You're overextending the logic here. Yes, that's a correct explanation of the difference between correlation and causation. But just because the B->A case matches logic doesn't mean that it works as a hypothesis. (In this case, I guess it would mean that children with speech delays are able to induce their parents to let them watch more TV -- that's absurd to the point of being nonsensical).
In fact, where there is a clear and sane hypothesis in play (e.g. "time spent watching TV is time not spent learning to talk") it almost always works out that further science shows the causation that you expect. That's true across fields, and it really shouldn't surprise anyone.
It's just that the experiments required to show that are harder. Simple demographic studies are a lot easier and cheaper. So you do those first, then work up the hard stuff when you know where to look.
Taking your point literally, it would never be useful to do demographics like this because you can't "prove" the causation. But of course that's ridiculous; these studies are immensely useful and improve all our lives.
"In fact, where there is a clear and sane hypothesis in play (e.g. "time spent watching TV is time not spent learning to talk") it almost always works out that further science shows the causation that you expect."
There can be more than one clear and sane hypothesis, even if it doesn't occur to you immediately. For example, Wilduck below suggests poor parenting as an underlying cause for both these effects.
Here's another example. It is well-known that there is a correlation between the global temperature and the levels of CO2 in the atmosphere (from looking at historical data). It is widely believed that higher CO2 levels lead to higher temperatures. On this basis governments have capped CO2 emissions, with real economic implications. This seems reasonable, doesn't it?
But some scientists who are working on this suggest an alternative explanation; here's the gist of it. There is a lot of CO2 trapped in ice, due to some historical reason that escapes me at the moment (it's not my field). As temperatures rise, ice melts and CO2 is released, leading to higher CO2 levels in the atmosphere. If this is the correct explanation, following your advice and putting caps on CO2 emissions certainly did not "improve all our lives".
"it almost always works out that further science shows the causation that you expect."
Evidence?
"Taking your point literally, it would never be useful to do demographics like this because you can't "prove" the causation. But of course that's ridiculous; these studies are immensely useful and improve all our lives."
Not at all. Sometimes correlation is enough. For example, if an insurance company finds a correlation between having red hair and being involved in more car accidents, they can use this information to their advantage.
> children with speech delays are able to induce their parents to let them watch more TV -- that's absurd to the point of being nonsensical).
Is it? A child that does not speak much takes more effort to engage with in play that does not get boring and tedious to the parent, and may get frustrated with the parent more easily from finding it hard to get their meaning across. In this case more TV would be a "solution" for both parties, and might be arrived at simply because it causes fewer tantrums and less stress for the parents.
I don't believe that's the most likely cause, but I also do believe one should be exceedingly careful about jumping to conclusions about cause based on "common sense" and correlation, because a whole lot of explanations that people would never even think of suddenly seem like common sense after the fact.
> In fact, where there is a clear and sane hypothesis in play (e.g. "time spent watching TV is time not spent learning to talk") it almost always works out that further science shows the causation that you expect.
In fact, where there is a clear and sane hypothesis in play (e.g. "time spent watching TV is time not spent learning to talk") it almost always works out that further science shows the causation that you expect. That's true across fields, and it really shouldn't surprise anyone.
It's just that the experiments required to show that are harder. Simple demographic studies are a lot easier and cheaper. So you do those first, then work up the hard stuff when you know where to look.
Taking your point literally, it would never be useful to do demographics like this because you can't "prove" the causation. But of course that's ridiculous; these studies are immensely useful and improve all our lives.