How do black box models work?
That is the mystery, isn't it? Black box models are such complex programs that no one really knows what's going on under the hood, and the computer itself can't tell us, either. They're not built with transparency in mind; they're designed for speed and power.
Despite the lack of transparency, we still use them in engineering, finance, and a wide range of other complicated disciplines. As long as you check their output and make sure it makes sense, then how the sausage gets made is a lot less important than that it gets made.
Black box model versus white box model
If you're wondering to yourself why we let something like a black box model happen -- well, it kind of happened accidentally, and then programmers decided that since experts were working with these machines, it might not much matter how they came to their conclusions. That's the black box model: it's answers without transparency.
A white box model is very different. It's instead built with transparency in mind, as well as restraints at each step in the process to force the machine to slow down and explain itself as it goes. A white box model tells you not only the answer, but how it got there, in case there are legal or ethical implications that need to be considered. They run slower due to these additional checks, but not necessarily slow enough to matter.