There is Absolutely No Free Will
There is a simple argument against free will, and it is uncomfortable mostly because it is simple. Every event has a cause, or it does not. If it has a cause, you did not author it. If it does not, you did not author it either. There is no third option in which you appear as an uncaused cause, and nobody has ever been able to say what that third option would physically be. Everything else in this post is commentary on that paragraph.
The title is deliberately absolute, so let me say exactly what is being denied. I am not denying that you deliberate, that you choose, that your choices matter, or that you are a coherent agent. I am denying one specific thing:
That at the moment of choosing, you could have done otherwise, with the same past, the same laws, and the same brain, and that you are the ultimate source of which way it went.
That is contra-causal free will. It is what most people mean when they are not being careful, and it is what I think does not exist.
The two horns
Take any decision, say reaching for the left cup instead of the right one. Either the outcome was fixed by prior conditions, meaning the state of your brain, which was fixed by your history, your genes, your environment and the laws of physics,
$$ \text{state}_t \xrightarrow{\ \text{laws}\ } \text{state}_{t+1} $$or it was not fixed, in which case something in the process was random. Determinism gives you a chain you did not choose, and randomness gives you a coin flip you did not choose. A random event is not a free one: if a decision came from a quantum fluctuation in a neuron, it is not more yours than one that came from the neuron’s prior state. It is less yours, and it is not even a reason.
This is the structure of van Inwagen’s consequence argument1 and of Galen Strawson’s Basic Argument2. Strawson’s version is the cleanest. To be truly responsible for what you do, you would have to be responsible for the way you are. To be responsible for the way you are, you would have to have chosen it, and to have chosen it you would have needed to be some prior way that made you choose it. That regress does not terminate. You did not pick your genes, your parents, your language, your early childhood, or the order in which you met the ideas that now feel like your own, and there is no point in the chain at which a self steps in from outside it.
Turtles all the way down
Robert Sapolsky has spent a career on the biology of behavior, and his argument in Behave (2017)3 and then Determined (2023)4 is essentially this regress filled in with neuroscience. Why did you do that? Because of the neurons firing a second earlier. Those fired because of the sensory input and hormones of the minutes before, which depended on your sleep, your hunger and your stress that day, which depended on your childhood, which depended on your fetal environment and your genes, which depended on your ancestors, the evolutionary pressures on them, and the culture that shaped all of it. At no timescale does a free agent show up.
The part I find most useful in Determined is not the chain itself but that Sapolsky goes after the standard escape routes one by one: chaos, emergence and quantum indeterminism4. Chaotic systems are unpredictable in practice, not undetermined in principle. Emergent levels of description are real, but they do not escape the causal closure of the levels beneath them. And indeterminism, as above, buys randomness rather than authorship. He also uses cases like Charles Whitman, who in 1966 killed more than a dozen people from the University of Texas tower and whose autopsy revealed a brain tumor pressing on the amygdala4. Most of us feel that this changes how we should think about his responsibility, and Sapolsky’s point is that it should change it without a threshold. The tumor is simply the one cause we happened to find, and we usually do not find the others.
The view from the inside
Sam Harris makes a different and, I think, complementary argument in Free Will (2012), and it is an introspective one5. Watch your next thought arise. You did not choose it, and you could not have, because you would have had to think it first. Your intentions, your reasons and the sudden preference for one cup over another all appear in consciousness, and consciousness is then retrospectively told a story in which it authored them. The feeling of authorship is real, but the authorship is not.
Harris also stresses something that tends to be underweighted: the argument is not that you are being forced against your will, but that the will itself is one more thing you find yourself with. You can be free from coercion without being the author of the desires you are free to act on. This matters for what follows, because it means the sense of freedom is evidence about the brain’s model of itself, not evidence about the causal structure of the world.
