Wittgenstein Resolves the Free Will Debate

Table of Contents

• The Basic Idea
• Applying that Idea to Free Will
• Flattening the Discourse
• Improving the Overall Quality of our Discourse
• Enriching the Conceptual Space
• Enhancing Discussions about Human Agency
• Revisiting Kevin Mitchells "Free Agents"
• Conclusions

I was recently revisiting some of the concepts I discussed previously in my first Free Will post, but with Wittgenstein lingering in my short term memory. In the past, my conclusions was that "yes, we obviously have free will", but that our understanding of the term causes confusion. But now after reconsideration, I am not even sure it's a deciadable yes or no question; rather, it's an ill-posed false dichotomy. With Wittgensteins Language Games fresh in my mind, I somewhat had an epiphany: this problem is fundamentally ill-conceived. The proposition P="Humans have Free Will" is not truth evaluable as stated, because "Free Will" is a pseudo-problem primarily caused by misunderstanding language.

The Basic Idea

Wittgenstein’s basic idea is that a lot of philosophy looks deep because language misleads us. His conclusion, especially in his later work, is not that philosophical questions are stupid, but that many of them are not real problems in the way they first appear. They are “pseudo-problems” because they arise when we take words out of the ordinary situations that give them meaning and then treat the resulting confusion as a profound discovery.

His core thought is that we use words successfully in everyday life: “know,” “think,” “mind,” “time,” “meaning,” “exists,” “cause,” and so on. In normal use, these words work because they are part of shared human activities. But philosophers often detach them from those activities and ask things like:

• What is the essence of meaning?
• Where is the mind located?
• Does time itself flow?
• How do words connect to reality?
• How can I know other minds exist?

For Wittgenstein, many such questions become puzzling because we are bewitched by grammar. A sentence can be grammatically well-formed while still pushing us toward a false picture. For example, because “the mind” is a noun, we may assume it names a thing, like a brain or a chair, and then start hunting for what kind of thing it is. Or because we can say “time passes,” we imagine time as a kind of river-like object. Grammar makes these ideas look natural, even when they may be conceptually misleading. So the philosophical problem is often not a missing fact but a misuse or misunderstanding of language.

His remedy is not to build a grand theory, but to clarify how words are actually used. He wants philosophy to look carefully at “language-games,” meaning the concrete practices in which words have their role. Once you do that, the mystery often dissolves. A simple example: If someone asks, “What is the meaning of a word?” they may expect meaning to be some hidden object attached to the word. Wittgenstein says: look at how the word is used. Meaning is not usually a ghostly thing behind language; it is bound up with use. The confusion came from assuming every noun must correspond to some inner entity.

That is why he compares philosophy to therapy. Philosophy does not mainly discover new objects; it untangles confusions. It shows the fly the way out of the fly-bottle. The “bottle” is the trap created by our own forms of expression. So his conclusion can be put like this: Many philosophical problems are pseudo-problems because they are generated by language operating outside its proper everyday contexts. We mistake grammatical form for metaphysical structure, then invent puzzles that seem profound but actually rest on confusion. The job of philosophy is to diagnose and dissolve these confusions by examining how language is really used.

Wittgenstein does not mean that all philosophy is worthless. He means that much of it goes wrong when it tries to explain what should instead be clarified. Traditional philosophy often asks, “What hidden reality corresponds to this word?” Wittgenstein asks, “How is this word actually used, and what confusion arises when we force it into the wrong role?”

Applying that Idea to Free Will

From a Wittgensteinian angle, free will is a strong candidate for a pseudo-problem, or at least a problem that gets artificially inflated by linguistic confusion. Instead of asking, “Does free will really exist?” as though “free will” names some mysterious inner object or metaphysical power, Wittgenstein would push us to ask how expressions like these actually function:

• “She did it voluntarily.”
• “He had no choice.”
• “You can decide.”
• “She was forced.”
• “He acted intentionally.”
• “I could have done otherwise.”

