I don't see where in the article he concedes that's impossible. — flannel jesus

You did.

Ok, so now that you know it's a thought experiment, and not a real experiment, and nobody thinks it's a real experiment and nobody is suggesting we conduct it in physical reality, — flannel jesus

that wasn't very clear from your phrasing that you were talking about the experiment being impossible

"So under libertarianism, the decision to do one action over the other ends up being arbitrary after all." But only under a condition defined to make it arbitrary, and that moreover is conceded to be impossible

I thought you meant the decision to do one action or the other is impossible.

Libertarians hold that free will is only possible in indeterminism. And given that indeterminism in regards to human choice literally means "Bob's choices are indeterministic if and only if, for some choices Bob makes, Bob2 might make a different choice from Bob1 even though everything is perfectly the same about them and their circumstances", then Libertarians by extention must necessarily believe that "Bob has free will if and only if, for some choices Bob makes, Bob2 might make a different choice from Bob1 even though everything is perfectly the same about them and their circumstances"

So you say the thought experiment is defined to make it arbitrary, but the thought experiment is quite literally *the only thing that distinguishes deterministic choice from libertarian choice*. So as far as I'm concerned, if one wants to take libertarianism serious, one MUST engage in a thought experiment like that. That's what libertarianism means. It means that the exact conditions Bob is in do not gurantee his choice, ie that Bob really, geniunely could do something entirely different even if literally nothing changed.

So the writer of the article isn't just defining some silly impossible thought experiment for shits and giggles, he's taking the central claim of libertarianism seriously and looking through the implications of it.

You say that, but then you confirm that Bob2 would always do the same thing as Bob1, which is what determinism means. — flannel jesus

I’ll adopt the SEP’s definition of causal determinism for clarity: “Causal determinism is, roughly speaking, the idea that every event is necessitated by antecedent events and conditions together with the laws of nature” (SEP entry on Causal Determinism). You’re right that, under this definition, if Bob2 always acts the same as Bob1 in a replay with identical conditions, determinism seems to hold. But my point is that physical determinism—the deterministic laws governing physical processes like brain states—doesn’t explain why Bob1 or Bob2 are acting in the way that they are. It doesn't provide the right causal story and so it leaves something important out of the picture.

Consider the explanation (or lack thereof) of a chance event as the encounter of two independent causal chains, which I have found to be variously ascribed to Aristotle (the general idea appears in Physics, Book II, Chapters 4-6 and in Metaphysics, Book V, Chapter 30, 1025a30-b5) or to J. S. Mill. This also has been illustrated by commentators with the scenario of two friends meeting by accident at a water well. If each friend went there for independent reasons, their meeting is accidental—no single physical causal chain explains it. But if they planned to meet, their encounter is intentional, explained by their shared purpose, not just particle movements. Intentional actions, like Bob’s, are similar: their physical realization (P2) isn’t an accident of prior physical states (P1). Reducing Bob’s action to the causal histories of his body’s particles misses why P2 constitutes a specific action, M2 (e.g., choosing to meet a friend), rather than random motion.

The determinist might argue, via van Inwagen’s Consequence Argument or Kim’s Causal Exclusion Argument, that since P1 deterministically causes P2, and M2 supervenes on P2 (M2 couldn’t differ without P2 differing), M2 also is determined by P1. But this overlooks two issues. First, multiple realizability: M2 could be realized by various physical states, not just P2. Second, contrastive causality: explaining M2 isn’t just about why P2 occurred, but why P2 realizes M2 specifically, rather than another action. The physical story from P1 to P2 ensures some state occurs, but not why it’s M2 specifically that P2 realizes non-accidentally.

What explains this is the rational connection between M1 (Bob’s prior beliefs and motivations, realized by P1) and M2 (his action, realized by P2). Bob’s reasoning—e.g., “I should go to the water well because my friend is waiting for me there”—links M1 to M2, ensuring P2 aligns with M2 intentionally, not by chance. This connection isn’t deterministic because rational deliberation follows normative principles, not strict laws. Unlike physical causation, where outcomes are fixed, reasoning allows multiple possible actions (e.g., meeting his friend or not) based on how Bob weighs his reasons. So, while physical determinism holds, it doesn’t capture the high-level causal structure of agency and the underdetermination of intentional action (i.e. what makes P2 a realization specifically of M2, when it is) by past physical states like P1. What further settles not only that P2 occurs, but also that P2 is a realization of M2 specifically, comes from the agent’s deliberation, which follows non-deterministic norms of reasoning, not merely physical laws.