What neuroscience adds, and what it does not
The experiments are usually presented as the heart of the case, and I think they are not, so I want to be careful here. Libet and colleagues (1983)6 found that a slow electrical build-up in the brain, the readiness potential, began roughly half a second before a voluntary movement and well before subjects reported being aware of the urge to move. Soon, Brass, Heinze and Haynes (2008)7 decoded a left/right choice from prefrontal and parietal activity several seconds before subjects reported deciding, though only modestly above chance. Fried, Mukamel and Kreiman (2011)8 did something similar with single neurons in human medial frontal cortex, with much higher accuracy at much shorter lead times.
These results are suggestive, but they are also contested. Schurger, Sitt and Dehaene (2012)9 argued that the readiness potential may simply reflect ongoing random fluctuations crossing a threshold, rather than a decision already made unconsciously, so reading the Libet result as proof of anything is too strong. That is fine, because the argument above does not need it. What the experiments establish, together with Daniel Wegner’s work in The Illusion of Conscious Will (2002)10, is something narrower and still important:
$$ \begin{gathered} \text{the feeling of consciously initiating an act is a report} \\ \text{produced by the brain, and it can be wrong.} \end{gathered} $$That is enough. The philosophy carries the conclusion, and the neuroscience only removes the intuition that we would know it if we were not free.
Physics does not rescue it
Laplace stated the classical picture in 181411: an intellect that knew all forces and positions at one time could compute the entire future. Classical mechanics is deterministic, and nothing in it has a place for an uncaused chooser. Quantum mechanics is the usual hope, because it appears to be indeterministic, but notice what we get if we take that at face value. Randomness at the level of ion channels is noise, and noise does not make a decision more yours. Whether quantum mechanics is fundamentally random is also interpretation-dependent, since the Schrödinger evolution is deterministic and interpretations such as Everett’s12 and Bohm’s13 keep it that way. Either way the structure of the first section stands. Einstein did not like dice, but the deeper point is that even a universe that does play dice does not thereby produce a self that rolls them.
Our own work: why it feels free
So there is a puzzle. If the argument is this short, why does the feeling of freedom survive it? I think the model we developed in Cognition as Non-Commutative Geometry14 offers a structural answer, and I want to state carefully what it does and does not claim, because it would be easy to abuse. In that post the proposal is that a sufficiently self-referential cognitive system should be modeled as a non-commutative algebra of operations \$\mathcal A\$ together with a state \$\omega\$ on it. The central step was that self-reference makes operations order-sensitive, because the system’s act of inspecting itself changes the state it is inspecting:
$$ AB \neq BA. $$What this does not do is introduce indeterminism. Non-commutativity is a statement about order, not about chance, and the dynamics generated by the pair \$(\mathcal A,\omega)\$, the modular flow
$$ \sigma_t^\omega(A) = \Delta_\omega^{it} A \Delta_\omega^{-it}, $$is a one-parameter automorphism group, which is as deterministic as anything gets. Given the algebra and the state, the flow is fixed. So the model gives no loophole for free will, and that is a feature rather than a failure.
What it does do is explain the shape of the illusion. A system that contains a model of itself, \$M(S)\subseteq S\$, has a particular problem: each time the model is applied, it changes the state it was meant to describe,
$$ S \xrightarrow{A} S', \qquad M(S) \text{ must now describe } S'. $$The self-model is always one step behind the system it models, so from the inside there is no vantage point from which the next state of the system is fully available before it happens. This is not Gödel’s theorem and I would not say it is, but it is structurally the same family of limits on self-description. The consequence is that the system cannot, from inside, predict its own next decision in the way an outside observer with the full state could. From the inside the future looks open: epistemically open and ontologically closed.
This is an old idea in a new vocabulary. Donald MacKay made the same point in 1960 in On the logical indeterminacy of a free choice15, arguing that a deterministic brain’s own state cannot be fully predicted by that brain in a way it is bound to accept, and the halting problem16 teaches the same lesson in computation: a perfectly deterministic program can be impossible to shortcut, so that the only way to know what it does is to run it. Running it, from the inside, is called deliberating. The feeling of free will is what unpredictability-from-within looks like from within.
