> Hardly. In fact, just the opposite is true: You can say it about virtually no theories outside a niche within modern physics.
Again I think you're missing the point. There's many ways in which you can construct theories on any topic. They're all just models: which can be seen by how these models often evolve over time, and yet were at each time seen as fairly valid. This is true in all the sciences, and is inherent to science. You make like that's a criticism of modern physics only, but I think it's more an effect of the point below about measurement.
> In other disciplines, e.g. biology or chemistry, it's effectively forbidden to work backwards from data and come up with a "solution" that fits but can't be validated in any way.
in any way: I think this is a mischaracterization. Fitting with the data is a validation in some way. You seek orthogonal validation via new testable predictions, which is fair enough, but it's also fair to consider the measurement problem again (mentioned below), and the process of development wherein advances may be made before testing is figured out, if the domain is of sufficient complexity.
> Chemical space is vast; if you have a complex molecule, there are any number of ways retrosynthesis will work; finding one such way, without empirically validating it, is of near-zero value.
That's the thing. These synths look like they work, on paper. The groups and charges move around right, but they don't actually work, which contradicts your point that you can't devise a large number of plausible alternatives in chem. The same is true of bio and metabolic pathways.
In practice, often we don't know how to test something at the time we sketch it out. I guess that's the consequence of being at the frontier, there's much unkown. So the analogy between these disciplines and physics fits, but perhaps not in the way you intended.
But on another level, the analogy doesn't work, as finding synths is more an eng problem. You don't have to understand why a synth works, for it to work...which is often the case. And the theories about why that was allowed by enthalpy or catalysis or whatever are often evolved over time. Whereas theory building is more focused on having a why that might bring insight.
I think your main problem here is the false dichotomy you see with "work backwards from data to solve the constraints" and "this doesn't validate it in any way". In fact, in the world of theories, fitting the data, is a pretty fucking great validation hahaha! :) But I do get the sense you are expressing about the futility, which I think is real, but just not the whole picture.
> The great theories of 20th century physics were empirically validated almost immediately. Even their thought experiments were subjected to intense scrutiny and debate. Now those theories are fodder for engineers who tinker with material devices like MRI machines.
almost immediately: is inaccurate. Theories can make many predictions, and some of GR and SR were only validated recently.
I guess you can say that our measuring ability has not caught up to our theorizing or mathematical ability. Which is regrettable, but not a condemnation of the theories, as you seem to think.
I get the impression you advance that modern physics is in the business of producing time wasting untestable theories. Which is a fair enough take. I just think there's more nuance there, and you risk maligning a good theory, with this broad stroke.
> How do you reconcile this with the frame-dragging effect? Namely, in that it shows that the distribution and movement of mass (not just the presence of mass/entropy) can influence the curvature of spacetime and thereby gravitational effects.
I don't see that it contradicts it, so it may not be the best counter example. The movement or spin of the mass that induces frame-dragging can be considered information as well.
> Information is degraded over time; the universe will eventually contain zero useful information content ("heat death") and will then be modellable as a homogeneous space subject to statistical fluctuation.
I know that's the model, but what if it's inaccurate/only a part of the picture? What if the information is elsewhere and we don't know how to measure it? There's precedent for that: there's information in quantum states that was not known to exist or be measurable before quantum theory.