But my point is that physical determinism—the deterministic laws governing physical processes like brain states—doesn’t explain why Bob1 or Bob2 are acting in the way that they are. It doesn't provide the right causal story and so it leaves something important out of the picture. — Pierre-Normand

Which is why I can't stress enough that "physical" isn't particularly important here. Afaik you brought up physical determinism. It isn't mentioned in the article and I didn't bring it up. Sure, maybe there are non physical things that go into deciding future states, future choices, future actions. That's not in question. Nobody is denying that. Whether you believe all that explains how the world evolves is physical or not doesn't seem to be to have anything to do with the argument at hand.

Are D and non-D exhaustive of all possibilities? It seems they might be, but it remains to be established. You yourself have said that non-D is like D only a little squishy, loosey-goosey, random. Free will, still undefined, is held to be possible only in a D world operating in a loosey-goosey way under some random influences.

We then have B1 and B2, held to be identical, or at least in identical circumstances. But can they be? Assuming both are alive, they're both dynamical systems, and with respect to their thoughts and feelings, in a state of constant flux and change. And for that matter so are their respective worlds. In particular, both Bs, having a choice to make, in consideration of their options, can prioritize them and think about them. They can distinguish and separate their "free" choices from their less free, and their choices which allow for no freedom. And of their free choices, they get to make a free choice. Thus they're living in a world that allows them to make free choices - not from the world itself, but from the way they are in it.

Whether there is any freedom in the world itself as it is in itself, I suspect there is not, except that beings in the world will never know enough to overcome what may seem equivalent to "free" occurrences. That is, de facto freedom.

So I may have to pay my bills and drive on the right side off the road - except that I am free to choose not to - but whether I have chocolate or strawberry ice cream today would appear to be a completely free choice.

And nothing in this about D or non-D worlds, or about libertarians. And if non-D is not defined, it must be taken as simply the negation of a D world. And if free will is ruled out of a D world, then it must then exist in any non-D world. And if non-D is simply a loose D, then it is a species of D, that is, a D, and therefore needs further definition to distinguish it from worlds that are not, then, either D or non-D.

Are D and non-D exhaustive of all possibilities? — tim wood

As far as functions go, or systems that evolve into the future, I believe so.

Free will, still undefined, is held to be possible only in a D world operating in a loosey-goosey way under some random influences. — tim wood

By whom?

Which is why I can't stress enough that "physical" isn't particularly important here. Afaik you brought up physical determinism. It isn't mentioned in the article and I didn't bring it up. Sure, maybe there are non physical things that go into deciding future states, future choices, future actions. That's not in question. Nobody is denying that. Whether you believe all that explains how the world evolves is physical or not doesn't seem to be to have anything to do with the argument at hand. — flannel jesus

It is directly relevant to the OP's argument since it directly refutes the claim that libertarianism inescapably faces an "Intelligibility problem" (Kane). Libertarianism need not face such a problem and only appears to do so because libertarians (and compatibilists too) have traditionally analysed alternate possibilities for action in terms of rollback scenarios that focus exclusively on the physical level of description of mental processes, while they've ignored important implications of multiple realizability and of the contrastive character of causation. The argument laid out in the OP's linked article also ignores those and as a result covertly sneaks in unsupported reductionistic assumptions.

I think you're reading stuff into it that isn't there. It doesn't say any of that explicitly, and I don't believe it says it implicitly either

By whom? — flannel jesus

This gets tiresome. You tell me. And I asked this above:

Let's try these:
Is free will possible in a D system?
Is free will possible in a non-D system? — tim wood

This gets tiresome. You tell me. — tim wood

You criticized both me and the writer of the article for being vague, but you keep on saying completely unqualified statements that, without more specifics, are hard to judge. I don't hold free will to be that. What do you mean "are held to be"? They certainly aren't universally held to be what you said, so by whom?

I think you're reading stuff into it that isn't there. It doesn't say any of that explicitly, and I don't believe it says it implicitly either — flannel jesus

I've read it carefully. It states: "According to libertarianism, if the clock were rolled back, then radically different things could happen than what happened the first time. This is because humans could choose differently the next time around even though all antecedent conditions including beliefs and desires remained the same."

