← All texts

Essay · mantisvex.com

You Are a Machine

Free will is not a power you exercise. It is a feeling your machine produces: the lossy rendering of a real fact, your unpredictability, misdescribed as the power to be a cause without a cause. This essay dismantles the power, spares the feeling, and explains why the machine manufactures it. The strong concept of free will is dead on both sides of the physical verdict: if everything is caused, you are not the origin of your acts; if some part is random, you are not the origin either. What follows spares no one, the author included: the one who wrote this text deserves no credit for having written it.

Essay 2026 ~39 min read
I

You are a machine (and I wish I were wrong)

You are a machine. That is neither an insult nor an image: it is a statement of inventory. A physical system, made of ordinary matter, each of whose states follows from the previous one according to laws that do not ask your opinion. And I want to put one thing down before anything else, because it changes how everything that follows should be read: I am not writing this with pleasure. Most texts of this kind are written by people who savor the demolition. Not here. I would rather conclude the opposite. I would rather be master of my destiny, author of my acts, origin of something. I searched for the exit with the energy of someone who wants to find it, and I did not find it. This text is the report of that search: a conclusion adopted against my own wish. An argument accepted reluctantly has been tested harder than the others; that is the only authority I claim.

First the foundation, because everything rests on it and it fits in one paragraph. Your thoughts are states of an organ. Every time the organ is touched, the mind is touched: a lesion changes the personality, a molecule changes the mood, an electrode triggers a gesture or an urge. There is not one documented case, not one, of the mind doing something its substrate was not doing. No extra ingredient has ever been needed to explain a human behavior: the machinery suffices, and it suffices every time. That the machine is made of carbon rather than silicon changes nothing about its status as a machine: the fabric matters less than the arrangement. I have argued elsewhere that felt experience itself, what it is like to be you, is a fact of organization and nothing more; this text does not depend on that, but it is cut from the same wood.

 one level down · the machine, exhibits for the record

Three families of facts, converging for over a century and a half. Lesions: since Phineas Gage (1848)1, whose character was transformed by an iron bar through the frontal lobe, we know that destroying a region of the brain destroys or deforms a function of the mind, up to and including personality itself. Chemistry: every psychoactive substance, from coffee to antidepressants, modifies cognition by modifying the substrate, and nothing else. Stimulation: ever since Penfield's mappings1, an electrode placed at the right spot on the cortex produces a movement, a sensation, and sometimes an urge to move that the patient claims as his own.

None of these observations proves by itself that the mind is the brain. Together they establish something more modest and sufficient for this essay: every intervention on the substrate reaches the mind, without known exception, and no mental phenomenon has ever required anything but the substrate to be explained. Dualism is not refuted in the logical sense; it is out of a job.

Now, the experiment. Do it with me, it takes two seconds. I am going to write a digit, whichever one I want, right now: 3. There. And I report to you faithfully what it was like from the inside: I felt a choice. I felt that 5 was available, that 7 was available, that nothing forced me, and that it was I who settled it. That feeling is real, I do not deny it for a second. I only ask the killer question: what does it prove? Answer: nothing. Strictly nothing. Because in a fully determined world, the machine that I am writes 3 and produces exactly this feeling, the feeling of having been able to write 5. And in a world where free will existed, it writes 3 and produces the same feeling. The inner experience is identical in both worlds. It therefore cannot tell them apart. Introspection, on this question, is an instrument whose resolving power is zero: it displays « choice » whatever world it runs in. The whole debate about free will begins with this humiliation: the only direct witness you possess is structurally incapable of testifying.

Then comes the temptation, and I know it in you because I have had it: proving one's freedom by an act. Writing the most absurd digit possible. Switching hands. Doing anything at all, on purpose, to show the causal chain you are not inside it. Watch the trap close: the impulse to make that gesture, where did it come from? It arrived in you. You did not fabricate it out of nothing: it surged, caused by this reading, by this challenge, by the state of your machine at this instant. An act undertaken to prove freedom is an act, and an act is a link. No gesture can step outside the chain to testify against it, since the gesture is the chain in the process of unrolling. This is the deepest lock in the whole case: the system has no maneuver, none, that escapes by construction the very thing it is supposed to test. You can rattle the bars as much as you like; rattling the bars is on the program.

This situation was described once and for all, in 1674, in a private letter2. Spinoza proposes a thought experiment of perfect cruelty: imagine a stone that a hand has thrown. It flies. Now give it consciousness. The stone, says Spinoza, would believe itself free: it would feel its motion as an impetus coming from itself, it would believe it wills to fly, chooses its trajectory, decides its landing point. And he drives the nail in: such is that human freedom which all men boast of possessing, and which consists in this alone, that men are conscious of their desires and ignorant of the causes that determine them. Three and a half centuries before the first electrode, everything is there. My digit 3 is the stone. The feeling of having been able to write 5 is the flight mistaking itself for a takeoff.