In ordinary life, these expressions already have clear roles. We use them in law, morality, explanation, praise, blame, education, and everyday relationships. They help us distinguish between acting under coercion and acting willingly, between accident and intention, between compulsion and deliberation. The philosophical confusion begins when we tear these expressions out of those practical settings and turn them into a metaphysical riddle:

• Is the will an uncaused cause?
• Could the self have chosen differently in exactly the same total universe-state?
• Is freedom compatible with causal determination?
• At what point in the brain does the “real chooser” intervene?

A Wittgensteinian suspicion is that these questions may look profound mainly because grammar encourages a misleading picture. “Will” is a noun, so we imagine it must name a thing. “Choice” sounds like an event produced by an inner agent, so we imagine a little commander inside us. “Could have done otherwise” gets recast from an ordinary practical judgment into a cosmic counterfactual about the entire universe being replayed. That shift may create the illusion of a deep problem where there was originally just a family of practical distinctions. For example, in everyday life, saying “She could have stayed home” often means something like: she was not physically compelled, not coerced, knew the options, and had the ordinary capacity to do either. It usually does not mean: if the entire history of the universe up to that instant were held fixed, some metaphysical faculty could still have broken through causality.

The second reading is a philosopher’s reconstruction, not the ordinary use of the phrase. And once that reconstruction is imposed, the debate starts to feel impossible.

We start with useful human concepts like responsibility, choice, coercion, regret, decision, and intention. Then we compress them into one abstract noun, “free will,” and ask whether it “exists” absolutely. That encourages us to search for the essence of freedom, as if all the varied uses had one hidden metaphysical core. The result is a verbal trap. Wittgenstein would likely say that there may be no single problem of free will here, only many different language-games being run together: moral responsibility, criminal liability, psychological self-control, deliberation, political liberty, absence of coercion, religious notions of the soul, and scientific causal explanation. Once these are disentangled, the giant philosophical problem may partly dissolve.

That said, a careful Wittgensteinian would not necessarily say the whole topic is empty. He would more likely say that some formulations of the problem are confused. Real questions remain, but they are more local and practical:

• What counts as coercion?
• When is someone responsible?
• How do addiction or mental illness affect agency?
• What is the role of intention in explanation?
• How should legal and moral practices respond to human psychology?

Those are not pseudo-problems. They are intelligible because they stay connected to actual human practices.

So a good Wittgensteinian summary would be that the “free will problem” becomes a pseudo-problem when it asks for the metaphysical essence of freedom beyond the ordinary uses of words like “choice,” “responsibility,” and “could have done otherwise.” The apparent depth of the debate may come less from a discovery about reality than from language being stretched beyond its normal role. The moment “Was the act free?” turns from a practical question into “Is there, behind action, a metaphysically independent faculty that escapes causality?”, the debate may already be in trouble.

There is also a likely objection: some philosophers would say this is too quick, because neuroscience and physics really do raise substantive questions about agency. The Wittgensteinian reply is not that science is irrelevant, but that science does not automatically validate the metaphysical picture. It may change how we apply concepts like responsibility or control, but it does not prove that the old philosophical riddle was well-formed in the first place.

Flattening the Discourse

One of the strongest practical payoffs of a Wittgensteinian approach is that it does not merely “debunk” a philosophical debate, it reorients attention toward distinctions that are actually useful.

The free will debate often becomes sterile because both sides are tempted into abstraction. The libertarian says: at the crucial moment, the person simply chose. The hard determinist says: no, the choice was just the inevitable output of prior causes. Both sides, despite appearing opposed, often share the same distorted picture. They both treat action as if it were best understood by isolating a single decision-point and asking what metaphysical status attaches to it. That is exactly the kind of picture a Wittgensteinian would distrust.