This is not entailed by my specific brand of libertarianism (although it is entailed by most other libertarian accounts, possibly with the exception of Kane's). I do acknowledge that if the clocks were rolled back, then the exact same things would necessarily happen (which shows that P2 is causally necessitated by P1 and, contrary to popular belief, fails to show that M2 is causally necessitated by P1). Therefore the premise needed for an Intelligibility problem to arise doesn't hold.

Twice and more than twice I have asked you direct questions which you have ignored. No good reason can be given for those evasions, other that you're playing games and not interested in your own topic. The floor is all yours.

you're asking me questions that don't matter to understand the thought experiment in question. They simply don't matter. They're as relevant as if you asked my favourite sandwich

right, and the reason, as far as I can tell, that your view doesn't have the intelligibility problem is because it actually isn't reliant on indeterminism at all. If you think it's perfectly fine to roll back the clock and see the exact same events play out every time, then you think it's perfectly fine to be in a deterministic universe. You have some unique semantic reason for calling your view "indeterminism" - it looks like that semantic reason revolves around the word "physical" for whatever reason - but I don't have any semantic reason to do the same, because I don't care about "physical". A system that evolves from the past to the future is either deterministic or not, and whether that system is "physical" or not is irrelevant. The system you described, where you roll back the clocks and everything happens the same, is the very definition of a deterministic system. That's what it means to be deterministic. Being deterministic has nothing to do with "physical", and everything to do with what you said would happen if you roll back the clocks.

So of course your view doesn't have the intelligibility problem. As far as the argument in OP is concerned, you don't have a libertarian view of free will.

For what it's worth, although I haven't read the article, I believe I get the argument, which seems very simple. If it were somehow possible to repeat a situation in which I made a certain choice and everything in that situation was exactly the same in every possible way, libertarian free will would entail that I could, this second time around, make a different choice.

If that is impossible, then libertarian free will cannot be the case, by definition. It follows that that free will means simply 'acting according to one's nature'. Since we don't create ourselves, this seems the most sensible notion of free will: that is a compatibilist notion.

The system you described, where you roll back the clocks and everything happens the same, is the very definition of a deterministic system. — flannel jesus

This is not the definition of determinism. It is not the definition offered in the beginning of the SEP entry on Causal Determinism that I quoted earlier, for instance. Likewise, the definition offered on the Encyclopedia Britannica website is: "determinism, in philosophy and science, the thesis that all events in the universe, including human decisions and actions, are causally inevitable."

The specific characterization that you offer, and that is indeed very popular in philosophical discussions about the alleged conflict between free will (or responsibility) and "determinism" specifically focuses on the rollback scenarios that you (and the article linked on your OP) have alluded to. But the ideas of causal inevitability of actions, or of their "[necessitation] by antecedent events and conditions together with the laws of nature", are usually justified by questionable inferences (from the causal closure of the physical and supervenience theses) rather than logical entailments from general definitions of determinism such as those from the SEP or Encyclopedia Britannica. In fact, the focus on physicalism as the ground metaphysical commitment is betrayed by the idea of rolling back the situation to a past instantaneous moment in time, where many libertarian philosophers posit bifurcations to occur, which makes sense for indeterministic physical systems but doesn't make sense when what are at issue are temporally protracted deliberative episodes and unfolding actions.

(The main problem that most libertarian and compatibilist accounts possibly share, on my view, is that their physicalist commitments blind them to the naturalistic structure of downward causation, which is not a species of Humean event-event causation, and that make the very idea of rollback scenarios irrelevant.)

If it were somehow possible to repeat a situation in which I made a certain choice and everything in that situation was exactly the same in every possible way, libertarian free will would entail that I could, this second time around, make a different choice. — Janus

Yeah, just wanted to add to this to make it explicitly, you would have to be perfectly the same too. But I think you already factored that in..

that is impossible, then libertarian free will cannot be the case, by definition. — Janus

It's actually not so much about it being impossible, but rather that it doesn't seem to give us free will in any meaningful sense if it is possible

It follows that that free will means simply 'acting according to one's nature'. Since we don't create ourselves, this seems the most sensible notion of free will: that is a compatibilist notion. — Janus

This does end up being my eventual conclusion, but isn't my immediate conclusion from the argument at hand. The immediate conclusion is just that incompatibilist notions of free will don't land for me.

But the ideas of causal inevitability of actions, or of their "[necessitation] by antecedent events and conditions together with the laws of nature", are usually justified by questionable inferences — Pierre-Normand

I have no idea why you think that doesn't apply to your view that it would always play out the same way. Whatever semantic distinction you're drawing between your view, that it would always play out the same way given the same conditions, and the statement "it's causally inevitable that they would play out the same way"... to me they're just the same thing. I don't see a difference.