So here is the program, and the rules of the game. I will not claim to prove a nonexistence: one does not prove the nonexistence of ghosts either. I will do better, and more honest: show that free will in the strong sense, the power to be a cause without a cause, the ultimate author who could really have done otherwise under exactly the same conditions, is a concept that does not survive its own description; that it dies on both sides of the only physical verdict possible, caused or random; that not one observation, ever, has weighed in its favor; and that it is necessary to explain nothing. A concept in that state is not entitled to the benefit of the doubt: the burden of proof is on it, and there it will stay. But before the demolition, a duty of cleanliness: my own camp defends this conclusion with arguments that are not arguments, and I will begin by executing them myself. Then the demolition itself, in two blades, which will need no laboratory. Then the most serious adversary, compatibilism, which I will present at full strength before striking it, because a caricatured adversary does not count. Then the real question, the one in the title: if the power does not exist, why does your machine manufacture the feeling? And finally the price, paid in full, without anesthesia: what dies with the ultimate author, and what it is like to know it.

II

The bad arguments of my own camp

Before striking the adversary, you clean your own house: that is the rule here. And there is cleaning to do, because the determinist camp, mine, defends a correct conclusion with an arsenal of which half is bad ammunition. Whenever a popularizer wants to bury free will in three minutes, he draws the same experiment, the same graph, the same shock phrase: « science has shown that your brain decides before you do ». I will begin by executing that argument, because I have used it myself, because it is fragile, and because a true thesis defended by weak evidence ends up looking like a false one.

The experiment first, at full strength. In 1983, Benjamin Libet fitted electrodes3 to the skulls of volunteers and asked them for something trivial: flex the wrist whenever the urge came, while noting, on a fast clock, the precise instant they felt the intention to move. The famous result: a rise in cerebral activity, the readiness potential, begins about half a second before the gesture, while the conscious intention is reported only two hundred milliseconds before. The pop translation, immediate and worldwide: the brain has already decided while « you » still knew nothing. The decision precedes the decider. Case closed.

Except the case does not hold, and it took thirty years to see where. In 2012, Schurger and his colleagues proposed a rereading4 that changes everything: the readiness potential may not be a decision being executed, but noise. Neural activity fluctuates constantly, like a swell; in a task where one waits for an urge to « come », the gesture launches at the moment that swell happens to cross a threshold. And here is the artifact: when you average the recordings by aligning them on the gesture, you keep, by construction, only the waves that were rising. The beautiful regular ramp on Libet's graph is not the portrait of an early decision: it is what randomness looks like once sorted by the finish line. Schurger's model reproduces Libet's data without any early decision existing anywhere. My camp's star argument therefore rests on a reading of the data that is no longer the best available. I put it away, and I note the lesson: it will hold for everything else.

 one level down · the fine print of the neuroscience of choice

Two supplements for whoever wants the full file. First, even without Schurger, Libet's experiment measured little: a spontaneous wrist flexion, with no stakes, no deliberation, no reasons. Extrapolating from it to the choice of a career or a lie was already a leap nothing licensed. Libet himself, for that matter, did not conclude that free will does not exist: he defended a conscious veto power, the freedom not to do.

Then, Libet's heiress: the study by Soon and colleagues (2008)5, in brain imaging, often summarized as « your choice can be predicted up to ten seconds before you know it ». The fine print: the prediction concerns a binary choice (left button or right), and its accuracy tops out around sixty percent. Predicting barely better than a coin, several seconds ahead, shows that early neural biases exist which tilt the scale. That is real, that is interesting, and it is very far from « the decision was already made ». A slightly loaded die is not a destiny.

Second piece of defective ammunition, in the other direction this time: quantum randomness. The objection returns in every conversation: « but physics has not been deterministic for a century; quantum indeterminism reopens the door ». Two replies, one short and one fatal. The short one: whether fundamental physics is deterministic is genuinely open, some interpretations of quantum mechanics are, others are not, and I do not need to settle it. The fatal one: even if your choices contained a share of authentic randomness, in what sense would that randomness be you? A die rolling in your skull is not an author. If I write 3 because a random fluctuation tipped a neuron, I am even less the origin of that 3 than if it flowed from my character, my history, my machine. Randomness does not manufacture freedom: it manufactures lottery. Hold on to that point, it is half of the next section: strong free will needs to escape both the cause and the die, and there is no third box.

Third piece of ammunition, to be handled with gloves: the illusions of agency. The literature is real and fascinating. People sincerely claim acts they did not commit: the ouija board that « no one » is pushing, facilitated communication where the facilitator's hand writes in the patient's place, Wegner's experiments6 where subjects are brought to feel authorship of a movement produced by someone else. And the reverse exists too: so-called alien hands, which their owner watches act as if they belonged to a third party. Correct conclusion: the feeling of being the author of an act is not a direct perception of causality, it is an attribution the brain computes after the fact, from cues, and it can err in both directions. Abusive conclusion, which my camp draws too fast: « therefore this feeling is always an illusion ». No. Showing that an instrument can err does not show that it always errs; your vision produces optical illusions and remains globally reliable. What Wegner really establishes, and it is exactly what I will need further on, is more precise and more interesting: the feeling of agency is a manufactured signal. A rendering, produced by the machine, according to rules, with failures. Not a window onto metaphysics.

