Reproducibility is not a problem here. Falsifiability is.

That said, this can't be "useful data" unless you are willing to accept non falsifiable interpretations. There's no way of formulating such an interpretation.

Whether or not we call it science is semantics, highly loaded semantics. We could stop calling it science, but then we'd have to also stop being derisive of unscientific methodology.

More realistically, maybe we should rename "the scientific method" to "the Popperian standard." After all, both the term science and its associated culture predate the definition/standard you are alluding to.

The researchers themselves haven't done anything wrong. They're researching interesting things in theory field, and publishing in the format journals demand... ostensibly to satisfy objections such as yours.

You can't have it both ways.

The process of science needs this useful data - a problem is that we currently have coupled the publication of these useful observations with scientific conclusions, which results in publishing lousy conclusions and not publishing valuable observations (e.g. negative results); perhaps it would be better to decouple these things so that people can (and would be motivated) to publish their primary studies/observations, where there would not necessarily be anything falsifiable, and the conclusions can be done afterwards by others, preferably combining multiple observational studies (thus being not only reproducible but reproduced), like e.g. meta-studies often used in medicine.

I’ve long stopped using the phrase “social sciences” and use “social studies” instead. (I include medicine here, or at least studies with 5% p-value… not science if compared to e.g. 5-sigma physics.)

How is it useful if it does not expand our understanding of the world? What use does it have?

You probably can’t stop other people from using words you believe or even know to be inappropriate. Thankfully it’s sometimes easy to spot science that’s not really science. It’s the stuff that seems legit but isn’t that’s a problem. I’d say the problem is actually bigger in harder sciences. Anybody can look at most social science research and see it’s dodgy. Collusion rings in comp sci research are harder to detect, just to pick an example.