The problem is not that “choice” is meaningless. The problem is that the word is made to do the wrong kind of work. In ordinary life, “choice” is not usually a name for a tiny metaphysical miracle or a micro-event whose essence we must discover. It belongs to a much wider practical grammar: capacities, reasons, pressures, habits, institutions, incentives, threats, available alternatives, competence, timing, ignorance, manipulation, fatigue, addiction, character, and social setting. Once we pull “choice” out of that web and ask whether it is absolutely free or absolutely determined, discourse flattens.

That flattening happens in two mirror-image ways. The libertarian flattening says: the person could have done otherwise, therefore the act is fundamentally attributable to them in a simple and ultimate sense. That move compresses an entire landscape into a point. It ignores how what a person can actually do depends on upbringing, learned repertoires, social options, fear, scarcity, status competition, trauma, habits of attention, local norms, cognitive bandwidth, and the structure of the environment. It treats agency as if it were maximally visible at the instant of action, when in many cases agency is distributed across a longer sequence: what the person was exposed to, what alternatives were intelligible to them, what self-control resources were available, what prior commitments they had made, and what institutional channels were open. The determinist flattening makes the opposite reduction. It says: since every event has causes, the decision is “really” just physics. But that often empties the concept of decision of its role. Of course human action is physically realized; that is not in dispute. The problem is the leap from “decisions have physical conditions” to “the proper or deepest description of decisions is at the level of physics.” In practice, that is like saying a market is really just particle motion, or a promise is really just air vibrations, or a legal judgment is really just neurochemistry. None of those are false in a thin sense, but they are useless as replacements for the human-level descriptions in which explanation, prediction, evaluation, and intervention actually happen.

This is where Wittgenstein helps. He pushes us to ask: what are we doing when we talk about deciding, choosing, being responsible, being constrained, being manipulated, being capable, acting under pressure? These are not all synonyms. They belong to different but overlapping language-games. Once we stop demanding that they all answer to one metaphysical master concept called “free will,” the discussion becomes more articulated.

Take choice architecture. A Wittgensteinian reframing makes it much easier to see that action is not only about what happens “inside” a person at one instant. It is also about how the environment presents options. Defaults, framing, timing, salience, friction, order effects, social proof, and informational structure all influence what people actually do. In ordinary language we already recognize this. We say that someone was steered, cornered, primed, set up, tempted, overwhelmed, rushed, confused, or guided. Those expressions are not marginal; they are central to how we understand agency. They show that our concept of action already includes situational shaping. The crude free-will debate often suppresses this by forcing a false alternative: either the act came from a pure inner chooser, or it was not really “the person’s” act at all. But that is not how our practices work. We routinely distinguish between degrees and modes of control.

The same is true of nudging. The interesting question is not whether nudges abolish freedom in some absolute sense. The interesting questions are: how do they alter attention, effort, and default behavior? When does a nudge preserve agency by helping people do what they already endorse? When does it become manipulative because it bypasses reflective capacities or exploits vulnerabilities? Those are more precise questions because they keep the concept tethered to actual use. They let us discuss transparency, reversibility, contestability, asymmetries of information, and whether the intervention expands or narrows a person’s effective control over their own conduct.

Opportunity sets are another major gain from this reframing. “He chose X” tells us almost nothing unless we know the menu from which X was selected. A person’s formal options and effective options can be radically different. Two people can both “choose” not to go to college, but one does so from a position of abundance and the other from a position of financial pressure, bad school quality, family obligations, and lack of credible support. To describe both simply as “their choice” is to erase the grammar of constraint. We ordinarily know better than this: we distinguish choosing among genuine alternatives from settling under pressure, accepting the least bad option, or responding to a rigged field. The free-will frame often blurs these distinctions because it obsesses over whether the person, at the final instant, authored the act. A more useful frame asks what alternatives were live, accessible, legible, sustainable, and socially supportable.