Balance of the cleaning, and it must be said bluntly: determinism is not a laboratory discovery of 1983, and it never needed electrodes. My camp loves neuroscience because a graph looks more serious than an argument, but that is a weakness dressed up as a strength: propping the thesis on a fragile experiment offers the adversary the pleasure of refuting the experiment and believing he has refuted the thesis. The real demolition of strong free will needs no scanner. It is structural, it fits in a few lines, its idea is older than electricity, and it kills the concept on both sides at once: it is the section that comes next.

III

The two-blade demolition

I promised a demolition without a scanner: here it is. Two arguments, two blades. The first received its definitive form in 1983, the same year as Libet's experiment, and the irony of the calendar is perfect: while a laboratory was producing the bad proof, a desk was producing the good one. The second fits in a regression anyone can unroll in their head, knowing nothing about anything. Neither rests on an electrode, a statistic or a discovery: they rest on what the words « your acts » mean. And taken together, they close the only two exits physics leaves open.

First blade: the consequence argument, in Peter van Inwagen's formulation7. Here it is, unadorned. If determinism is true, then your acts of today are the consequences of the laws of nature and of the state of the world in the distant past, say before your birth. Now, what happened before your birth does not depend on you: you were not there. The laws of nature do not depend on you either: nobody votes on them. And what follows necessarily from things that do not depend on you does not depend on you. Conclusion: your acts do not depend on you, in the sense that mattered; you have never been able to do otherwise than what you did. Three premises, one transfer principle, one conclusion. That is all, and it is enough. Forty years of literature have gone after it; the detailed balance sheet is in the block below, but the summary fits in one line: no one has produced a clean refutation, only ways of redefining « able to do otherwise » so that the conclusion stops hurting. Remember that maneuver: it is the entire subject of the next section.

And savor the irony, because it is worth the detour: van Inwagen believes in free will. He is a convinced libertarian. He forged the sharpest blade ever produced against the coexistence of free will and determinism, then held it by the other end: free will exists, he concluded, therefore determinism is false. Same argument, reversed grip. He sacrifices physics to save freedom; I pull the other way. But note what his wager requires in order to succeed: it is not enough that determinism be false. Indeterminism must, on top of that, be able to supply freedom. A world riddled with chance must make you an author. And that is exactly where the second blade comes down, and it leaves nothing.

Second blade: the basic argument, formulated by Galen Strawson8. You do what you do, in a given situation, because of what you are: your preferences, your character, your reasons, the state of your machine at that instant. To be ultimately responsible for what you do, you would therefore have to be ultimately responsible for what you are, at least in the respects that decide. But to be responsible for what you are, you would have to have made yourself: to have chosen, somewhere in the past, your own character. Yet that choice you would have made because of what you were at that moment. And to answer for that choice, you would have to have chosen that earlier state; a choice that would have flowed from a state earlier still; and so on, with no floor. The regression stops only at an impossible being: a being that would have created itself, out of nothing, before existing, in order to answer for what it is. Cause of itself. No one is cause of itself. Nothing, anywhere, is cause of itself.

Now measure what the second blade does that the first did not: it is totally indifferent to the physical verdict. Whether the chain is strict or riddled with chance, the regression holds. Insert a quantum die wherever you like in the history of your character: the throw does not thereby become yours, one does not sign a roll of dice. You remember the previous section: there is no third box between cause and chance. The first blade closes the box marked « caused ». The second closes both in a single motion. Strong free will, the power to be the ultimate origin of one's acts, therefore dies whatever world you run in: deterministic, you are not the origin; indeterministic, you are not either. And understand the nature of this death. It is not an observed absence, like that of unicorns, which an expedition could overturn tomorrow. It is a concept that does not survive its own description: every time one tries to write down in black and white what being cause of oneself would be, one writes a regression or a miracle. Centuries of philosophy have tried without producing a single coherent description. I am not betting on the next one.

 one level down · the state of the front, forty years on

On the first blade's side: van Inwagen gave the argument a precise logical form, resting on a transfer rule (known as rule beta): if no one can do anything about p, and p entails q, then no one can do anything about q. Technical counterexamples were proposed against the original version of the rule (McKay and Johnson, 1996)9; refined versions were built that withstand them. The real front never stabilized there. It settled on the meaning of « able to do otherwise »: compatibilists read it as « I would have done otherwise if I had wanted otherwise », a conditional reading under which the argument no longer bites. In other words: the blade is acknowledged as valid against the strong concept, and the dominant defense consists in changing the concept.

