I’ve already written twice about the views of Christian List on free will, views called to my attention by Michael Shermer (see here and here). My interpretation of List’s views are that while he’s a physical determinist for molecules, he thinks that there is true libertarian “you could have done otherwise” free will on the level of individual behavior. That is, at a given time a person has the possibility of deciding to make any of several choices, and is not constrained by the laws of physics to make only one. That kind of free will, List says, is an “emergent property” that arises in “agential states”: the state of a person contemplating an action.  List’s view is frankly dualist, lacking a convincing scenario for how free will arises in people’s behavior. One might also call it “compatibilist” because, he says, free will on a higher level (the “agential level) is compatible with physical determinism on a lower level.

But the difference between familiar compatibilists like Sean Carroll and Dan Dennett on the one hand, and compatibilists like List on the other, are that with List, determinism of molecules is compatible with pure indeterminism of human behavior. For nearly all other compatibilists, determinism of molecules is reflected in determinism of behavior, but the compatibilism arises when “free will” is construed as something other than libertarian free will (e.g., as Dennett’s “complex input-output relationships”), so that those compatibilists use “free will” in a sense different from how it’s construed by most people and all Abrahamic religions—dualism.

List and others have pulled what I call “The Feser Gambit” on me, though. Theologians like Feser, when you criticize an article they wrote, say something like, “You can’t criticize me unless you’ve read these 1000 pages of my previous writings, which will show you to be wrong.” Likewise, List and his supporters have told me that I can’t properly criticize him unless I’ve read his big paper on free will.

Well, I have. It’s called “Free will, determinism, and the possibility of doing otherwise“, and you can get it free at the link. It’s long and a bit complicated, and I have to say that it’s almost above my pay grade. (It was published in 2014 in Noûs, volume 48, pages 156-178.)

Having read it, however, I think there’s a central flaw in List’s paper that invalidates his argument for agential could-have-done otherwise free will.

Before I pinpoint that flaw, which I’ll try to do briefly, you will see from the paper that List really is a true free-willer of the libertarian stripe. That is, he really thinks that, with an agent in a given deciding state (he’s not clear about what an “agential state” is), under identical conditions, an agent could make any of a number of choices. Such an agent at the point of “deciding” does not make a single decision mandated by the laws of physics, but has the possibility of making different choices. List takes free will to mean that at a given time, there is the possibility of an agent—the “decider”, to use G. W. Bush’s jargon—to make more than one decision.

List’s “compatibilism” that leads to true libertarian free will depends on his assumption below:

Supervenience and multiple realizability: There exists a (many-to-one) mapping σ from S into S such that each physical state s in S determines a corresponding agential state σ(s) in S, but the same agential state s in S may be realized by more than one physical state s in S. [Bold vs roman characters represent given symbols.]

You’ll see below what he means by many-to-one mapping: it’s mapping from world histories of molecules onto “agential states” of behavior in a given situation.

But what is an “agential state”? It’s not clear what List means, even in this paper, but it appears to be the condition an agent is in when about to make a decision. I didn’t get much help from the definition below, as the parts I’ve put in bold are confusing:

Agential state. Let me introduce the term “agential state” to denote the state of an agent and his or her macroscopic environment as specified by the relevant higher-level theory of human agency. There are various ways of making this definition more precise; it may sometimes be useful, for example, to represent the agential state explicitly as a pair consisting of the agent’s intrinsic state and the state of the environment.But for the general purposes of this paper, I can set these details aside. What matters is that an agential state, while supervening on (being fully determined by) the underlying physical state of the world, is more coarse-grained than that physical state.

What he apparently means by this is that there are many “world histories”—the historical sequence of events that puts all molecules and particles in the universe in a given state—that can correspond to a single agential state. This is the crux of List’s argument, as given in both words and in a figure below. In other words, a given agent about to make a free choice could have arrived at that “agential state” by a number of conceivable world histories.  Because the agential state is “coarse grained” in this way, the underlying physical states could go their separate paths at the moment of  decision, somehow giving the agent the possibility to make two or more decisions as the world histories bifurcate, leading the agential states on new paths.

Here’s how List puts it (I’ve bolded the parts implying libertarian free will):

I will argue that free will, even when understood as requiring the ability to do otherwise, is compatible with determinism. Crucially, I claim that this is so even when the ability to do otherwise is interpreted modally, as the possibility of doing otherwise, rather than in some weaker conditional or dispositional sense. The key idea is that although determinism implies that only one future sequence of events is physically possible given the current fully specified state of the world, the more coarsely defined state of an agent and his or her macroscopic environment can still be consistent with more than one such sequence, and thus different alternative actions can be possible for the agent. The notion of agential possibility will be defined and defended in detail below. In particular, I suggest that this notion – and the notion of free will analyzed in terms of it – is no less scientifically respectable than other higher-level notions we routinely employ in intentional explanations, such as beliefs, desires, and intentions. However, I will also identify conditions under which new scientific developments might force us to give up this compatibilist view. Notably, these conditions would challenge not only free will, but also our established intentional approach to explaining human behaviour.

. . . My main point follows immediately. Given the multiple realizability of agential states by physical states, it is perfectly possible for the many-to-one mapping σ from S into S to be such that determinism at the physical level is consistent with indeterminism at the agential level. While any physical history (in Ω) may have only one possible continuation at any time, namely the history itself, there can be two or more distinct agential histories (in Ω) that coincide up to time t but then branch out in different directions.