Systems that constrain choice become visible once we stop imagining agency as a self-enclosed spark. Crime is a good example. Saying “it was just their choice” often treats the criminal act as though it emerged from a vacuum. But real cases usually involve developmental and situational sequences: neighborhood violence, poor schooling, substance abuse, criminal networks, absent guardianship, unstable work, humiliation, hypervigilance, retaliatory norms, low trust, and repeated exposure to immediate-payoff incentives over long-horizon planning. None of this magically absolves a person. But it changes the description of what kind of control they were exercising, what capacities were available, and what interventions could matter. Responsibility does not disappear; it becomes more textured. One can still hold that the person acted intentionally and should answer for the act while also rejecting the simplistic idea that the act is intelligible as a bare atom of choosing.

This is one of the most important places where Wittgenstein is clarifying rather than excusing. He is not saying that once we see complexity, nobody is accountable. He is saying that accountability itself is a family of practices with many purposes: condemnation, deterrence, repair, protection, moral address, rehabilitation, reaffirmation of norms, and recognition of victimhood. The metaphysical free-will debate often confuses these by pretending there is a single question underneath them all: did the person possess ultimate contra-causal freedom? But our actual practices do not depend on such a test. They depend on distinctions like intention versus accident, compulsion versus ordinary temptation, capacity versus incapacity, informed versus misinformed action, sustained pattern versus isolated lapse, remorse versus indifference, and responsiveness versus unresponsiveness to reasons.

Decision theory also becomes easier to talk about once the free-will metaphysics is loosened. Decision theory is not mainly about whether the chooser transcends causality. It is about structured reasoning under conditions of uncertainty, tradeoff, limited information, and limited computational capacity. It studies preferences, incentives, risk tolerance, dynamic inconsistency, strategic interaction, and bounded rationality. Those topics are fruitful because they do not ask for the hidden essence of “free choice.” They ask how agents actually navigate alternatives. A Wittgensteinian would appreciate that these are not attempts to name the metaphysical source of action but to model particular aspects of practical reasoning. That is why they can be empirically connected and operationally useful.

The focus on sequences of decisions is especially important. The usual free-will debate loves the snapshot: at 8:03 p.m., did the person freely choose or not? But many human outcomes are path dependent. An addiction, a career, a crime pattern, a marriage, a debt spiral, a political identity, a health trajectory, or a violent feud is rarely explained by one isolated act. These are temporally extended formations. Small earlier moves alter later menus. Habits shape perception. Repetition lowers friction. Commitments create identities. Social expectations lock in behavior. Sunk costs and escalating commitments narrow exits. Early failures can damage confidence; early successes can expand ambition. What matters is not just whether a single act was free in some absolute sense, but how a person’s option space was gradually built, narrowed, widened, corrupted, stabilized, or redirected over time.

That is where terms like cumulative effects and path dependency become indispensable. They allow us to see that a later “choice” may be both genuinely attributable to a person and also deeply conditioned by earlier structure. A person can be both agent and product of a trajectory. The free-will binary obscures this by encouraging the thought that either the person is a sovereign originator or a passive mechanism. Human life is usually neither. People are shaped beings who also participate in their own shaping. They inherit patterns and modify them. They are constrained and improvisational. They are neither acausal gods nor billiard balls.

This connects with emergence. Saying that decisions are “just physics” fails because human agency is a higher-level pattern with its own explanatory grammar. Not a spooky substance, but a real organization of behavior, sensitivity to reasons, self-monitoring, temporal projection, norm-following, and social interpretation. An intention is not a second substance floating above the body. But neither is it replaced by talk of molecules. It belongs to a level of description at which persons deliberate, justify, promise, regret, and revise. This is what emergence is useful for here: not mystification, but resistance to bad reduction. The higher-level pattern is physically realized without being explanatorily obsolete.