On the second blade's side: the standard reply does not attack the regression, which is watertight. It attacks its starting point: « ultimate responsibility was never required; ordinary responsibility, that of courts and parents, suffices and survives ». Same maneuver. Note the pattern, for it is striking: faced with the two blades, the majority defense does not say « the argument is false ». It says « the argument kills something we do not need ». That may be true. It is precisely what must be judged head-on, not sideways.

So here is the state of play, and it has something strange about it that must be faced before moving on. The strong concept is dead twice over, cleanly, publicly, and has been for a long time. And yet the majority of professional philosophers do not join my camp. Not because they would contest a premise of the above: most sign all of it. But because they hold that the strong concept was never the right target; that the freedom that matters, the one the word « free » really names in real mouths, is something else, which comes out unscathed from my two blades and from all the determinism in the world. That position is called compatibilism, it is the majority view in the profession, it has real arguments, and it deserves better than the caricature usually inflicted on it, including by the best sellers of my own camp. So I will present it at full strength, as if I believed it. Then I will say why I think it is a subterfuge. A respectable subterfuge, brilliant even. But a subterfuge.

IV

The respectable subterfuge

Here, then, is the adversary, and first the rules of engagement. The compatibilist contests nothing of what precedes. Not a premise, not a blade, not a fact. He signs determinism, he signs the machine, he would probably sign this text up to this point. His disagreement bears on a single point, and it must be stated exactly, because everything plays out there: he holds that the word « free » never referred to what I have just buried. That the freedom that matters, the freedom of real mouths and real lives, does not oppose the caused to the uncaused, but the constrained to the unconstrained. The founding example is Hume's10, and it is formidable. Two men, on a quay, fail to save a drowning child. The first is in chains. The second watches, shrugs, and walks away. Both are fully determined: every neuron, every cause, all the way down. And yet you feel a difference between them, and that difference is not an illusion: it is real, physical, describable. The second acted according to his own desires, his own reasons, his own deliberation, without hindrance. That, says the compatibilist, is what « free » has always meant. A court never asks whether your atoms obeyed physics. It asks whether you had a gun to your temple.

And the position does not stop at that example: it has floors, and they are well built. Frankfurt11 distinguishes first-order desires (I want to smoke) from second-order desires (I want to want to smoke, or I wish I no longer wanted to): free is the one whose floors line up. The happy smoker and the addict who hates himself are both determined; only one is a slave, and the difference between them is once again real and physical. The same Frankfurt produced, two years earlier,12 the most disturbing thought experiment in the file. Imagine a neurosurgeon has implanted in your skull a device ready to intervene if you were about to choose A. But you choose B, on your own, for your reasons: the device never fires, it remains a spectator. You could not have done otherwise, since the device would have prevented it. And yet the intuition, nearly unanimous, holds you responsible for B: everything came from you. Frankfurt's conclusion: responsibility never required alternative possibilities, it requires that the act flow from your own machinery. Note what that targets: the « you could never have done otherwise » of my first blade, disarmed at a stroke, or at least ruled off topic. Dennett, finally, puts the roof on the edifice13: a system capable of representing its options, weighing them, responding to reasons, correcting itself, a system you can make change its conduct with an argument, possesses, he says, all the varieties of free will worth wanting. To demand on top of that an uncaused causation is to demand a magic no one can even describe. That is why this position is the majority one among professionals: it saves everything people care about, responsibility, the difference between the voluntary act and the epileptic fit, without denying a single result of science. Freedom is rehoused inside the causal order instead of being raised against it.

And now, before striking, I want to grant what must be granted, without reservation, because it is true: everything above describes real differences. Constrained and unconstrained: that exists. Desires aligned and desires at war: that exists. Responsive to reasons and closed to reasons: that exists, and these distinctions are among the most important that can be drawn about machines like us; everything that follows in this essay will keep them preciously. If the debate were about whether these properties exist, compatibilism would have won before starting. But the debate is not about that. It is about a label, and about what the label carries with it.

For here is the charge, aimed at the right place. I do not reproach the compatibilist with denying physics: he is sitting on it. I charge him with a bait and switch. The word « free will » had a job. It was not there to distinguish the chained man from the unchained one: « constrained » already did that. It was there to fund something else: merit and blame as metaphysical debts. The pride of the one who succeeds, the guilt of the one who errs, the retribution of the one who harms, all of it rested on the idea that the agent could really have done otherwise, right there, under exactly the same conditions, and that he is the origin of what he did. It is that concept, the strong one, that the two blades killed. The compatibilist then arrives with a replacement product, non-constraint, perfectly real, sticks the original's label on it, and proclaims that nothing was lost. It is redefining « ghosts » as « memories of the dead » and announcing, technically triumphant, that ghosts exist. The maneuver has a history, and I am not the first to name it: Kant called it a wretched subterfuge14, adding that the freedom thus saved is worth that of a turnspit, which carries out its motions by itself once it has been wound up. William James spoke of a quagmire of evasion15. The grievance has been the same for two centuries: when people learn determinism and feel that something is being taken from them, what they feel leaving is the strong version. Answering « but look, non-constraint is still here » does not return the lost object: it changes the subject with confidence.

