Is libertarian free will incoherent?

Question

This is getting long. Here's the TL;DR;

---

The determinism-randomness dilemma has been discussed a lot both on this forum and in philosophy as a whole.

What I was trying to get at here is that the concept of Libertarian Free Will, even when considered by itself (and not against determinism or randomness) has severe conceptual problems.

What about "selecting" or "deciding" is agency-conferring? Even if we grant will as a third causal option, it is differentiated as a brute impulse vs anything responsibility bearing.

To the extent we can explain these "willful impulses" we are back to citing reasons, which beget questions about why we chose to give those reasons, and so on. At the end, we are left with a brute impulse.

So, even granting the existence of a third causal option ("will") it seems like at its core its just an extra-physical impulse, hardly worthy of doing any moral or deliberative work by itself.

---

Edit: 12/2/2024: I realized my presentation below came across as defining away free will from the get go. The intention was to actually challenge the implication that determinism and randomness exhaust the causal landscape.

I've clarified my definitions (and adjusted the terms -- thanks Pertti) as the original ones were too vague. Hopefully that will eliminate the charge of question begging/circularity :)

---

I am trying to understand the coherence of libertarian (contra-causal) free will.

To define my terms (expanded to not imply I am "begging the question" against free will):

Deterministic: An event E is deterministic iff there exists a set of conditions sufficient for it to occur (i.e., sufficient set)
Indeterministic: An event E is random iff there does not exist a sufficient set of conditions (note: we can have purely necessary conditions)
Deliberate: When a decision D by an agent A serves as the sufficient cause of event E.
Random: An indeterministic event E that is not deliberate.

The big question here is if the following is true (quantifying over the domain of antecedent sets of facts/conditions):

Proposition 1
"Event E is deliberate due to A choosing D" → ∃Q: "Q is sufficient for A choosing D"

With this framing, my major struggle is understanding how "A choosing D" is not the result of some combination of deterministic and random causes.

If we truly have agency then we need to have control over our decisions. This leads me to another proposition for a sufficient condition for identifying when A can be said to have decided D:

Proposition 2
IF (the set O of available options includes an option D and at least one more option other than D) and (A can provide an account/explanation of why they chose D) THEN A has decided D.

Some may argue with this definition as it excludes cases where we feel we "decide at random" (for example on a test where we don't know the answer). But I think to include "decided randomly" as "decisions" in the "deliberative" sense begs the question we are trying to resolve -- was that "random choice" really an example of you exercising agency (deliberating), or was it giving up agency and letting the neurons go their course?

If we assume Proposition 2 is true, then Proposition 1 is true within this domain. Every choice has a reason which itself must have a reason etc.

But what if we challenge that? What if you say "No, I gave you my reason but I have no reason for my reason." To me the counter is that you really didn't exercise contra-causal agency, you just played out the inevitable reasoning of your brain, which inexplicably provided you this reason for your decision.

So, when trying to understand our choices we end up going down a rabbit hole of chained reasoning (therapists make a lot of money this way ;) or we end the chain at a brute feeling/impulse (a type of brute fact).

In either case, I fail to see where we have any control above and beyond the control embedded in our neurology and acting through normal physical processes.

---

Example

If an agent A decides to have ketchup on their hot dog (event K), but they could have chosen another condiment (¬K) in what sense did they have such freedom to do otherwise?

If we ask A "Why did you choose K?" then A can give some reason ("I like the taste of K better") or admit to being arbitrary ("I just picked K for no reason").

But if A gave a reason, then if we "wound back the clock" the agent would still posses the preferences and antecedent conditions that led them to choose K, so they would still choose K. Their preference for K (and any environmental factors) is a sufficient condition for A choosing K. They couldn't have picked mustard in that situation.

If we say they could have chosen otherwise, under the exact same conditions, then there is a possible world accessible from the actual one where A suddenly likes mustard in that situation in instead of ketchup - but why? Where did this come from? How is this type of freedom anything like the agency we want for moral agents. If eating mustard were morally bad then we seem to be unable to fully account for why we acted morally in a given situation - at some point there will be a brute impulse driving the more sophisticated machinations of our mind.

If A has no reason, then they are just a random condiment picker (random).

Where is there a need to posit a process that somehow modulates action potentials in our neurons in in a way no pre-determined by physical laws?

--

Answer 1

Kevin Mitchell, a scientist working in the field of genetics and neuroscience, published: “Free agents. How evolution gave us free will. Princeton Press (2023)”.

