Weirdness I noticed about Kant's theory of the categorical imperative: he says that the CI is "synthetic," in the second Critique using the very imposing phrase "sic volo, sic jubeo" to characterize the weird way the CI "announces" itself to us.
So, the first thing is that the "Act only on that maxim..." expressions of the CI, as imperative sentences, don't have subjects that can contain their predicates. They don't have subjects at all, so much, although in at least American English a subject is "understood" for these (usually "You"). How can an imperative be "synthetically" justified without two different concepts to mediate in the first place, intuition or no?
Kant offers an even weirder explanation in the Groundwork (IIRC) when he converts the pure imperative into an ought-sentence and says that the "I" in "I ought to..." is synthetically united with the "ought to..." in the way of divided rational beings (those whose inclinations can potentially conflict with their reasons) not being necessarily united with right actions. In other words, any statement like, "So-and-so did the right thing," would have to be synthetically justified (except if "So-and-so" is God, Kant argues). However, the conditional "an ought is a would, would that one were only a rational and not also an empirical being" is really analytic, on this account: "If I were only able to act reasonably, and if X were the reasonable thing to do," is like a sentence-scale subject and "I would do X" is then the predicate "analytically contained in" it.
Moreover, Kant offers what turn out to be very clear examples of what an analytically justified imperative would be, even as a categorical one: he refers to commands like, "Do good and avoid evil," or, "Act according to the truth." (Allen Wood's writings taught me this, but I don't remember which one right now.) And that's because "good" contains "do" (a positive imperative representation) and "evil" contains "avoid" (a negative imperative representation), and action-according-to-truth just refers to how "right/good" are prescriptive analogues of the assertoric "true."
So, what to do? I have a more involved idea but as far as Kant exegesis goes, the answer I'm going with is: when Kant uses that "sic volo" phrase, he's basically saying that the CI is:
Synthesize [something] a priori.
I.e. the CI is a command to engage in some kind of a priori synthesis.
From what Kant says about the typic of practical judgment, I think this [something] is possible experience in general, but since as he notes, there's no receptive moral intuition (our moral feelings at most are more like acts of moral imagination, which is still spontaneous and not perceptually receptive, and so not as of external objects), the categories have to be used just as a twelve-fold system of the relevant intrinsic character. Hence his table of categories of freedom simulates the universal laws of physics, and hence his reference to forms of the CI in terms of "universal laws of physics" (I know he says "nature" (really, the German word for it?) but by ancient translation isn't that the same as "physis"?).
Finally: since Aristotle AFAIK, the concept of substance is understood as if the world itself contains sentence-like slices of reality (Quine's phrase I think) and some parts of the world are like subjects, in those sentences, that for some reason "can never serve as predicates." Kant plays off this concept when he speaks of the CI in terms of something that must always be an end in itself and never only a means to another end: this goes back to how Kant originally paired the logical type of categorical judgment to his transcendental type of substance. Since agents who contain reason and hence reason's own inner distinctions (as between the hypothetical and categorical levels of the understanding) are hence "categorical" beings, it follows that it is all and exactly these agents who in some way are categorically imperative. Now, that all actually looks pretty analytic to me so it doesn't quite fit with my surface-theory of Kant's claim about the CI being justifiable synthetically a priori but that's a story for another time...
Q: Does this seem like a good way to approach that exact one of Kant's many weird claims?