It remains to deal with the two floors that seemed to parry my blades, and with one empirical fact. Frankfurt first: his device is a marvel, and it establishes what it establishes, namely that our attributions of responsibility track not alternative possibilities but the provenance of the act. Very well. But provenance is precisely the territory of the second blade, and the second blade could not care less about alternatives: the act flows from your machinery, yes, and your machinery flows from a past you never signed. Frankfurt wins his battle and does not touch the war: he shows where our intuitions of responsibility look, not that what they find there suffices to fund merit. Dennett next: « all the varieties of free will worth wanting » is a magnificent sentence, and watch it work. Who decides what is worth wanting? The implicit criterion is: what a machine like us can actually have. It is a peace treaty with the available, rebranded as victory. One may sign the treaty, I sign it with both hands; one does not then parade as if the stronghold had not fallen. As for the empirical fact, it must be given honestly because it cuts both ways: when ordinary people are questioned about deterministic scenarios, their answers depend on the framing. Described abstractly, a determined world seems to them to exclude responsibility: incompatibilist intuition. Described concretely, with a revolting crime, they blame anyway: compatibilist intuition. The lesson is not that the folk vote for one camp; it is that the ordinary concept contains the strong ingredient, that it comes out whenever one thinks coldly, and that the compatibilist redefinition therefore does amputate something people believed they possessed. A word, finally, on the best seller of my own camp: Sapolsky is right about the conclusion16 and dispatches this particular adversary in a few irritated pages. That is the one place where this text claims to do better: you have just read compatibilism at full strength, and the verdict that follows did not need to weaken it first.

 one level down · exhibits appended to the trial

Kant's turnspit is not an image I lend him: it is in the Critique of Practical Reason (1788). Aiming at the compatibilist freedom of his day, that of an act « not externally compelled » yet entirely determined from within, Kant writes that such a freedom would be worth no more than that of a turnspit, which, once wound up, carries out its motions by itself. The « wretched subterfuge » targets the same position, in the same passage. James, a century later, in The Dilemma of Determinism (1884), ranks compatibilism among the « quagmires of evasion » under which the real problem lies buried.

The data on ordinary intuitions come notably from Nichols and Knobe (2007)17: in a universe explicitly described as deterministic, a large majority of subjects judge that no one in it is fully morally responsible (abstract condition); faced with a concrete, emotionally charged crime in the same universe, a majority blames the agent (concrete condition). The authors read this as emotion swaying the judgment; compatibilists read it the other way, abstraction as the distortion. The point I draw from it is more modest and robust to both readings: the strong ingredient is part of the folk concept, since a cold framing is enough to bring it out.

One honesty note, finally: there exist compatibilists who refuse the maneuver I describe, admit that strong merit falls with the strong concept, and rebuild responsibility on purely forward-looking grounds (protect, deter, repair). Those ones sell no substituted product: they display the new price. My quarrel with them is almost nil; it will resurface, pacified, in the last section.

The verdict, then, and it fits in two sentences. Compatibilism is right about every fact and every distinction: keep it all, the non-constraint, the floors of desire, the responsiveness to reasons, those are the true coordinates of what matters in a life. It is wrong about the transfer of warranty: those real properties cannot finance what the label promised, merit and blame as metaphysical debts, and sticking them under the old name amounts to selling the house while letting the buyer believe the foundations are included. Which leaves the file in a remarkable state: the power never existed, in any possible world; the neighboring properties exist but do not replace it; and yet the feeling, for its part, is there, universal, insistent, identical in the compatibilist, in me and in you, factory-installed on every unit of the machine. A rendering that nothing founds but that everyone runs cannot be dismissed: it has to be explained. Why does your machine manufacture, at every instant, the sensation of a power it does not have? That is the question in the title, and it is the section that comes next.

V

What the feeling is for

Let us pose the question in the terms it deserves. The feeling of choosing is universal, insistent, costly to produce, and factory-installed on every unit of the machine. Now, biology does not willingly maintain universal, costly equipment for nothing: a trait that persists everywhere has generally paid for its place, or travels attached to something that paid. Careful with the word « paid »: evolution has no goals, it makes nothing in order to. A rendering like this one persisted because it made the machine perform better, or it would not have held on. What follows is therefore of a different register from the previous sections, and I announce it without detour: the blades were demonstrations, this is a functional hypothesis. It is reasonable, it is supported, it remains a hypothesis. But it answers the only question left standing: why does your machine manufacture the sensation of a power it does not have?

