I'll start with an extended SEP quote from the J. L. Austin article (when I copy text here from the SEP, it drops the italics, etc. formatting, and normally I would go back through the copy and fix that; but this quote is really long and the number of fixes I would have to make are prohibitively many):
Austin considers this issue in his “Ifs and Cans” (1956a). There, he discusses and rejects attempts in G.E. Moore 1912 and Nowell-Smith 1954 to provide accounts of what we can do on which it being true that we can do things (/refrain from doing them) is compatible with its being determined by our circumstances that we don’t in fact do them (/refrain from doing them). Austin thinks that his objections to the accounts on which he focuses provide partial support to the view that our ordinary claims about what we can do are incompatible with determinism.
The first proposal of Moore’s that Austin considers is that the claim that someone, S, can do something, A, is equivalent to the following claim: S will A, if S chooses to A. Austin argues that Moore’s first proposal is based on a mistaken view about the functioning of the claims of the form: “S can A, if S chooses”.[25] The second proposal of Moore’s that Austin considers is the claim that “S can A” is equivalent to a claim of the form “If it were that C, then S would B”. For example, “I could have holed the putt” might be taken to be equivalent to “If I had tried to hole the putt, I would have succeeded in holing the putt.” Here again, it seems that the proposal might be of service in sidestepping the challenge posed by determinism. Suppose that in actual circumstances C, I don’t try to hole the putt. According to determinism, it follows that it is impossible that (C and I do try to hole the putt). But that is consistent both with its being possible, in slightly different circumstances, that I try to hole the putt, and with it being the case that, if I were to try, I would succeed. Austin doesn’t pursue the proposal in detail, though his discussion of Nowell-Smith pursues connected issues (219–230). However, in a footnote, Austin presents an important putative counterexample:
Consider the case where I miss a very short putt and kick myself because I could have holed it. It is not that I should have holed it if I had tried: I did try, and missed. It is not that I should have holed it if conditions had been different: that might of course be so, but I am talking about conditions as they precisely were, and asserting that I could have holed it. There is the rub. (1956a: 218 fn.1)
Austin’s thought here is that attributions of this sort manifest our belief that,
…a human ability or power or capacity is inherently liable not to produce success, on occasion, and that for no reason (or are bad luck and bad form sometimes reasons?). (218: 218 fn.1)
Now a committed determinist would claim that the events that constitute such failures must be determined—and in that sense explained—by the circumstances at and before the failure. But Austin believes—for reasons in effect considered above—that the existence of such an explanation would make it the case that, in fact, the golfer could not have made the putt in the circumstances as they precisely were.
We have here, then, a point at which Austin expresses his view that ordinary attributions of ability, power, and capacity are incompatible with the thesis of determinism. In response, the compatibilist is forced, I think, to deny that its being true that a golfer could have holed the putt, or even that he could have holed the putt in precisely the same conditions, entails that he would have holed the putt in a perfect duplicate of the actual world. Leaving that issue to one side, Austin presents the proposed analysis with a case of masking—a case in which, although ability is retained, successful exercise of the ability is somehow precluded, for instance by outside interference (for discussion of masking, see A. Bird 1998; Clarke 2009; Fara 2008; and Johnston 1992). The challenge facing the defender of the analysis is to spell out the analysis so as to cope with masking. Arguably, meeting the challenge depends upon provision of a non-circular specification of all possible masks. Austin would, I think, claim that it is impossible to meet the challenge. For on his view, abilities are sometimes masked brutely, without any specifiable mask. Even if he is wrong about that, it remains an open question whether the challenge can be met, or whether the endless heterogeneity of potential masks makes it impossible to provide an explanatory specification.
Now a gamebook is usually known (in English) as a "choose your own adventure" book, after the most famous series of gamebooks in English. The way they proceed is they have a set of scenes culminating in a choice. Sometimes the choices are declaratives: "If you choose to do X, then you go to page Y." But usually, IIRC, the choices are conditional imperatives: "If you choose to do X, then go to page Y." I wrote my own gamebook one time, and I encoded a choice in an even more exotic way: the scene culminated in a single imperative sentence, "Go to page X," except the imperative was a question, "Go to page X?" The idea was that your answer to the question constituted the making of the choice: "Yes," meant going to page X, whereas, "No," meant stopping one's progress in the book altogether. ♍
The "rub" is that gamebooks do not present choices in the Moorean/related sense that Austin criticized. They do not have parallel scenes leading to parallel imperatives. They work with the "same initial conditions, different outcomes" image of free will: the "principle of alternative possibilities."
Are gamebooks "evidence" that Austin's thesis on the ordinary sense of "free will," is the correct thesis? (I don't mean whether incompatibilism about free will in itself is true; I mean whether the ordinary-language description of free will is incompatibilist. There is another fundamental description of free will at stake, though: "the causal background style necessary and sufficient for moral responsibility." It is as a satisfaction of this latter that a Frankfurt mesh is supposed to count as free will: certainly not because a Frankfurt mesh is what we "ordinarily" mean by "free will," seeing as the dialectical illustration of such a mesh involves an esoteric kind of possible mind control.)
♍This played into my theory that strong (AKA incompatibilist) free will is metaphysically equivalent to (or congruent with) our ability to ask questions. Among other things, considered that this ability allows our most interior consciousness, our thoughts and imagination, to "step back" (c.f. Korsgaard!) from the concept and assertion of causation. More exotically, imperatives, not being truth-apt, cannot be identical to the meaning of deontic assertions, yet they seem intimately bound up with them, to the point that Mackie, for example, spoke of "objectively prescriptive properties" along these lines: the semantics of a deontic assertion encode imperatives, so that making the assertion is semantically simultaneous with issuing the imperative. And then it is a special class of questions, however, that can take imperative syntax and logically transconstruct an assertoric function therefrom: namely, "Why?" questions. The primitive erotetic function of, "Why?" will reconfigure an imperative: "Why do that?" then allows for, "J is why do that." And this is the preimage of deontic assertions (as assertions of moral reasons/"whys").
So all that being said, if it is via an ur-elementary erotetic function, that deontic assertoric functions are able to be given, then if "ought" implies "can" in such a way as to imply "can" in terms of "the principle of alternative possibilities," it is again our erotetic capacity that establishes our strong free will. QED