1. His preface takes Denett as representative for the compatibilist position: “This compatibilist view of free will is widespread among philosophers and scientists, but it cannot explain how free will or indeed any form of real agency can possibly exist in a deterministic universe. […] I do not think that either the strictly deterministic or the compatibilist position is satisfactory.”

2. Mitchell describes his own position: “[…] living organisms do things, for reasons, as causal agents in their own right. They are driven not by energy but by information. And the meaning of that information is embodied in the structure of the system itself, based on its history. In short, there are fundamentally distinct types of causation at play in living organism by virtue of their organization.”

3. IMO Mitchell’s book opens up the search space for an explanation of free will within the causally closed space of scientific explanation. Mitchell introduces biological evolution as a new dimension when investigating agency of organisms from the cell to higher developed animals like humans.

   The principle of downward causation - always a bit suspicious in science - appears in a new light.

   For a talk of Mitchell see here.

   On request I can email a copy of the preface of Mitchell's book.

Added: I remember that Mitchell's ideas and his book were the subject of a former post on this blog.

Answer 2

Libertarian (contra-causal) free will is a philosophical position which holds that choices (particularly human choices) are acts of will, not causally determined events. In other words, say person X comes to a fork in a road and takes the road on the left. There are two possible explanations for this event:

• Determinism: turning left is the mere summation of all previous events in the universe, so X was always going to take the left road, and there is no possibility that they might ever have taken the right road.
• Libertarian free will: X chose to go left, but could have chosen to go right; the event occurred as a matter of X's will, and is not bound to the summation of all previous events in the universe.

This philosophical notion doesn't depend on knowing the element or mechanism that makes free choice possible. It simply asserts that free choice is possible (somehow or other), and works from there. Nor does it make that assertion lightly, since the common human experience is that we make choices that are seemingly free within the constraints of given situations.

The confusion in this question arises because the question presumes a deterministic worldview in which all events must have determinate causal explanations, and so it interprets the concept of 'will' (oddly) as randomness. But will isn't random; will is purposive. Randomness is merely a modern convention which models unknown effects as though they are undiscovered and unpredictable deterministic processes. Libertarian free will asserts that purposive behavior cannot be reduced to deterministic principles, known or unknown, predictable or unpredictable.

Answer 3

Short answer

Yes. I have not seen a definition of libertarian free will that clearly logically closes, and captures everything meant by libertarians about free will.

BUT -- I have not seen a definition of causation that clearly logically closes, and captures everything meant by users of "cause".

All definitions of science have suffered the same problem -- Both Popper's and Lakatos' definitions have been shown to have exceptions.

The similar failure for physicalism even has a name: Hempel's Dilemma.

Most philosophers try to gloss over the issue, but definitional shortcomings are a feature of basically every philosophic viewpoint. This is not a fatal problem. We get by with reasonably approximate definitions.

Longer Answer

The best description I have found is on this blog web site by the Information Philosopher, Bob Doyle, on agent causation: https://www.informationphilosopher.com/freedom/agent-causality.html SEP also discusses agent causation here: https://plato.stanford.edu/entries/incompatibilism-theories/#3 And Helen Steward wrote a good book outlining the plausibility of Agent Causation https://www.amazon.com/dp/0198706464.

These discussions are not quite as explicit as I think they need to be on the three preconditions needed for libertarian free will. I consider them to be:

1. Our physical universe is probabilistic. This is the dominant view of what quantum mechanics tells us, and indeterminacy is known to leverage up to be relevant at macro scales due to chaos phenomena, so this criteria is reasonably taken as true of our world.
2. Our physical universe is not causally closed to agents influencing it. Things like the near consensus that irreducibly emergent phenomena are causal, the placebo effect where our awareness is causal, the evolutionary tuning of consciousness, and every one of us experiencing our willing causing our actions every day, make this highly plausibly true of our world.
3. Causal logic must include at least a third option besides "deterministically caused" and "probabilistic-or-random". This would be a third option, agent causation. There are infinite possible logics. https://www.cambridge.org/core/journals/think/article/guide-to-logical-pluralism-for-nonlogicians/EDFDFA1C9EB65DB71848DABD6B1 What logic applies to what aspect of our world is an empirical question. That we have a three state logic for causation is entirely possible, and exploring possible "agent causation" logic structures until we find one that matches what we have observed of our world is entirely reasonable.