Patterns and archetypes also become discussable in a healthier way. Often what matters is not one choice but a recurrent form: the self-sabotaging pattern, the retaliatory pattern, the avoidance pattern, the domination pattern, the rescue pattern, the short-term reward pattern, the prestige-seeking pattern. These are intelligible only across time and context. To ask whether one episode manifested “free will” is often less illuminating than asking what pattern this episode belongs to, how stable that pattern is, what cues trigger it, what social environment reinforces it, and what practices might interrupt it. The free-will debate narrows vision; pattern-language widens it.

Control, too, becomes more nuanced. Control is not all-or-nothing. We regularly speak of diminished control, partial control, regained control, self-control, situational control, strategic control, and control over some variables but not others. Someone may have control over whether they speak but not over whether they feel anger. They may control short-term impulses poorly but manage long-term planning well. They may be unable to stop a craving but able to build routines that reduce exposure. These gradations are central to law, ethics, therapy, education, and policy. A metaphysical free-will framing often distorts them by asking for a single threshold where “real freedom” appears or disappears.

Once we shift the frame, the conversation becomes more practical and morally serious. Instead of asking whether the criminal act was “really free,” we can ask:

• What capacities did this person have at the time?
• What alternatives were genuinely available?
• What social and institutional conditions structured the option set?
• Was there coercion, manipulation, desperation, addiction, or ignorance?
• Is this act part of a pattern?
• How responsive is this person to incentives, sanctions, treatment, or moral dialogue?
• What forms of accountability protect society while also reducing recurrence?
• Which interventions widen future control rather than merely punish past behavior?

Those questions are richer because they allow multiple truths at once. Someone can be blameworthy and also deeply shaped by circumstance. They can deserve sanction and also be a predictable product of institutional failure. A system can be unjust without every individual within it being exonerated. A person can have agency without being an uncaused causer.

Likewise, against reduction to physics, we can say something more precise than “that’s absurd.” The problem is category confusion. Physics describes one level of reality. Human action-talk describes another. When deciding whether someone acted intentionally, whether a confession was coerced, whether a contract was informed, whether a policy produces perverse incentives, or whether a nudge respects autonomy, appeal to microphysics is not false so much as irrelevant. It does not answer the question being asked. It is like answering a question about whether a poem is ironic by giving a chemical analysis of the ink.

Wittgenstein would say that the meaning of these terms lies in their use. “Choice,” “control,” “constraint,” “intention,” “responsibility,” and “freedom” do not all get their sense by pointing to one hidden metaphysical ingredient. Their sense comes from the roles they play in our practices. That is why philosophical fixation on ultimate freedom can be counterproductive: it distracts from the distinctions that actually guide judgment and action. This has consequences across fields.

• In ethics, it encourages movement from abstract desert toward differentiated responsibility, moral formation, and the conditions under which people become more answerable to reasons.
• In law, it supports more careful distinctions among intent, recklessness, negligence, coercion, diminished capacity, predation, and systemic mitigation.
• In public policy, it makes room for debates about institutional design, education, poverty, policing, incarceration, labor markets, addiction treatment, and urban planning as conditions of agency rather than external background.
• In behavioral science, it aligns with the study of bounded rationality, framing, defaults, and temporal discounting without implying that persons are mere puppets.
• In psychology and therapy, it supports talk of agency as scaffolded and recoverable rather than present in all-or-nothing form.
• In political philosophy, it helps move from the myth of purely self-made individuals toward a clearer view of how liberty depends on material and civic preconditions.
• In organizational settings, it helps explain why bad outcomes are often products of incentives, reporting structures, and normalization processes rather than isolated bad actors alone.

That last point matters. One reason the classical free-will frame can be ideologically useful is that it individualizes outcomes too quickly. “They chose it” becomes a way of refusing to inspect systems. But the opposite ideology is also possible: “the system caused it” can become a way of refusing to address individual conduct. A Wittgensteinian untangling resists both simplifications. It asks us to use the right concepts in the right places. Sometimes the right question is about personal culpability. Sometimes it is about institutional design. Often it is both.

