> There is actual research in the domain of language complexity! (Which, to be fair, is still a somewhat more niche topic in linguistics)
I know little about this as well, but there’s been some research done by…… John McWhorter, I think? A quick search through my Big List of Things to Read (3000 articles and counting!) gives https://elementy.ru/downloads/elt/mcwhorter_creole_grammars.... and https://benjamins.com/catalog/slcs.71, which you may want to have a look at if you’re interested. David Gil is also well-known for his insistence that Riau Indonesian is simpler than most other non-creole languages.
> https://journals.plos.org/plosone/article?id=10.1371/journal...
Ah yes, this one. I remember carefully reading it a while ago, and I ended up very sceptical about its conclusions. For one thing, I don’t agree with rankings such as ‘Isolating > Concatenating’ and ‘Agent > Agent & Patient’ (though these do admittedly seem to be common assumptions). The ranking as a whole is a bit weird also — I’d love to know how they concluded that Tiwi is less complex than Dyirbal, for example. But I’m more suspicious of the linear fits, which appear extremely bad — I’m not convinced you can conclude anything from a fit of R²=8.9% (for Afro-Asiatic) or R²=14.1% (for Sino–Tibetan). In fact, only one of the graphs has R²>90% at all. And I note that they very carefully avoid giving p-values — except for the overall ‘complexity against population’ graph, a very significant correlation which however disappears when languages are split into groups by family (Simpson’s Paradox again?).
(Though keep in mind that I’m not great at statistics and any of the above could be incorrect.)
Also, they rely entirely on WALS, which isn’t a great source of linguistic data. It’s good for other things, but I very regularly find that a good proportion of its data is incorrect. It’s fine if you want to, say, see roughly what proportion of languages use active–stative alignment, but not for much else. (Though I suppose you could argue that the CLT applies and the errors will cancel out. But as I said, I’m bad at statistics and wouldn’t know how correct this is.)
(Personally, I’m sceptical that there’s a link between complexity and population. Witness Swahili, the lingua franca of a good chunk of East Africa. Or Quechua, the language of the Inca Empire. The main reason most widely-spoken languages are ‘simpler’ is that they disproportionally come from IE, ST and Austronesian, which aren’t the most morphologically complex families anyway.)