To date, I don't think Doyle or Steward or any other free will advocate I have read has fully nailed an entirely correct logic of agent causation. However, as noted in the short answer, this is a common problem for philosophy, and therefore not a fatal one. The pursuit of Agent Causation thinking is an active and fruitful Research Programme in philosophy.

Answer 4

Willed: Brought about by the action of an agent
Deterministic: An event E is deterministic iff it has at least one sufficient cause
Random: An event E is random iff it is not deterministic.

There are some problems here. First, you are defining randomness to be anything not deterministic. But in libertarian free will it is supposed there is a third option, so it's not proper to just assume there isn't. And in fact, in mathematics, there are other options. Mathematically, "randomness" just means something using one particular abstract structure out of infinite possible others - the abstract structure of random variables. And for that matter, "deterministic" in mathematics also means a certain kind of abstract structure that involves time-varying states which depend on previous states. There are plenty of other structures than those two!

For example, a causal filter could be taken as a simplified model of determinism. A filter in this context is a transformation applied to a time-varying input signal to produce a time-varying output signal. A filter is causal if the output depends only on inputs from the past or present.

But non-causal filters aren't random! Like causal filters, they're the result of applying a formula to obtain a definite output. For non-causal filters, this formula just involves terms from the future. Would you say non-causal filters are deterministic? That doesn't seem right. They're determined by a formula, but unlike what we normally think of as "deterministic," their next state is not determined by past states.

Now, I'm not saying that libertarian free will involves some sort of influence from the future; this was just an example to show that there are more options than just "deterministic" and "random."

I'm also not saying libertarian free will makes sense. Yes, our choices, when we introspect, do seem to be caused by our mental state - our beliefs and preferences. And if they weren't caused by our mental state, we would seem to ourselves to be acting randomly - we wouldn't be in control! That would be quite upsetting to us. To be in control is nothing more than for our conscious preferences to determine our actions.

Answer 5

Willed: Brought about by the action of an agent

That definition is way to broad. A stone rolling down a mountain can bring about an action. Though a stone rolling down a mountain is by most accounts an inanimate object without a will. So either you need to hide complexity in the word "agent" idk "a being that can will", which would make that definition circular or you would attribute a will to actions that people would otherwise group as deterministic, which makes this distinction kinda superfluous. Either way that seems to lack crucial information.

Deterministic: An event E is deterministic iff it has at least one sufficient cause

This definition makes it sound as if singular events could be caused by singular causes. And the example with the condiments further points in that direction. The problem is, that is most likely not the case. Singular events like the choice of condiments would not be considered "deterministic", UNLESS, EVERYTHING is deterministic.

Like suppose you have a strong preference for ketchup, you like the texture, the taste, the color and whatnot. Now suppose I change the lighting so that it appears in the most ugly shit and disgusting color you could imagine, suppose I crank up the heat to simmer away the texture and taste, that I mess with the pressure to manipulate your taste buds and whatnot. Would "preference for ketchup" still be a sufficient reason, meaning a reason that if you knew only this you could tell the outcome of the choice 100%, anytime?

Because if you can't then that wouldn't be deterministic, because you the cause you'd identify would be leading, but still not conclusive and sufficient. Meaning while almost predictable, this event would nonetheless be random.

So actual determinism would also require certain lab conditions, those lab conditions require external conditions for the universe and so on, so that in order for any particular event to be deterministic the entire universe and everything in it would need to be deterministic, otherwise it would just be random and the determinism is just an approximation.

So if, as you state:

Random: An event E is random iff it is not deterministic.

Then, as Jo Wehler has mentioned in a comment, you'd include a thing and it's exhaustive opposite so "From Will follows Something or nothing" will be trivially true and also pretty useless as that tells you nothing about what "willed" means or what it particularly entails.

Also as others have already pointed out, if randomness is just indeterminism or non-determinism then this includes a lot more than total arbitrariness. "If under most earthy conditions in the last [your age years] you'd prefer ketchup over everything else", that would be a fairly good approximation to make an educated guess in terms of determining your choice of condiment when faced with a multitude of options. However ultimately it would still be random, because despite that you can't say with 100% certainty that just because ketchup is on the menu it will actually be picked here.

Now the irony is, you can still pick ketchup, you can also pick it because you like it and you can also pick it so often that people think they know you well enough to deduce from "[you] and [ketchup on the menu]" that the amount of ketchup will have shrinked after your meal, yet that's still not a sufficient cause, as libertarian free will still assumes you have a free choice and could decide on a whim to pick something else if you wanted.

