While you are correct conceptually, most scientific research using statistics does in fact attempt to provide evidence against no-effect (the null hypothesis).

The p-value of a scientific study is the probability that the given data would have been observed, given that there is no effect. Hence why small p-values can be associated with the success of the alternative hypothesis (i.e. what a scientist actually thinks will happen).

We should remember Popper (1).

You can't prove an hypothesis is correct. All models are temporal until they are disprove.

(1) - https://en.wikipedia.org/wiki/Falsifiability

I know some people who tap a can of Coke before they open it, because "that prevents it from spilling". Each time they do that, they are more convinced that their hypothesis is correct.

But is it correct? To find out, we need to try to invalidate it. Try opening the can without tapping it first. If the Coke doesn't spill, the hypothesis is clearly wrong. It the Coke does spill now, it's quite strong evidence in favor of the hypothesis.