>>13394281Lets ignore the question of whether causality and causation actually exist.
We pragmatically determine a thing to be causal through experiment. An example of a high quality experiment is running statistically high-powered double blind randomized control trials.
That is. We take a large number of things T. We randomly assign those things T to group A and B, in a way that the experimenters dont know which group a thing T is in. And if thing T is a human, then T also doesn't know which group it is in.
Then group A gets X, group B gets fake-X. X might be a medicine, a therapy, a surgery, or some other baseline.
We then look for outcome Y. Some quantitative variable like blood glucose, or some qualitative variable like death.
After the experiment is conducted, the data is labelled and analyzed by statisticians who don't know what the labels are referring to, so they don't bias the analysis.
Then the analysis is given to those who know what the labels mean.
If X (relative to fake-X) causes some statistically significant change in Y, then we say X caused the change in Y. We say this because the change is not accounted for by anything else. There is no experimental error.