Now you could go on as to whether that "want" that desire is ultimately free or whether that desire is determinism creeping in by developing preferences over time through the external influence of the environment, thus making you a product of the environment rather than a free autonomous agent or whether the environment rather restricts the amount of viable options but if, even in a situation where you have just one viable option, your will could still oscillate with respect to how you feel about that option and whether you accept it as your choice or reject it, even though you'd end up doing it either way.

Though with regards to the question, actually you made "random" so broad, that you can fairly easily stuff libertarian free will in that category which makes this not incoherent at least not with your argument.

Answer 6

Let's get more precise and break "random" down further into its 2 possible meanings: (for completeness' sake, if nothing else)

• Probabilistic: behaves according to some probability distribution on a fundamental level (many deterministic things follow probability distributions, while they're still non-probabilistic on a fundamental level).
• Arbitrary or spontaneous: is not probabilistic and is not determined by causes. It's unclear whether such things are even possible, but let's put that issue aside.

One would also need to define deterministic as being exactly and fully determined by its causes (but maybe this is functionally the same as having a "sufficient cause"). Otherwise this could include behaviour not determined by causes.

Some might also object to the label of "arbitrary or spontaneous", but either way, "deterministic + probabilistic + arbitrary" seems to be an exhaustive list of all possibilities, since "arbitrary" includes anything not included in the first 2 possibilities. Note: I address something being a combination of these things below.

---

From here, it does indeed seem that libertarian free will doesn't make much sense (although it isn't necessarily incoherent).

• Something that is non-deterministic has to be either probabilistic or arbitrary.
• Probabilistic and arbitrary things are necessarily free from any influence of our intents, desires, memories, emotions or anything else that makes up our personality.

Proponents of libertarian free will seems to be left with one of the following options:

• Accept that "free will" is indeed free from the things that make up our personality. This makes it something that seems to reduce or remove our agency, rather than providing it.

  If one says there's "a third thing", aside from deterministic and probabilistic (and which somehow isn't arbitrary), one would seemingly still have to bite the bullet on this point, because being fully determined by intents, desires, memories, etc. is necessarily deterministic.

• Assert that there is some part of one's personality that is both meaningful and fundamentally ill-defined.

  This position doesn't seem particularly tenable. It would suggest e.g. that you have a meaningful position on whether you like chocolate, yet whether you like chocolate doesn't have a defined value at any given point in time. It seems contradictory.

  Note that someone might like chocolate because they like the taste, and might dislike chocolate because it causes undesirable effects because they have lactose intolerance. This is still perfectly consistent with a deterministic view - note that "because" parts above, which provides causal reasoning for those feelings towards chocolate, and that would be well-defined even if those feelings might not boil down to a binary "like"/"dislike" overall feeling towards chocolate.

• Define our personality to exist "outside" of ... whatever, such that it's outside of causal consideration.

  If one already accepts the existence of some vague underdefined non-physical consciousness (as many people do), adding a vague "free will" to this isn't much of a stretch. But this doesn't solve the problem as much as it just draws an arbitrary line, puts free will on the other side, and only considers what's on this side.

One could also say that any given event breaks down into a combination of deterministic, probabilistic and arbitrary components. But this wouldn't solve the problem with libertarian free will. One could apply this argument for each component, and similarly conclude that the components which aren't deterministic are free from what makes up our personality, and thus they don't provide agency in any meaningful sense. Whereas the ones that are deterministic... are deterministic, which wouldn't do much for the claim of non-deterministic agency.

---

---

---

The similar failure for physicalism even has a name: Hempel's Dilemma.

Hempel's dilemma itself fails because it presupposes the coherence of the non-physical, as I detailed in another answer.

the three preconditions needed for libertarian free will. I consider them to be:

1. Our physical universe is probabilistic. ...

As argued above, probabilistic effects don't seem to provide any reasonable definition of agency. If you pick one thing above another 50% of the time, and there's no rhyme or reason behind that, then the choice is arbitrary. If there is a reason, if it's based on parts of one's personality, then it follows from that, instead of being fundamentally probabilistic.

2. Our physical universe is not causally closed to agents influencing it. Things like the near consensus that irreducibly emergent phenomena are causal, the placebo effect where our awareness is causal, the evolutionary tuning of consciousness, and every one of us experiencing our willing causing our actions every day, make this highly plausibly true of our world.

