Sean Carroll wrote a compatibilist discussion of free will that Jerry Coyne, being a strong determinist, argued against. I want to look at Coyne’s objections and why they don’t really touch Carroll’s or an compatibilist’s position, and then at what objections can be levelled at Carroll and almost all compatibilists.
But first, a quick summary of Carroll’s position. Carroll likens free will — and probably more accurately, choices — to baseball. Baseball doesn’t really exist at the atomic level, for example. Sure, particles and atoms move around and with enough knowledge of it you might be able to predict what was going to happen next and trace through the whole game that way, but you wouldn’t really get anything interesting about the baseball game out of that, and you couldn’t usefully describe a baseball game using particle terminology. You’d be missing something important about the emergent property — as Carroll puts it — of baseball, which should include things like what a pitch is, what a bunt is, what a home run is, and so on and so forth. Thus, there are interesting concepts about baseball that you couldn’t pick up from doing physics; at some point, you need to enter into the “baseball” model and use terms and descriptions from that model to understand and express those concepts.
This is not a new idea, and is generally a reaction to eliminativist paradigms. If one tries to say that we should describe biology or chemistry or psychology or economics purely in terms of a lower field — like physics — we often find that useful concepts get left out. There’s just something missing from that level of description that we find really interesting about those actions if we try to reduce it to the lower level. Thus, while the concepts may reduce to lower levels — and even that isn’t certain — they aren’t eliminated by them.
Carroll thinks that free will is one such thing, but it’s probably more accurate that he thinks that “choice” is such a thing. His argument seems to boil down to the idea that modelling humans as choice machines is useful and introduces useful concepts even if all choices are ultimately determined by the strict physical level. We “choose”, even if choice isn’t the sort of grand, anti-causal thing that libertarian concepts of choice insist on. Choice is a meaningful concept even if that choice is ultimately determined.
Coyne objects that this will cause problems wrt punishment and the like:
The problem here is that maybe criminals are no more “responsible” for their crimes than are miscreant “young children” and “the mentally ill.” If all acts are, as Carroll believes, physically determined on a macro scale, then there’s no difference between these groups, and any lawbreaker is as guilty as any other.
The problem is that this works for a strong determinist position, but not for a compatibilist position like Carroll’s. Carroll is arguing that our concepts of choice and responsibility are mostly find as long as we don’t try to stuff it into an acausal spirit thing. For Carroll, choosing is a material process determined just like any other material process. But since it is a material process, it is impacted by the same potential issues and impacts as any other process. Thus, there can be physical differences that impact this proces, and thus impact choosing. One of those is immaturity; the physical choosing systems aren’t developed enough to be able to make choices that confer what we call responsibility. Another is damage; the systems may be damaged in such a way that means that their ability to choose is impaired. This, then, is how we separate young children and the mentally ill out in these cases. But there’s no more issue with this than there is in our classification of apples as food, despite the fact that when an apple isn’t ripe it shouldn’t be considered food, and that when it’s rotten we don’t consider it food either. All of these are strict physical notions and stages in the “life” of an apple, but we can treat those stages differently with respect to how they impact our classifcation of “food”. Carroll can indeed do the same thing with the classification of the material process that we call “choice”.
Thus, for Carroll, what choice and responsibility are is just this sort of material — and thus determined — process. Ultimately, this is how compatibilist positions work, by either redefining what choice and responsibility mean so that we don’t need that strong a version and so don’t clash with determinism, or by weakening determinism so that there’s some room for choice. Carroll, like Dennett, chooses to weaken the concept of choice, and gives us “free will worth wanting”. That’s a concept of free will and choice that preserves all of those philosophical issues that Coyne sees, but still preserves determinism. So Coyne’s objections are misplaced, because he’s foisting a definition of choice and responsibilty on Carroll that Carroll is actually rejecting (whether Carroll’s really aware of that or not).
However, this does lead into the real objection to Carroll, and surprisingly it’s almost the same one that Coyne makes: does this preserve notions of responsibility? Coyne’s initial rejection doesn’t work because Carroll can sidestep that by saying, as I argued, that that’s not the defintion of choice he’s working with. But we can reply by challenging whether that redefinition of choice is indeed anything like choice. Carroll’s view of choice is like a computer’s way of choosing: take the system state and do some switches based on that. And yeah, that’s a form of choice. But is it the form of choice we need or think we have? Is it a form of choice that can actually ground real notions of responsibility? Can his form of choice really preserve that distinction, or are all forms of determination really at odds with the concept of choice as it needs to be to work?
Carroll, like all compatibilists, is trying to have his cake and eat it, too, by preserving our intuitive notions of choice and responsibility while maintaining that it is indeed always determined. You need more than to say “This is another level of description” since the issue is over whether those sorts of choices really count as choices. Carroll can argue all he wants that considering ourselves as intentional or choosing machines is a nice model, but that won’t in any way address whether those really are choices or only pseudo-choices. Does Windows really choose to crash or shut down a program, or is that only a pseudo-choice based on us interpreting that choice as more intentional or free than it is? That’s the question that Carroll must answer, and he isn’t even close … at least, not yet.