Here is a fictional story of a science research.

An institution wants to investigate how rats see things. Researchers put a cat near a rat, then the rat escaped. They repeated this procedure for a considerable number of times, using different kinds of rats and cats, and found that the rat will always run when there is a cat near it. Thus, they drew a causality that a rat, when running away from a cat, can see cats.
Then they broke the legs of rats, and put cats near the rats. All repetitions of procedures show that none of the broken-leg rats can run when there is a cat. Thus they came to conclusion that the rat can’t see cats when their legs are broken. In other words, the rat see things via their legs.

This conclusion, of course, sounds absolutely ridiculous. However, you think this conclusion is ridiculous in that your knowledge and experience tell you that eyes are exactly what allow animals to see things. Thus you will further argue that, the broken-leg rats can’t run because the legs are used for running instead of looking, so they can still see but they can’t run. This argument is correct. Nevertheless, the research topic is trying to reveal the true nature of a black box, an unknown mechanism that you can only get outputs from inputs but that you don’t know what happened inside the box. Therefore, if the same research procedure is put onto a research topic about which you have absolutely no prior knowledge, do you still think such research is ridiculous?

If you know philosophy, you might have already noticed that this is actually the Humean Problem I am talking about. The researchers drew an incorrect causality between the rats’ legs and rats’ ability to see from observing whether the rats will escape from cats but ignored that running – an ability actually provided by legs – also played a vital role that makes rats escape from cats. However, if the research topic is to reveal the nature of a black box, how should the researchers notice there are such vital parts they have ignored?

Let’s review how we disproved the research in the story. We argued that rats with broken legs didn’t run away from cats because legs are used for running instead of seeing, and that running is also a vital part of escaping from cats. We may notice that we are taking other parts of a rat into considerations. So, if researchers have thought in this way, they will be able to make other hypotheses then will do experiments to verify it. For example, in addition to breaking the rats’ legs, researchers will try to break something else like ears, tails, eyes instead. My phrasing might sound inhumane, but you can try to find alternatives that will achieve the same effects (for example, using blindfolds and earplugs, but keep in mind that Humean Problem could argue these humane measures could be ineffective because we don’t know if blindfolds can hinder the functionality of eyes and earplugs can block the functionality of ears).

As we can see, in the Humean Problem, we are facing critics who always deny the causality we are making by acknowledging the correlations we found but denying these are the causalities. They’ll claim there are something unknown which are actually the causality and the correlations are merely coincidents.

Is the Humean Problem the doomsday of science? The answer is yes if we treat science as truth itself because Humean Problem can always deny the causality. In order not to see the Humean Problem as the doomsday of science, we should instead treat science as something that tries to get as much close to the truth as possible so that, when Humean Problem advocates are denying us with concrete assumptions, we can turn these assumptions into hypotheses and verify the hypotheses in experiments. We will be able to claim such correlation is the closest to causality but others aren’t.

As you can see, to learn the true nature of a black box, we should make good use of the idea of Humean Problem to aid avoiding treating false correlations as causalities.