"Every time we send 5 fire trucks to a fire, the damage is 10x than when we only send one fire truck. We've observed this 1000 times. And still you don't believe that the fire trucks are causing the damage."
In this case it should be absolutely clear that A (lots of trucks) aren't causing B (lots of damage), but rather a third aspect, C (size of fire) is causing both A and B. Insisting that A causes B will result in completely counterproductive interventions, like "send only one truck to all fires".
The same thing could be true for cardiovascular fitness. If people are sick, they're much less likely to running or hiking up a mountain. So rather than poor cardio fitness (A) causing high mortality (B), it could be that a third thing, sickness (C) is causing both A and B. If that is the case, then shaming people who are sick into trying to exercise, instead of making them healthy enough so that they feel like running, is likely to make things worse rather than better.
How do you tell the difference? Well the "gold standard" is randomized controlled trials. Pick 3,000 random people. Tell 1000 of them to exercise more, and 1000 of them to exercise less, and 1000 leave alone, and compare. If the "exercise more" group is healthier at the end of 10 years, that's decent evidence that "exercise more" is a useful intervention.
Failing that, you can think of other possible confounding factors and control for them. Don't just ask how much they exercise; ask how old they are, and how well they are, how stressed they are, and loads of other factors which might both cause both A and B, and use statistical methods to detect whether one of those factors is actually a better explanation than "A -> B".
"Every time we send 5 fire trucks to a fire, the damage is 10x than when we only send one fire truck. We've observed this 1000 times. And still you don't believe that the fire trucks are causing the damage."
In this case it should be absolutely clear that A (lots of trucks) aren't causing B (lots of damage), but rather a third aspect, C (size of fire) is causing both A and B. Insisting that A causes B will result in completely counterproductive interventions, like "send only one truck to all fires".
The same thing could be true for cardiovascular fitness. If people are sick, they're much less likely to running or hiking up a mountain. So rather than poor cardio fitness (A) causing high mortality (B), it could be that a third thing, sickness (C) is causing both A and B. If that is the case, then shaming people who are sick into trying to exercise, instead of making them healthy enough so that they feel like running, is likely to make things worse rather than better.
How do you tell the difference? Well the "gold standard" is randomized controlled trials. Pick 3,000 random people. Tell 1000 of them to exercise more, and 1000 of them to exercise less, and 1000 leave alone, and compare. If the "exercise more" group is healthier at the end of 10 years, that's decent evidence that "exercise more" is a useful intervention.
Failing that, you can think of other possible confounding factors and control for them. Don't just ask how much they exercise; ask how old they are, and how well they are, how stressed they are, and loads of other factors which might both cause both A and B, and use statistical methods to detect whether one of those factors is actually a better explanation than "A -> B".