Either everything would always play out the same way or it wouldn't. If it would, to me that's determinism. As far as I'm concerned that's all determinism means. Same input, same output.

follow up to my previous post, I wanted to ask an LLM just for fun. An LLM is of course not a replacement for a human expert or an individuals thought, but when it comes to how words are used i think they can at least have interesting insights. Perhaps this is interesting, perhaps not:

https://chatgpt.com/share/67d246ea-488c-8002-9544-35fd937426e2

... follow up to my previous post, I wanted to ask an LLM just for fun. An LLM is of course not a replacement for a human expert or an individuals thought, but when it comes to how words are used i think they can at least have interesting insights. Perhaps this is interesting, perhaps not: — flannel jesus

I asked follow-up questions to the LLM you used, o1-mini (which is an excellent model) but it only generated thinking episodes and no main answer. This likely is a temporary issue with the OpenAI servers. So I asked your question again to GPT 4.5, and then supplied my follow-up questions/suggestions. I don't agree 100% with GPT 4.5's characterization of the issue, but it got the gist of it. (Its mentions of McDowell, Sellars and Brandom likely is a result of GPT 4.5 having accessed some if its stored memories about our past interactions, which is a bit weird since I had recently turned the memory feature off in the setting, but maybe this merely prevents the recording of new memories.)

you would have to be perfectly the same too. But I think you already factored that in.. — flannel jesus

Yes, that was what I had in mind: all conditions being exactly the same including oneself.

It's actually not so much about it being impossible, but rather that it doesn't seem to give us free will in any meaningful sense if it is possible — flannel jesus

What I meant with that is if in that 'rewind' scenario we could not make a different choice, then libertarian free will would seem to be ruled out by definition, since any consistent notion of libertarian free will requires that we could have made a different choice. The idea being that our will would not be determined by conditions but would be in some unfathomable sense causa sui.and sui generis.

But, you know, in simple terms, if everything is happening the same way every time you replay with the same starting conditions, that is in a very straight forward way what it means for something to be deterministic, right?

ChatGPT said:

Yeah, if every time you reset the system to the same initial conditions, it evolves in exactly the same way, then in the most straightforward sense, that’s what we mean by a deterministic system.

The extra philosophical layers we discussed—about high-level properties, rational norms, and interpretation—introduce complications about how we describe and explain what’s happening, but they don’t change the basic fact that, at the fundamental level, the system itself is unfolding in a fully determined way.

Would you say, then, that the distinction you're drawing is less about whether the system is deterministic and more about whether determinism at the physical level is the right kind of explanation for high-level rational agency?

Determinism doesn't have a whole lot of criteria. It really doesn't matter to me what you wrap your concepts in, no matter how complicated you want to make it. Supervenience, norms, multiple realizability - all very worthwhile, I'm sure, for you to think about, but it doesn't change the simple criteria of determinism - if you are guaranteed the same output from the same input, then it's in the most plain straight forward way deterministic.

(I didn't get a share link when I asked before, so I just asked again. Slightly different response the second time around. https://chatgpt.com/share/67d28f66-d8b4-8002-893b-ebe19bc39430 )

The point I was trying to make is that it's not outlandish or an unfair manipulation of language when I say, the criteria of determinism is simple, it's simply, if you give it the same input you get the same output. If you start in the same state the next events happen the same too.

It doesn't matter what you want to tack on top. That criteria isn't concerned with any extra layers of analysis you'd like to do. You can do them, you can not do them, the simple criteria of determinism doesn't mind either way. If a system meets the criteria, then it's deterministic.

So if you believe free will exists in that system you described - that system where, given the same starting state you get the same events after - then as far as the logic in the op is concerned, you're not a libertarian, and it's not arguing against your idea of free will. You semantically think it is, because you call your idea "libertarian free will", and the article uses that same term too. But I personally wouldn't call your concept of free will presented here "libertarian". I would call it compatibilist. And you're right when you say it's not vulnerable to the intelligibility argument.

What I meant with that is if in that 'rewind' scenario we could not make a different choice, then libertarian free will would seem to be ruled out by definition, since any consistent notion of libertarian free will requires that we could have made a different choice — Janus

You are correct, of course, I'm just saying that's not the direction of argument being made here.