First function: bookkeeping. A machine acting in a world must distinguish, continuously and without delay, « I did this » from « this happened to me ». Without that tag, no fine motor control, no learning: impossible to attribute an outcome to the right cause, hence impossible to correct the aim. The feeling of agency is that tag. And recall what section II established in passing, for the pieces assemble now: Wegner showed that this feeling is a computed attribution, produced after the fact from cues (the intention preceded the act, the act resembles the intention, no other cause is visible), and that it can err in both directions. A direct perception of causality could not err like that; a manufactured signal can. The pathologies confirm it by breakdown: the so-called alien hand is the labeling machine that has stopped stamping; the delusion of control is the labeling machine stamping what does not belong to it. One never sees better that a mechanism exists than at the moment it breaks.

Second function: simulation. When you « feel » that 5 and 7 were available at the moment of writing 3, you are not reporting a property of the world: you are reporting an activity of the machine. To deliberate is to run branches offline: the machine represents several possible continuations, evaluates them, executes one. The options were indeed « available », but only in the sense that the simulator represented them; the error, the only one, consists in reading representational availability as metaphysical availability. And the feeling of an open future is simply what it is like to be a system in the middle of comparing branches before collapsing onto one. Even regret, that seemingly gratuitous torture, earns its wages here: replaying the untaken option is training at a discount, an error exploited without having been committed twice. A machine that never felt « I could have done otherwise » would learn nothing from its branchings: the counterfactual is a teaching tool, not a testimony from beyond causality.

Third function: social grip. Blame and praise are behavior-modification technologies, and note the delicious paradox: they work precisely because we are machines. To blame an agent is to inject an input that reconfigures his future outputs; to praise, the same, opposite sign. On a being endowed with strong free will, whose acts would spring from nowhere, these levers would bite on nothing: you cannot train an uncaused cause. It is on the determined that sanction gets its grip. And a community of agents who feel like authors, who anticipate having to answer, offers the social lever its handle: the feeling of authorship is the interface through which the group steers its members. I state it this way, in pure mechanics, and I stop there: the literature that claimed to show experimentally that shaking the belief in free will makes people dishonest has fallen in replication, and I refuse to lean on it. The detail is in the block, because this house cuts its convenient ammunition, even when it serves its story.

And now the centerpiece, the one that changes the status of everything else: the feeling does not merely render services, it tracks a fact. That fact has a name, computational irreducibility, and it is established on systems far simpler than you. For most sufficiently rich systems there exists no shortcut: no formula delivers state number one thousand without unrolling the thousand steps; the fastest simulation of the system is the system itself. Apply it to the machine you are. You are determined, each state follows from the previous, and you are incompressible: nothing and no one can know your next output at a lower cost than letting you run, or executing a complete copy of you, which is not a shortcut, it is you in duplicate. Your future is therefore closed in the metaphysical sense and open in the only operative sense there is: unavailable before it happens, for you as for everything else. That, I believe, is what the feeling of freedom renders on screen. It is to incompressibility what color is to wavelength: a rendering, real as experience, false as physics, and plugged into something true. The machine reads its own unpredictability from the inside and displays it in the only form an interface knows how to give: « it is open, it is my move ». The illusion is not a hallucination of nothing. It is the lossy compression of an exact fact.

One guardrail before concluding, because I have just walked along the edge of the very precipice I dug in the previous section. I accused the compatibilist of a bait and switch; have I not just committed one, offering incompressibility as the « true referent » of the feeling of freedom? No, and the difference fits in one word: the warranty. The compatibilist sticks the old label on the new property so that the warranty follows: merit, blame, debt. I name what the feeling tracks, and I attach nothing to it. Incompressibility describes; it does not found. It makes no one responsible in the strong sense, it restores no debt, it explains why the rendering exists and what it aims at, full stop. Naming the residue is not selling it under the old brand. Which leaves the file complete, and the bill on the table: the power never existed; the feeling exists, it works, and it aims at a real fact; but merit, blame, pride and guilt had been underwritten on the power, not on the feeling. It remains to count what dies with it, and what it costs to know. That is the last section.

 one level down · two precautions

First precaution, the convenient ammunition I refuse. A famous study (Vohs and Schooler, 2008)18 claimed that exposing subjects to a text denying free will made them more inclined to cheat afterward. It became the standard citation for the « dangerous truth »: even if true, the idea should stay hidden. Large-scale replication attempts failed to recover the effect. I could have used it one way (the feeling is so useful that dislodging it derails the machine) or the other (nothing happens, the fear was unfounded): I use it in neither. A datum that does not survive replication is not a datum.

Second precaution, the fine print of irreducibility. The concept comes from Wolfram19, who isolated it while studying cellular automata: systems with childishly simple rules, some of which produce a complexity that no analysis condenses. Not all systems are irreducible, and the property is not provisional ignorance: it is a mathematical fact about the process, not a limit of our instruments. That a brain is subject to it is highly plausible, not strictly demonstrated; it is the conjectural link in the centerpiece, and I mark it as such. One clarification, finally: irreducibility does not forbid prediction « in principle », it forbids cheap prediction. A Laplacean demon could compute your future, provided it executed a complete copy of you at real speed or worse: the fastest model of you is you. That is exactly the gap of which the feeling of freedom is the rendering.