Questionable claim about consensus aside, this relies on a false dichotomy where consciousness is either "strongly emergent" (something which we have zero verified examples of) or non-physical. And it emphasises the "interaction problem" with both of those views, i.e. how would that interact with the physical world. This isn't an argument against the position that consciousness reduces to brain chemistry (similar to how every other complex system reduces down to its parts). If anything, the 2-way causation between consciousness and the physical world seems to suggest that consciousness exists alongside the things it influences and is influenced by, rather than existing in some other place entirely, in a place people can't explain, and existing in a way people can't explain, and people can't explain how that links back to the physical world to make sense of the aforementioned causal link.

3. Causal logic must include at least a third option besides "deterministically caused" and "probabilistic-or-random". This would be a third option, agent causation.

This is basically just saying that a precondition of libertarian free will is that the coherence of libertarian free will must be shown. Like... yeah, I agree. But the question is whether libertarian free will is coherent. And it seems kinda trivial to say that something must be coherent for us to say it exists. Proponents of libertarian free will presenting that as a meaningful point indicates how little they've got in support of their position, to not even be able to present a positive case for the coherence of their position.

This philosophical notion doesn't depend on knowing the element or mechanism that makes free choice possible. ... The confusion in this question arises because the question presumes a deterministic worldview in which all events must have determinate causal explanations, and so it interprets the concept of 'will' (oddly) as randomness.

The criticism is precisely the lack of any mechanism that could make that make sense, and the apparent contradiction in saying that we have meaningful agency, while implicitly rejecting the notion that said agency is in any way affected by our personality.

Libertarian free will relies on vagueness and handwaving. That could indeed said to be the strength of poorly justified claims (in that it makes it hard or impossible to argue against them). But it's also their weakness, in that you can't justify them, which makes it a good reason to reject such claims.

Answer 7

Libertarian free will is not incoherent. Only your strawman of libertarian free will is.

Your fallacy is wrong dichotomy. Deterministic vs. random is a wrong dichotomy for multiple reasons:

• There are no deterministic events in reality. Determinism is not an option.
• The opposite of deterministic is indeterministic, not random.
• The opposite of random is deliberate.

The real dichotomy is deliberate vs. random. Libertarian free will is the ability to make deliberate choices in the absence of determinism.

Answer 8

You ask if libertarian free will is incoherent. The problem is that you strawman the position of libertarian free will.

You state that something willed is either willed in a deterministic way or by random chance. That is simply not the libertarian position. To then prove that your strawman is incoherent doesn't prove anything.

Libertarianism states that determinism is incompatible with free will and since agents have free will, determinism must be false. This is perfectly coherent in itself. Then of course, as a libertarian, you have to explain how agents exercise free will and that's where the trouble for libertarians starts.

I don't think we have any free will whatsoever.

Answer 9

Pursuant to your earlier query regarding modal realism, I should like to bring up, firstly, that the theory of negation at its intersection with modal logic is very involved. One problem I've been having myself concerns the following:

1. If A is not possible, then A is not actual, either. Or: ~◊A > ~A
2. If it's not possible that it's not possible that A, then what? ~◊~◊A seems like it could be separably interpretable as (~◊)(~◊)A "at the same time" so that we reduce to ~~A = A (waiving the option of paracompleteness), or first as ~◊(~◊)A: ~◊~A = Necessarily A, or (~◊)~◊A = ~~◊A = ◊A. It is wildly unclear to me how to apply (1) correctly; and my intuition is no help, since all three options seem intuitively plausible and subtly absurd.

As far as disjunction goes, there is a theory of imagination and the concept of impossible worlds where there seem to be (meta-)possible nonprime disjunctions, i.e. cases where, "A or B," is true even if neither A specifically nor B is true "by itself, already" (see Berto[17]). Per the theory of demi-negation, maybe we'd say that A and B are demitrue by occurring in a true nonprime disjunction. At any rate, this general approach to the consistency of hard free will alternates between:

• Paraconsistent routes (modifications of the theories of negation and conjunction)
• Paracomplete routes (modifications of the theories of negation and disjunction)

So a drastic option would be to suppose that hard free will is a physical dialetheia, a case where a certain contradictory conjunction is true. Not, "Free will does and doesn't exist," but more like, "Everything is random and not everything is random" ("symmetrically," "Everything is determined and not everything is determined"). Coherence, as more than mere consistency, perhaps as nontriviality (one definition I've seen in this context), is then obtained if the paraconsistent treatment of the theory is stable enough (for the time being). Generally, the line of thought could be, "I am free to agree or disagree with the law of noncontradiction, so my intellectual will occupies a space that transcends that law."

