> simply not very useful for achieving the goals that a dictionary has.
Apparently the dictionary's goals are not the same as my goals. Sure, I want to know about incorrect usages; but I also want to know that those usages are incorrect.
Say that an English speaker speaks a sentence, and an English listener understands it. If they both agree that it was an English sentence -- including no jargon -- and they both agree on the meaning of the sentence, then that's correct English. If the two individuals have also never met before, then it's certainly correct English.
That's what "defined by usage" means. English does not have a language regulator or language academy.
English does have lexicographers, though. I should be able to turn to a dictionary to find out what "literally" means, and it's regrettable that that inquiry will tell me the word has two directly-opposed meanings, without noting that one of them is wrong.
I think you are mistaking the map for the territory. I think you are blaming the data (actual usage) when the model is wrong (dictionaries).
"Literally" is allowed to have a valid and true definition of "figuratively" because exaggeration and hyperbole are used for rhetoric and expression. That's a vital and popular way language is used, and in the case of "literally" it is so commonly encountered that it worth noting in descriptive texts that it's common.
FWIW that's also how languages work that do have an academy. The academy may be a participant in the process but they don't control it, that's basically just a relic of before we understood how languages work.
> that's basically just a relic of before we understood how languages work
Relics that unfortunately maintain significant influence in at least France and Germany.
When I grew up and was just learning to write, orthography was "revamped" by literal committee, in some cases even going so far as deriving a new spelling via false etymologies.
It wasn't really a big deal for me practically, but it just seems bizarre.
I understood what you meant I was just trying to gently point out that it's not a goal that is valued or even taken seriously by people who study language, including lexicographers. Some curiosity about why that is could take you pretty far here.
I know why it is; I know that language changes, and I know that one task of lexicographers is to record those changes. But I don't think they should give the same status to a usage that is just a decade old, and restricted to casual chit-chat among teenagers, as one would accord to a usage that is established over centuries.
According to TFA, "nonplussed" is from Latin "non plus" (no more), and it doesn't seem to have ever had that meaning in English. So I don't really care about "nonplussed". I do care about "literally" (by the letter), and the fact that lexicographers treat its usage to mean "figuratively" as perfectly legitimate. At least, the dictionaries should point out that the version I consider wrong is slang.
The result is that the word "literally" can't now be used in precise discourse, and you have to find some awkward circumlocution. This kind of abuse makes the language less expressive, and is cause for regret.
I had to check m-w's entry for "literally". While they do include a sense 2 that is similar to "figuratively", there is a substantial note afterwards and an FAQ that does a great job of explaining the status of the two senses and a touch of their history. They give some facts that contradict your above, as well (related to the timeline, as well as what constitutes slang). In case you're interested: https://www.merriam-webster.com/dictionary/literally
I do think this is what the dictionary should say. It describes actual usage, including disagreements about it.
By the way, I do think there is a place for linguistic prescriptions, it's just in style guides rather than dictionaries. E.g., fine for the Economist to decide "literally" will only be used in its literal sense in their pages.
Apparently the dictionary's goals are not the same as my goals. Sure, I want to know about incorrect usages; but I also want to know that those usages are incorrect.