VI

What dies, and what it costs

Let us do the accounts, and first the loss column. What dies with the ultimate author is the metaphysical ledger: merit and blame as debts, pride and guilt as claims. No one has ever deserved anything in the strong sense. The one who succeeds did not forge his talent, nor the appetite for work that made it bear fruit, nor the capacity for effort itself: effort is a card like the others, dealt with the hand. The one who harms did not choose the machine that chose: his indifference, his violence, his inability to weigh what should have weighed, all of it happened to him like the color of his eyes. The lottery goes all the way down; there is no floor, underneath, where a self-made self would be waiting to receive its due. It is brutal, and it is the direct consequence of the two blades: nothing is owed to one who is the origin of nothing, and there is no account to make anyone pay.

The asset column now, because it is not empty. There survive intact all the distinctions section IV granted: constrained and unconstrained, desires aligned and desires at war, responsive to reasons and closed to reasons. There survives incompressibility, which makes of every machine an unrolling that nothing summarizes. And there survives something this text demonstrates by existing: arguments work. You are a machine that a piece of reasoning can reconfigure, and that is a remarkable property, perhaps the most precious in the lot. But note the status of every line in this column: everything there describes, nothing founds. Non-constraint guides intervention, it does not manufacture merit. Incompressibility explains the feeling, it creates no debt. The ledger stays dead; it is the navigation instruments that survive.

Then comes the most human objection in the file, and it is due to the other Strawson, the father. Never mind your demonstrations, he says in substance: gratitude, resentment, indignation are not theories one abandons, they are attitudes constitutive of life among persons. You can prove whatever you like; faced with betrayal you will feel the bite, and faced with the gratuitous kindness, the warmth. The purely objective stance, the one that would regard everyone as a mechanism to manage, is untenable at the scale of a life; hence the metaphysical debate idles, the practice justifies itself. He is right about the fact and wrong about the conclusion, and the gap between the two is the whole point: that a reflex is ineradicable does not make it well founded. Fear of the dark survives perfectly well the demonstration that there is nothing in the dark; it keeps occurring, it has merely stopped being right. Resentment will keep rising in me, I predict it the way one predicts weather; but I now know it covers no debt, and that knowledge is not nothing: it modulates. A rage that knows it strikes a dealt hand cools faster than a rage that believes it strikes an author. Which is exactly what the old master who opened this text promised: understanding the causes does not remove the affect, it transforms it. The stone opened the essay; Spinoza may close it.

The practical consequences fit on one page, and the heaviest concerns punishment. If no one is the origin of anything, retribution dies: making someone suffer to settle a fault is keeping accounts on an empty ledger, a bookkeeper's torture. What remains standing is nonetheless enough for everything a society must do. Protect: you isolate the dangerous as you isolate the contagious, without hatred, to the strict minimum, and, note it, with a duty of care that retribution never knew: you do not punish a patient for being a carrier, you treat him, and quarantine obliges you to look for the cure. Deter: incentives bite, and section V said why, they bite because we are machines; you cannot train an uncaused cause, you can steer the determined very well. Repair, reconfigure, prevent upstream, where the machines get built. None of this is soft leniency: it is honest bookkeeping, often more demanding than vengeance, because it forbids paying oneself in suffering and obliges one to produce results. And here I return the promised handshake: to the compatibilists who admit that strong merit has fallen and rebuild responsibility looking only forward, I have nothing to oppose but the label. We keep the same ledger; they have bound it under the old cover.

There remains the personal price, and I promised it without anesthesia. The one who wrote this text deserves no credit for having written it. It came out of a machine I did not choose, assembled by a past I never signed, driven by a need for coherence I did not install. The one who reads it will deserve no credit for having understood it, and the one who rejects it, no fault: everyone does what their hand does. The pride I will feel all the same in publishing it will keep occurring, section V explained why the rendering runs; but it has changed status before my eyes: a dashboard light, pleasant, useful even, and holding no claim on the universe. Guilt likewise: a recalibration signal, no longer a debt. Is it like something, to live knowing it? Yes: less than one fears, more than nothing. One loses a throne of which one had only ever occupied the feeling; one keeps everything else, including the feeling. And I close the opening confession, because it deserves its answer: I wanted to be wrong. I searched for the exit hoping to find it, and the hope itself was a gear, and the search an unrolling, and this text is what a machine of this type produces when you let it run honestly on this question. I did not find the exit. I found why there could not be one, and why the door painted on the wall is so convincing. It is less than I wanted. It is all there was.

 one level down · the file for the last section

The objection from the reactive attitudes comes from Peter Strawson, « Freedom and Resentment » (1962)20, the century's most influential article on the question: it opposes the participant stance (that of human relationships, saturated with gratitude and resentment) to the objective stance (that of the clinician before a mechanism), and holds that no metaphysical argument can make us adopt the second full-time. The text's reply concedes the psychological fact and refuses the justificatory leap: the ineradicable is not thereby vindicated.