That would be similar (or "dual") to Kant's theory about the matter. On the one hand, he says that a person's actions at any given time can be fully explained by empirical determination. However, it's impossible to comprehend the full series of causality "at infinity" (in reverse), so there is something indeterminate about the series of causality as-a-whole.^A (Alternatively put: Kant denies absolute causal/temporal closure.)

More abstractly, the concept of determination itself can be construed as not-determinate-by-its-own-functionality, so if it's determinate "at all," it's relative to a higher-order/level version of the concept, and the sequences of type levels itself "transcends" the determinate/indeterminate distinction. Or as the SEP entry on determinates and determinables puts it:

For the seemingly distinctive character of determinables and determination may enter into the best case for the claim that there are genuine features of reality that are less than maximally specific—a claim which, if true, has profound implications for a wide range of philosophical issues.

Now if the conception of determination is (subtly, but very) inconsistent, then of course a conception of hard free will, based in such part on the other conception, will be inconsistent: it will inherit the contradiction generated beforehand. Then it's open to the no-hard-free-will theorist to be paraconsistent about their own base concept, yet this would undercut their further appeals to mere internal consistency in judging the alternative base concept. (For rather than start from the concept of rigid temporal causality, one might start from one's discursive representation of hard free will, and define rigid outside causality in terms of a negation of hard free will.) A naive next stage in the dialectic might include, "We need as few contradictions as we can manage," something like a paraconsistent Occam's razor, and then we might say, "So postulating contradictory hard free will on top of contradictory determination is one contradiction too many," or we wouldn't think to talk in such a naive way, etc.

---

^AIf everything is finitely determined, then it is finitely determined, how many things are finitely determined. Finite determination is finitely determined itself. Then, after a finite number of determinations, finite determination will "run out," and everything after is infinitely determined. So then not everything is finitely determined, and we have a conditional of the form (A > ~A), which allows us to think through (A > ~A) > ~A, which is not exactly contradictory, but is contra-connexive (if my understanding is correct). So we might say that, besides paraconsistent and paracomplete, we also have an option of paraconditional glosses of hard-free-will theories (perhaps para-this or para-that glosses for whichever isolated connective/operator).

Answer 10

The core issue is that when talking about decisions and responsibility of admissible agents (e.g. humans), agents are considered indivisible (without a formal delineation of a single agent). As such any formal reasoning just breaks down in the extremes:

• When a person kills another while sleepwalking, who is responsible?
• At what age is a human responsible and can go to jail?
• Can a person with identity dissociation (aka multiple personalities) be responsible and non-reapinsible at the same time?
• Can humanity as a whole be held responsible for climate change? How does that responsibility trickle down to people living in harmony with nature, or babies?
• What non-humans (animals, computers, plants) can carry responsibility?

The flavor of libertarian free will that entrenches itself on the matter of responsibility posits an indivisible self that makes the decision, magically. Like an eternal soul.

At the lastest since the introductions of computer, philosophy however had to deal with the possibility that all humans as agents have a mind that is divisible and works in separate independent processes. And brain research found out different parts of the brain apparently being influential in different parts of that process.

And that causes the coherence issue, there is a concept of decisions and responsibility that works only with agents as indivisible entities, and there is an apparent reality of agents being divisible "machines".

Compatibilism does not conflict with that, it allows for certain machines to be philosophically valid bearers of responsibility (though maybe not legally with current laws, laws still need to catch up with technology).

But libertarians believe that if agents are divisible in any way (even say if in a magic fairy dimension of souls and rainbows where all real decisions are made, souls were divisible into smaller parts), responsibility as a concept falls part, and society will break apart, and there will be rape and murder and pillaging since nothing matters anymore. That's why it's important for libertarian free will that something exists that is indivisible and makes decisions and bears responsibility.

So the issue of libertarian free will is not just that a decision must be made independently of prior states of the universe, but also that the agent of decisions is indivisible, in contrast to how computers or theoretical neural networks work.

That does not make libertarian free will incoherent, it is rather a form of wishful thinking that is not theoretically impossible, but at odds with all we know already about brains, minds and computers.