I think you can only "prove" things about the system you construct. While this is true in a literal sense, often the question is not whether a model is definit

I think you can only "prove" things about the system you construct. While this is true in a literal sense, often the question is not whether a model is definitively correct, it is whether a simpler model is sufficient to explain real-world phenomena where more complex models have been previously hypothesized. In those cases, we should generally prefer the simpler explanation, even if neither is yet proven in a formal sense.