The quarantine model comes from Derk Pereboom (Living Without Free Will, 2001)21, who names his position « hard incompatibilism »: a society's right to protect itself from carriers of danger, on the model of its right to protect itself from carriers of disease, under the same constraints: minimal restriction, dignity maintained, duty of treatment, no suffering inflicted for its own sake. It is the most worked-out forward-looking reconstruction in the literature, and the one this section largely joins.

One neighboring position deserves to be named in order to be set aside: the illusionism of Saul Smilansky22, for whom all of this is true and should stay quiet, the belief in free will being a social pillar it would be dangerous to shake. This text answers it by existing, and by section V: the only experimental datum that supported the danger of the truth did not survive replication. You do not protect people from a fact with a superstition; you teach them to live with the fact.

One last thing, and it is the close. The starting question, the one everyone asks, was « am I free? », and it was badly cut: it demanded a power no possible world can give, caused or random, there is no third box. The rectified question, the only one left and the only one that works: what does a machine do with the rendering of its own unpredictability, once it knows what it is? Answer: exactly what it was going to do, but knowing it. And that knowledge is not a spectator: it is an input like any other, which reconfigures the machine that receives it. You are not free of what this text will do to you. It will do something to you nonetheless: that is how one steers a machine that cannot be steered, one feeds it. You are a machine. You are a machine that has just read this, and that is already recomputing itself with it. This is not freedom. It is better than nothing, and it is exactly what there is.

SOURCES

References

  1. On lesions and cortical stimulation: J. M. Harlow, « Recovery from the Passage of an Iron Bar through the Head », Publications of the Massachusetts Medical Society 2, 327-347 (1868); W. Penfield & T. Rasmussen, The Cerebral Cortex of Man, Macmillan (1950).
  2. B. Spinoza, Letter 58 to G. H. Schuller (1674), in the Correspondence; the conscious stone and the line « such is that human freedom… » appear there.
  3. B. Libet, C. A. Gleason, E. W. Wright & D. K. Pearl, « Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential) », Brain 106 (3), 623-642 (1983).
  4. A. Schurger, J. D. Sitt & S. Dehaene, « An accumulator model for spontaneous neural activity prior to self-initiated movement », PNAS 109 (42), E2904-E2913 (2012).
  5. C. S. Soon, M. Brass, H.-J. Heinze & J.-D. Haynes, « Unconscious determinants of free decisions in the human brain », Nature Neuroscience 11, 543-545 (2008).
  6. D. M. Wegner, The Illusion of Conscious Will, MIT Press (2002).
  7. P. van Inwagen, An Essay on Free Will, Clarendon Press, Oxford (1983).
  8. G. Strawson, « The Impossibility of Moral Responsibility », Philosophical Studies 75 (1-2), 5-24 (1994).
  9. T. McKay & D. Johnson, « A Reconsideration of an Argument against Compatibilism », Philosophical Topics 24 (2), 113-122 (1996).
  10. D. Hume, An Enquiry Concerning Human Understanding (1748), Section VIII, « Of Liberty and Necessity ».
  11. H. G. Frankfurt, « Freedom of the Will and the Concept of a Person », The Journal of Philosophy 68 (1), 5-20 (1971).
  12. H. G. Frankfurt, « Alternate Possibilities and Moral Responsibility », The Journal of Philosophy 66 (23), 829-839 (1969).
  13. D. C. Dennett, Elbow Room: The Varieties of Free Will Worth Wanting, MIT Press (1984).
  14. I. Kant, Critique of Practical Reason (1788); compatibilist freedom is there called a « wretched subterfuge » and compared to that of a turnspit.
  15. W. James, « The Dilemma of Determinism » (1884), reprinted in The Will to Believe and Other Essays in Popular Philosophy, Longmans, Green & Co. (1897).
  16. R. M. Sapolsky, Determined: A Science of Life Without Free Will, Penguin Press (2023).
  17. S. Nichols & J. Knobe, « Moral Responsibility and Determinism: The Cognitive Science of Folk Intuitions », Noûs 41 (4), 663-685 (2007).
  18. K. D. Vohs & J. W. Schooler, « The Value of Believing in Free Will », Psychological Science 19 (1), 49-54 (2008); effect not recovered by subsequent large-scale replications.
  19. S. Wolfram, A New Kind of Science, Wolfram Media (2002); the concept of computational irreducibility.
  20. P. F. Strawson, « Freedom and Resentment », Proceedings of the British Academy 48, 1-25 (1962).
  21. D. Pereboom, Living Without Free Will, Cambridge University Press (2001).
  22. S. Smilansky, Free Will and Illusion, Oxford University Press (2000).