There is a second consequence worth noting. In the NCG picture learning is not only a change of state, \$\omega\to\omega'\$, but also a change of algebra, \$\mathcal A\to\mathcal A'\$, meaning a change in which operations are available at all. A person who has learned to reason recursively, or to speak a second language, or to recognize a pattern of manipulation, has a different space of possible actions than before. This is the part of “freedom” that is real. It is not the ability to step outside the causal chain but the growth of the set of operations the chain can run through, a quantity that can be larger or smaller, can be nurtured or destroyed, and matters morally and politically a great deal.
All of this is conjecture, in the same sense as the original post. The identification of that mathematics with cognition is the hypothesis, and what I am claiming here is only the conditional: if cognition has that structure, the sense of free will is expected, and the fact of free will is not.
Spinoza had it
None of this is new, and the person who said it most cleanly had no neuroscience at all. In the Ethics (1677)17, Spinoza argued that people think themselves free because they are conscious of their actions and ignorant of the causes that determine them. In his letter to Schuller18 he gave the image of a stone set in motion that, if it could think, would believe it was moving of its own will. Spinoza did not conclude that nothing matters. He concluded that freedom needs to be redefined: freedom is not the absence of necessity but acting from your own nature rather than from external compulsion, and more than that, understanding the causes that move you. That is the only freedom I think exists, and it comes in degrees.
It is also the notion that matters when we ask what it would take for any mind, biological or artificial, to be free, and it is the thread connecting this post to the Visprax announcement19. A free mind is not an uncaused mind. It is a mind whose actions flow from its own structure, and that understands that structure.
What changes, and what does not
What does change is how we treat people who cause harm, and I think it changes almost everything. If nobody ultimately authors their character, then there is no threshold at which someone becomes deserving of suffering, because it is turtles all the way down for the person who steals and for the person who kills. The tumor in Whitman’s brain is just the cause we happened to find, and there are always others we haven’t. Seen this way, imprisonment as we practice it, with its aim of making someone pay for what they did, makes no sense to me, and I think future generations will look at it the way we look at the stocks and the dungeon: as something medieval, built on a picture of the person that biology has made untenable. The one thing that survives is the narrow case where someone is an active danger to others and has to be kept apart from them, which is what Sapolsky calls the quarantine model420. That is a very different thing from prison. We keep a child with a contagious illness at home without hatred, without a sentence, and with a clear expectation that they will rejoin everyone once they’re well, and the same logic would demand that any restraint be as limited, humane and temporary as possible, paired with real effort to understand and treat the causes. Almost everything else, the punishment, the stigma, the lifelong record, the revenge dressed up as justice, falls away once you take the argument seriously. We won’t stop feeling the pull toward retribution, since it is an old emotion that served social species well, but we can stop building institutions on it.
The strongest objections
A post with this title should say where it is most likely to be wrong. The first objection is compatibilism, the majority position among professional philosophers: in the 2020 PhilPapers survey21, roughly three in five accepted or leaned toward compatibilism, around a fifth toward libertarianism, and only about one in nine toward no free will. Dennett (Elbow Room, 198422; Freedom Evolves, 200323) and Frankfurt (1969)24 argue that what we care about in free will, acting on our own reasons, without coercion, and responsive to reasons, is fully compatible with determinism, and that demanding contra-causal freedom is demanding something incoherent and then complaining that it is missing. I think this is the most serious opposing view. My reply is that it changes the subject: the compatibilist freedom is real, it is close to Spinoza’s, and I have used it above, but it does not give what ordinary talk of desert and ultimate responsibility demands, and the honest thing is to say so rather than keep the word. Compatibilists would answer that “free will” in the everyday sense never meant the contra-causal thing and that I am the one redefining it, which is a real disagreement about what the words refer to and not only about the physics.