The deeper philosophical payoff is that this reframing restores plurality. There is no single axis called “free will” on which every meaningful question about agency lies. There are many axes: voluntariness, awareness, coercion, competence, self-command, practical rationality, availability of alternatives, quality of information, temporal stability of preference, susceptibility to manipulation, social support, developmental history, institutional context, and responsiveness to reasons. Once these are separated, discourse becomes less theatrical and more diagnostic.

The free-will debate often flattens the topography of human action into a binary. Either total authorship or total causation. Either sovereign chooser or physical mechanism. But lived action has contours. It has gradients, thresholds, channels, bottlenecks, loops, and sedimented history. Human beings act within structures, through habits, across time, under interpretations, with varying powers of revision. A language that can describe only a single decisive moment is not rich enough. So the Wittgensteinian contribution is not merely negative, not just “the debate is confused.” It is constructive. It clears away a misleading picture and returns us to a broader field of description where more serious inquiry can begin. Instead of asking for the metaphysical essence of freedom, we can investigate the ecology of agency. That ecology includes:

• how options are presented,
• which alternatives are really available,
• how institutions shape conduct,
• how patterns form,
• how feedback loops reinforce behavior,
• how earlier actions narrow later choices,
• how capacities can be trained or damaged,
• how social meanings affect decision-making,
• and how accountability can be calibrated to realities of both person and system.

That is a much more fertile conversation. It does not solve one giant philosophical riddle. It replaces the riddle with many better questions. The traditional free-will debate often mistakes the metaphysics of a single decision for the reality of human agency. A Wittgensteinian reframing shifts attention from isolated acts to the grammar of action: capacities, constraints, systems, histories, patterns, and practical contexts. That makes possible more nuanced discussions of responsibility and control without collapsing either into moralism or into reductionism. And perhaps the most important gain is moral seriousness. Once we stop using “choice” as a blunt instrument, we can become more exact about both blame and help, both judgment and reform. That is not evasion. It is an improvement in description, and better description is often the beginning of better ethics and better politics.

Improving the Overall Quality of our Discourse

The free-will issue is not just an abstract metaphysical debate. The picture one adopts has social, moral, and practical consequences. A bad picture of agency does not stay in the philosophy seminar room. It migrates into how people interpret success, failure, blame, poverty, crime, merit, identity, and self-improvement.

For examples, libertarian free will often lends itself to self-serving interpretations of the world. Not always, and not necessarily in every person who believes in it, but the structure of the view makes that temptation very available. If one imagines each person as standing outside causal structure in some especially strong way, then outcomes become easy to moralize. Success starts to look like a transparent reflection of superior choosing. Failure starts to look like a transparent reflection of inferior choosing. The thicker the belief in self-origination, the easier it is to read the world as a scoreboard of moral worth. That has a predictable psychological appeal. It flatters the successful. It protects self-esteem. It turns contingency into merit. It lets people narrate their lives as authored triumphs rather than as mixed products of effort, luck, timing, institutions, inherited resources, networks, stability, health, and circumstance. It is satisfying to think: I am where I am because I chose well. It is much less satisfying to admit: I may have worked hard and exercised judgment, but I also benefited from conditions I did not create.

That self-congratulatory tendency is not an accidental side effect. It follows quite naturally from a picture of agency that overstates the independence of the chooser from the conditions of choice.

Once that picture is in place, it easily feeds attribution errors. The classic move is to over-explain others’ behavior by character and under-explain it by situation, while doing the reverse for oneself when convenient. If I succeed, I see discipline, talent, foresight, grit. If someone else fails, I see laziness, irresponsibility, bad values, weak character. The background conditions recede. Structural obstacles become invisible. Constraints are reinterpreted as excuses. And this can scale from individuals to groups. That is where the danger deepens. A flattened free-will framework makes it very easy to take recurring disadvantage and read it as evidence of recurring bad choices, and then to take recurring bad choices as evidence of some inherent deficiency. The path is disturbingly smooth: bad outcomes become bad decisions, which become bad judgment, which becomes bad character, which becomes defective culture, which finally becomes some supposedly essential feature of the group.