And here’s his figure, with an explanation, showing how he thinks this happens:

Figures 1 and 2 show a simple example. Figure 1 represents the physical level, Figure 2 the agential one. The dots in Figure 1 represent different possible physical states, and the lines connecting them different possible physical histories, over five time periods (t = 1 to t = 5). Thus S is the set of all the dots, and Ω the set of all the lines. It is easy to see that determinism holds: there is no branching in any of the possible world histories. Now suppose the agential state of the world supervenes on the physical one and is multiply realizable. Specifically, all physical states that lie inside the same rectangular cell in Figure 1 correspond to the same agential state. At time 1, for instance, the supervenience mapping σ maps the three left-most physical states to one agential state, and the three right-most physical states to another. Figure 2 displays all the resulting agential states and histories. Here the thick dots constitute the set S of possible agential states, and the thick lines (to be precise, all the linear paths that can be taken along the trees from bottom to top) constitute the set Ω of all possible agential histories. It is easy to see that there is no determinism at the agential level in the present example. The agential histories branch out in various ways. As this example shows, such agential-level indeterminism is entirely consistent with determinism at the physical level, jointly with supervenience and multiple realizability.

Note how three different world histories (the three dots in the third box from the left at t = 1) map to a single “agential state” at the same time. But then that agential state bifurcates, reflecting the different paths of the antecedent histories.  The bifurcations occur when the underlying world histories in an agent diverge to the point that a different choice is made. As List says, this confers true libertarian free will. (My emphasis below.)

However, as I have suggested and will defend further in the next section, this does not settle the question of what actions an agent can and cannot do. This is because the appropriate frame of reference for asking whether a particular action is possible for an agent is not the one given by fundamental physics, but the one given by our best theory of human agency, and such a theory employs a more coarse-grained state space than the physical one.

It’s not clear, either, what he means by “our best theory of human agency”. That is irrelevant to whether one’s “choices” are inevitable given the laws of physics.

The problem with List’s idea. While it’s conceivable that different histories of the universe or its constituent parts could produce someone in the same “agential state”: that is, a guy standing in front of an ice-cream counter trying to choose a flavor, this does not solve the problem—how to give that person libertarian freedom to choose different flavors. For at a given moment, a person about to make a “choice” is in only one physical state, not different physical states. And that one physical state, even if arrived at by different past histories, has only one future history: the history mandated by the laws of physics (with perhaps some slight differences due to quantum indeterminacy, which doesn’t confer agency). In other words, there is no way that an individual at a given moment, with everything held constant in the Universe, could have behaved otherwise. The bifurcations that List sees as corresponding to different free “choices” cannot occur, even if different past histories produce an identical configuration of the Universe.

And make no mistake about it: the possibility of different choices at a given moment by a given agent in a given situation are what most people think of free will. List agrees that this is indeed what free will is, and that it’s truly libertarian. But what we have here is not a series of histories that allow such coarse-grained bifurcations, but a single state, represented by a person in a situation where most people think “choice” is possible. For example, a robber with a gun points it at a cashier, who refuses to hand over money. That is a single situation represented by a single individual with a gun. Perhaps it’s possible that different conceivable histories could have produced the identical situation (and remember, by “identical” I mean “the configuration of the universe is identical at that moment”), though I see that as unlikely. But even if they could have, at the moment when the robber’s finger is on the trigger, it is the configuration of molecules in the Universe at that instant—the configuration of the particles in his brain and the cashier’s brain, in the surroundings, and so on, that determine what the robber will do. The past has become irrelevant: all that matters now is the laws of physics applying to the physical state we’ve arrived at.

And so—or so I think—there are no free choices possible, no bifurcations of behavior at that moment that can be selected by “agential” will.  List has erred, and has not given a credible account of libertarian free will on the individual level.

At the end, List describes two situations in which his theory could be refuted:

1.) “A successful reduction of psychology to physics—perhaps the dream of some neuroscientists—such that determinism at the physical level would imply determinism at the psychological level.”  Here he clearly shows that he thinks these two levels are only loosely connected. But even though we’re not yet able to connect psychology to molecules and physics, that is the only possibility we can think of save some spooky phenomena we haven’t yet understand. Sean Carroll thinks this is not possible: that we fully understand the physics of everyday life, and though Sean is a compatibilist in confecting a non-libertarian definition of free will, he (like Dennett) is still a pure determinist on the “agential” level.

2.) “A paradigm shift towards a deterministic theory of psychology, even in the absence of a reduction of psychology to physics.” While this sounds easier to attain, it is not, for it demands that we have enough information to predict with complete certainty what somebody is going to do, even if we don’t know the underlying physics. I maintain that we can’t attain such a theory without knowing at least lower-level molecular phenomena—the behavior of neurons, neurotransmitters, and so on—to arrive a near-perfect predictability of behavior and of choices.

While #2 seems unattainable, it follow immediately if goal 1.) is attained, and goal 1 is true according to our best knowledge of physics.

So I have read List’s Big Paper, and I am not convinced that it comes anywhere near giving evidence for libertarian can-do-otherwise free will. It “succeeds” by making a sleight of hand: an unwarranted claim that an agent in a given state could do otherwise because that “given state” could reflect different molecular histories. But a given agential state is a single configuration of the Universe, and its past is irrelevant. Once that state exists, its future is determined.