The second is libertarianism. Robert Kane (The Significance of Free Will, 1996)25 and agent-causal theorists such as Chisholm26 argue that indeterminism at the right moments in the right neural processes, or a distinct kind of causation exercised by agents, can supply genuine authorship. The standard worry is the one from the start of this post: it is unclear how an undetermined event can be more controlled, and agent causation seems to posit a causal power with no physical account. The third is the experiments, which, as noted, are contested, though nothing above depends on them. The last is our own model. The NCG framework is a conjecture, and the step that matters here, that self-reference yields a structural gap between self-model and system, would need to be tested and could fail. If it does, the feeling of freedom still needs some other explanation, but the argument of the first sections does not depend on it.
What would change my mind
It is worth saying what evidence would count. I would need an account of a physical event in a brain that is neither fixed by prior states nor random, and that is nonetheless attributable to a self in a way that makes the self responsible for it. Not a gesture toward one, but an account. I do not think this can be provided, because the two categories, caused and uncaused, exhaust the possibilities and neither one is authorship. But that is exactly the kind of claim a counterexample would refute, and I would be happy to be shown one.
The actual claim
Put compactly, caused or uncaused, neither gives ultimate authorship. The feeling of freedom is real: it is produced by a system that cannot fully see itself from within, and that therefore experiences its own next state as open. The usefulness of freedom is also real, in the form of the expanding set of operations a mind can perform and the understanding of the causes that move it. But the metaphysical freedom, a self standing outside the causal chain and steering it, does not exist. There is no free will, and there is still a you. It is just not the thing you thought was steering.
Scope and limits
This post argues against one specific thing, contra-causal free will, and it does not claim that deliberation, choice or agency are illusions. The title is intentionally stronger than the weakest version of the claim, but not stronger than the argument. I do not think we can prove that no physical process could ever underwrite authorship. I think no one can say what such a process would be, and that this is the same thing.
The neuroscience is not load-bearing. The Libet-style experiments are contested, and the argument in the first sections stands or falls on the logic of caused versus uncaused, not on any readiness potential. The proportions I give from the 2020 PhilPapers survey are approximate readings of its free will question, and since respondents could endorse more than one answer they should not be read as precise.
The section on our own work is a conjecture in the same sense as the original post. It supports a conditional only: if cognition has the structure proposed there, then the feeling of free will is expected without the fact of it. It does not establish that cognition has that structure. The paragraph on punishment is also a different kind of claim from the rest. It is a normative extension of the argument that I find compelling, and it does not follow from the metaphysics alone. Sapolsky’s quarantine model is paraphrased from his public statements, and people who accept the argument about free will can still disagree about deterrence, incapacitation and what institutions should look like in practice.
