[fasteddie31003]

That’s true in a narrow sense—every observation records something that has already happened. But in science, observations can be tested, replicated, and used to predict future outcomes. The kind of "research" I'm skeptical of draws broad, causal conclusions about unique, unrepeatable past events where none of that is possible.

[layer8]

Usually these are predictions made by a model that has explanatory power for things that we can observe. The model might be wrong, or there might be a better model. That’s always the case in science. Observations that confirm a model also increase the credence for its predictions that we can’t directly observe. It means that given our best current understanding of X, it also implies Y. Yes, Y might be wrong, but then that implies that something is likely also wrong with our current understanding of X. The predictions (or retrodictions) aren’t black and white. They always have some associated level of credence, which depends on how well we think we understand the kind of system we are talking about.

[drob518]

It’s always a theory, but what choice do you have? You can’t rerun the experiment again under controlled conditions. Your only choice is to theorize or not. Sure, there is more possible error in such theories compared to other theories where you can rerun the experiment multiple times to test it, but that doesn’t mean that a theory that can’t be tested is wrong.

[AStonesThrow]

It’s actually seldom a Theory. In fact, I’d be surprised if scientists were eager to form “Theory on the Formation of Jupiter” [or Venus or the Moon] because those are already quite specific subjects that should be derived from generalized theories.

We’re thinking of hypotheses and proposals and extrapolations here. A few scientists analyzed and compared existing data, and they estimate things and interpolate and ... guess ... And the narrative develops as they give facility tours and answer questions for journalists, 8-year-olds, and 8-year-old journalists.

Theories are hypotheses which were tested and survived falsification attempted against them. You cannot adequately falsify “800 million years ago, and 12 parsecs away...” but you can enjoy your colleagues’ version of the story over drinks with a jazz band.