I could say the same about acceptability. You either accept something or don't. Grudging acceptance is still acceptance, right?

Meanwhile, you can assign a mathematical measure of partial correctness for, say, sets of statements, simply as the fraction of correct statements over the total. In linguistics, you can assign measures to both acceptibility and correctness simultaneously by taking the fraction of people who "accept" a construct as "correct".

I'm not even sure how many levels of meta I'm at, so I'm done.

> Something is either correct or it isn't, is the point.

    Correct \Cor*rect"\ (k[^o]r*r[e^]kt"), a.
    [L. correctus, p. p. of corrigere to make straight,
    to correct; cor- + regere to lead straight: cf. F.
    correct. See {Regular}, {Right}, and cf. {Escort}.]

    Set right, or made straight; hence, conformable to
    truth, rectitude, or propriety, or to a just standard;
    not faulty or imperfect; free from error; as, correct
    behavior; correct views.
     [1913 Webster]
  
           Always use the most correct editions.    
           --Felton.

Granted, 1913 webster is almost an archaic resource at this point, but it's interesting that its very first example of use is "most correct".

> Something is either correct or it isn't

If you are building truth tables, then something is either correct or it isn't. But that's certainly not the case when using words to communicate ideas between humans.

After all, haven't each of us taken a multiple choice test with instructions to choose the most correct answer?

When it comes to diction, there are always a variety of choices. You make those decisions based on whichever word or phrase works best given the circumstances.

It's no different than choosing the most appropriate programming language for a given project; plenty will work, but ultimately one is the most logical given who will be writing it, reading it, using it, maintaining it, deploying it, etc.

Einstein's theories about the universe are more correct than Newton's. Neither are "correct".

Reminds me of an old quote by George E. P. Box - "All models are wrong, but some are useful"

Guys -- use whatever the fuck you'd like.

Except in any domain where correctness is not black and white.