You are both correct: correlation does not imply causation.

But, it seems to me, rather unsatisfactory to leave it there.

In life we have to continually make decisions: do we do this exercise or that, do we take a particular supplement or not, do we start to take levodopa now or not. Often there is little or no data to help us make our decision. Often we are forced by circumstances to make a decision now.

In these situations we are forced to rely on experience (often barely relevant) and whatever data we may find (often poor quality).

With this in mind, let's look at the blueberry example mentioned in the article. I agree the evidence in favour of the blueberry is weak: the positive correlation is likely to be due to confounding issues. But, since the confounding argument is not certain to be correct, the evidence in favour of them is still positive.

Put another way, on the evidence presented in the article, and ignoring cost, is there a better course of action?

John