References
van Inwagen, Peter. An Essay on Free Will. Oxford University Press, 1983. The standard statement of the consequence argument for incompatibilism. ↩︎
Strawson, Galen. “The Impossibility of Moral Responsibility.” Philosophical Studies 75, no. 1–2 (1994): 5–24. https://doi.org/10.1007/BF00989879 ↩︎
Sapolsky, Robert M. Behave: The Biology of Humans at Our Best and Worst. Penguin Press, 2017. ↩︎
Sapolsky, Robert M. Determined: A Science of Life Without Free Will. Penguin Press, 2023. ↩︎ ↩︎ ↩︎ ↩︎
Harris, Sam. Free Will. Free Press, 2012. ↩︎
Libet, Benjamin, Curtis A. Gleason, Elwood W. Wright, and Dennis K. Pearl. “Time of Conscious Intention to Act in Relation to Onset of Cerebral Activity (Readiness-Potential): The Unconscious Initiation of a Freely Voluntary Act.” Brain 106, no. 3 (1983): 623–642. https://doi.org/10.1093/brain/106.3.623 ↩︎
Soon, Chun Siong, Marcel Brass, Hans-Jochen Heinze, and John-Dylan Haynes. “Unconscious Determinants of Free Decisions in the Human Brain.” Nature Neuroscience 11, no. 5 (2008): 543–545. https://doi.org/10.1038/nn.2112 ↩︎
Fried, Itzhak, Roy Mukamel, and Gabriel Kreiman. “Internally Generated Preactivation of Single Neurons in Human Medial Frontal Cortex Predicts Volition.” Neuron 69, no. 3 (2011): 548–562. https://doi.org/10.1016/j.neuron.2010.11.045 ↩︎
Schurger, Aaron, Jacobo D. Sitt, and Stanislas Dehaene. “An Accumulator Model for Spontaneous Neural Activity Prior to Self-Initiated Movement.” Proceedings of the National Academy of Sciences 109, no. 42 (2012): E2904–E2913. https://doi.org/10.1073/pnas.1210467109 ↩︎
Wegner, Daniel M. The Illusion of Conscious Will. MIT Press, 2002. ↩︎
Laplace, Pierre-Simon. Essai philosophique sur les probabilités. 1814. English translation: A Philosophical Essay on Probabilities. ↩︎
Everett, Hugh, III. “‘Relative State’ Formulation of Quantum Mechanics.” Reviews of Modern Physics 29, no. 3 (1957): 454–462. https://doi.org/10.1103/RevModPhys.29.454 ↩︎
Bohm, David. “A Suggested Interpretation of the Quantum Theory in Terms of ‘Hidden’ Variables. I.” Physical Review 85, no. 2 (1952): 166–179. https://doi.org/10.1103/PhysRev.85.166 ↩︎
Aliei, Saeid. “Cognition as Non-Commutative Geometry.” Saeid Aliei. The model discussed in this section. https://saeidaliei.github.io/posts/ncg-cognition/ ↩︎
MacKay, Donald M. “On the Logical Indeterminacy of a Free Choice.” Mind 69, no. 273 (1960): 31–40. https://doi.org/10.1093/mind/LXIX.273.31 ↩︎
Turing, Alan M. “On Computable Numbers, with an Application to the Entscheidungsproblem.” Proceedings of the London Mathematical Society s2-42, no. 1 (1936–37): 230–265. https://doi.org/10.1112/plms/s2-42.1.230 ↩︎
Spinoza, Baruch. Ethics. 1677. See especially Part I, Appendix, and Part III, Proposition 2, Scholium. ↩︎
Spinoza, Baruch. Letter 58 to G. H. Schuller, 1674, which contains the image of the stone. ↩︎
Aliei, Saeid. “The Seed and the Act: Announcing Visprax and UnsafeAI.” Saeid Aliei. https://saeidaliei.github.io/posts/visprax-unsafeai-announcement/ ↩︎
Sapolsky, Robert M. “Robert Sapolsky on Free Will: Science & Society.” Interview, Thought Economics. Discusses the quarantine model of dealing with harmful behavior. https://thoughteconomics.com/robert-sapolsky ↩︎
Bourget, David, and David J. Chalmers. “Philosophers on Philosophy: The 2020 PhilPapers Survey.” Philosophers’ Imprint 23, no. 11 (2023). https://philarchive.org/rec/BOUPOP-3 ↩︎
Dennett, Daniel C. Elbow Room: The Varieties of Free Will Worth Wanting. MIT Press, 1984. ↩︎
Dennett, Daniel C. Freedom Evolves. Viking, 2003. ↩︎
Frankfurt, Harry G. “Alternate Possibilities and Moral Responsibility.” The Journal of Philosophy 66, no. 23 (1969): 829–839. https://doi.org/10.2307/2023833 ↩︎
Kane, Robert. The Significance of Free Will. Oxford University Press, 1996. ↩︎
Chisholm, Roderick M. “Human Freedom and the Self.” The Lindley Lecture, University of Kansas, 1964. ↩︎