1. Selecting on the dependent variable. How often have you seen a magazine article with a title like “Five Habits of Successful People”? The implicit message in such articles is that if you were to develop these habits, you would be successful too. The articles never discuss how many unsuccessful people have the same habits or all the other dimensions on which successful and unsuccessful people differ.
  2. Believing that correlation implies causation. A common example goes like this: children who watch more television are more violent. From this data, people deduce that watching television causes children to be violent. It is possible, but there are other potential explanations.
  3. Believing that events that happen in a sequence are causally related. B follows A so A must cause B. Often there isn’t just one A, but lots of As. And the B doesn’t instantaneously follow A.

Beyond this, people also tend to interpret vague claims such as X causes Y as X causes large changes in Y. (There is likely some motivated aspect to how this interpretation happens.)