Nunc per speculum

The difference between physics and metaphysics is not as clear-cut as one might assume at first. Science works by making metaphysical assumptions on the nature of the world, and deriving from those assumptions experimentally testable conclusions. Failure to disprove the conclusions gives us confidence in the plausibility of our metaphysics but it never definitively demonstrates it. One day, a person clever enough might set up an experiment which is in disagreement with our theory, forcing us to rethink our understanding of the world.

This constitutes the major limit of science: the impossibility to ascertain the truth, only approximations of it. Videmus nunc per speculum et in ænigmate. Never facie ad faciem, face to face: absolute knowledge of the truth pertains to the realm of divine revelation, not science. We are stuck with this. A person committed to the scientific method, as I am, ought to be committed to its limits, and this is one of them.  

We can make metaphysical assumptions about the world, but can never prove them, only disprove them, given enough time and ingenuity. Einstein assumed that gravity was really the curvature of spacetime, and derived General Relativity from it, which has been confirmed time and time again by countless experiments. Does that mean that we should start believing in our hearts that gravity is identical with spacetime deformation; that gravity, in a metaphysical sense, is just spacetime warped in different ways? Einstein would have cautioned against that, I believe (or maybe not? Human beings are very protective of their children, intellectual or otherwise). It could be a number of different things. Our universe could be a simulation, and gravity just a module in it. Or, as many physicists now believe, gravity and spacetime could both be emergent from primitive non-temporal and non-spatial entities.

The point is, we don't know. We can make educated guesses on the nature of the world, but we don't really know, not like Plato wanted, and we will never. Our knowledge is tentative. 

Falsification is one limit. There are others. Our metaphysics might be wrong because it doesn't predict the right result in every given experiment. In other words, it has a regime of validity, as it's probably the case with Einstein's spacetime. Fine. But even in the regime of validity where our theory works well, can we be sure that the picture of the world it gives us is the right one, or even accurate?

Not really. Take quantum mechanics as an example. There are many ways to understand what a quantum particle is actually doing in quantum mechanics, i.e. many interpretations of quantum mechanics. The mathematical theory is of course exactly the same, as it should, as we better not mess with the experimental results, but the theoretical frameworks that give rise to it can be completely different and entail completely different metaphysics. 

Particles can be waves that collapse under observation (objective collapse theories), or they can be point particles that take every possible available path from A to B simultaneously (Feynman's path integral), or they can be part of an objectively real universal wavefunction that never collapses (many worlds). The metaphysics in each of these three interpretations is completely different, ranging from one world with interacting waves that collapse under observation to a possibly infinite amount of parallel non-interacting and deterministic universes where collapse doesn't even exist. 

Nevertheless, all these theories give rise to the same mathematical formalism, and therefore the same testable predictions. Which one is the correct one? It is mostly a matter of taste. Someone might be able to find differing testable predictions for each of these theories, but the main point remains: within a given regime of validity, the mathematical formalism of a physical theory might arise from completely different metaphysical assumptions on the nature of world, and we have no objective way of deciding which one accurately describes our universe, because we are stuck with the limited epistemological tools science give us. 

In other words, there's a problem of multiple realizability at play here: a scientific theory can run on multiple metaphysical platforms, and we have no way of selecting the correct one, the one that actually describes the world that we see.

I'm not an anti-realist. I believe Nature is one, and comes in one form (or perhaps not? speculations for another blog post), so there is a Truth somewhere. It is just inaccessible to us.