> 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 wa

> 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. I think your main problem is that you only read half of the comment you responded to. It has two parts: > There's something distinctly unsatisfying about this sort of paper, because the number of potential solutions to the problem they're trying to solve is in principle vast, and you can arrive at any number of them if you laboriously work backwards and fit your equations to the data. Which you've adequately responded to. But it also has: > Without very rigorous efforts towards empirical validation, in novel ways if required, this sort of thing is just another wholly speculative theory to add to the large and growing pile. Which you haven't. Taken together, the parts are obviously talking about a class of theory that is fit to data AND does not bother checking the theory against other data or making predictions in reality. Your examples of theories that evolve don't seem relevant, as they're evolving because people made efforts to test the theory out and account for unexpected results (or, at the very least, look at datasets that weren't the ones that the theory was specifically designed to account for).