One way to frame the issue would be to look for connections and differences between the following imperatives:
- Do x.
- Decide (or choose) to do x.
- Intend to do x.
- Decide to intend to do x.
- Intend to decide to do x.
I think of the difference between (1) and (2) as involving a hidden disjunct: if we say, "Choose x," we implicitly refer to some other option, something ~x, so that the imperative "should" read, "Choose between x and ~x," or, "Choose to do x or ~x." Now, to say that the choice is involuntary, or deterministically caused, or whatever, is not to say that the resolution of the choice is determined strictly, but only that the setting of the choice problem itself is.
Note also the following peculiar "commands":
- Possibly do x.
- Necessarily do x.
- Decide to possibly do x or ~x.
- Possibly decide to do x or ~x.
- Decide to necessarily do x or ~x.
- Necessarily decide to do x or ~x.
With respect to (7), to me the idea of commanding someone to do something that will be necessarily done, is "absurd," or at any rate for other reasons I think that imperative syntax is unintelligible if we suppose that imperatives necessitate compliance therewith. When I tell someone to do something, I don't have a sense that my telling them what to do would force them to comply with my directive, and I wonder why I would feel angry with someone if they didn't comply, yet their lack of compliance was owing to some independent cause that did compel their noncompliance. (Though see also the issue of reactive attitudes and moral responsibility.) —With respect to (10) and (11), either command seems perhaps trivial, at least if we assume that the law of the excluded middle applies to actions/choices. (And then though Zermelo was at pains to claim that the axiom of choice "has nothing to do with" action theory as such, yet by virtue of the connections we have discovered between the LEM and Zermelo's special rule, ironically it has turned out that there might be some bearing on the practical question, from the theoretical side of things as such.) But so if the presence of a decision problem is necessitated, it could be in these terms, and yet just for that reason, it is not clear (to me, at the time of this writing) that said prior necessity (and hence involuntary conditionality) is inconsistent with the resolution of the internal disjunction itself being contingent, as far as its outcome goes.
- Pick a number between 0 and ℵ0.
- Randomly pick a number from 0 to absolute infinity (anti-zero, or V).
- Randomly decide between 0 and 1.
- Pick a real number at random.
- Intend to randomly pick a number from surreal -ω to ωωω.
- Decide to intend to randomly pick such a number.
- Randomly intend to randomly decide to pick any number from the range of (13).
And so on and on... (I mention these "options" only to highlight the ambient question of randomness and free will, modulo decision problems being caused even if their solutions are not. I have an intuition that, "Randomly decide to do x," and, "Decide to randomly do x," are both wrongheaded, yet there is a subtle difference between them, so their wrongness is not identical (but see (19) below).)
Addendum. Three other peculiar imperatives:
- Spontaneously do (or decide to do) x. (This might be what someone means to say when they "wrongly" say, "Randomly do (or decide to do) x." It might be equivalent to, "Do x for no reason at all," or perhaps, "Do x just for the sake of doing x.")
- Do something you would only be able to do if you had incompatibilist free will/when given the choice between x and ~x, do that which you could only do (or decide to do) if you had incompatibilist free will.
- Pair doing x or ~x with the outcome of tossing a quantum coin (a coin that somehow lands on heads or tails due to quantum fluctuations). Intend to do x if the coin comes up heads, ~x if it comes up tails.
An edit come quite lately... Let's say the following sequence represents a possible implementation of (retrograde) iterated decision operations:
- ORelim(A OR B) = the act.
- (A OR B) = The fact that I will have done A or I will have done B.
- ((A OR B) OR NOT(A OR B))
- (((A OR B) OR NOT(A OR B)) OR NOT((A OR B) OR NOT(A OR B)))
- &c. ... (I think the increase = 2 times 2 times 2 times ... such that, even if countably many times, we end up with uncountably/Continuum-many disjuncts)
If we had to decide-to-decide forever and ever as such, we would, I think, have to interject a moment of ORelim at every stage. I'm not saying that's possible or impossible; for now, anyway, it just sounds like a logically interesting description (or: a description of a logically interesting structure, or whatever along that line).