Instead of asking what opportunity sets were available, what incentives were present, how institutions were arranged, what historical pressures shaped the group, what cumulative disadvantages narrowed later options, or how stereotypes themselves altered decision environments, the observer jumps straight from outcome to essence. The result is moral reification: contingent, historically produced patterns get redescribed as fixed traits. This is why unflattening the conversation is beneficial. Once we refuse to treat action as a bare point-event, it becomes much harder to turn social patterns into moral essences. We are pushed to ask better questions.

• Why do similar behaviors cluster in similar environments?
• What kinds of scarcity produce short-horizon reasoning?
• What institutions reward or punish trust?
• How does repeated exclusion alter risk tolerance?
• How do policing, schooling, housing, labor markets, and social narratives shape perceived options?
• What choices are technically available but not realistically actionable?

These are not excuses. They are explanatory disciplines. They slow down the rush from observation to judgment.

A Wittgensteinian point is helpful here. Once “choice” is no longer treated as the name of a pure inner act detached from context, the temptation to derive an essence from outcomes weakens. You stop asking, “What kind of person or group must they really be?” and start asking, “How are these patterns of action embedded in practices, incentives, histories, and constraints?” That shift is morally important because it resists premature condemnation without erasing responsibility.

It also improves how we think about our own lives. The self-serving version of libertarian free will does not just distort our judgments of others; it distorts self-understanding. It invites people to over-credit themselves for favorable trajectories and undersee the scaffolding around them. They may forget how much their conduct depended on family stability, mentors, social capital, timing, health, legal protection, or simply being spared catastrophic setbacks. This produces arrogance in the fortunate and moral suspicion toward the less fortunate.

There is also a political consequence. A society saturated with that picture will tend to underinvest in structural reform because it interprets structurally produced outcomes as individual revelations. Why improve schools, neighborhoods, addiction treatment, transportation, or labor protections if outcomes are already assumed to be faithful readouts of personal choice? The metaphysic becomes a justificatory device. Existing distributions look deserved because they are treated as accumulated acts of will.

That is why the critique reaches beyond philosophy into ethics and politics. A simplistic doctrine of free will can function as a moral technology for naturalizing inequality.

Unflattening the false dichotomy can indeed improve decision-making itself. Moving away from the crude contrast between “I am absolutely free” and “I am merely determined” does not weaken agency. In many cases it strengthens it. Better models of human action can lead to better action. Why? Because good agency depends on accurate self-understanding. If you misunderstand how decisions are shaped, you are less able to shape them well.

Suppose someone adopts the heroic picture of choice: every important outcome is basically a matter of deciding right in the moment. That person may neglect the architecture around the moment. They may fail to notice how fatigue changes judgment, how defaults pull behavior, how repeated cues sustain habits, how social context affects standards, how earlier commitments create later leverage, how incentives bias attention, or how certain environments systematically shrink imagination. They may overrate willpower and underrate design. But once you acknowledge path dependency, incentives, framing effects, and environmental structure, a richer practical intelligence becomes possible. You stop treating agency as a one-time exertion and start treating it as something cultivated through positioning. That is crucial: many good choices are made before the moment of visible choice.

• You choose what environment to enter.
• You choose what temptations to remove.
• You choose what routines to automate.
• You choose what skills to build.
• You choose what people to be around.
• You choose which commitments will bind your future self.
• You choose which options will remain open and which will close.

That is agency too, and often more consequential agency than the isolated decision point people like to moralize. This is where path dependency becomes practically transformative. If current options depend on prior trajectories, then rational action includes managing trajectories, not merely selecting from a menu at the last second. A wise agent asks not only, “What should I do now?” but also, “What sequence am I entering?” and “What future option set will this create?” That frame is far more realistic. A career, a relationship, an addiction, a debt pattern, a fitness habit, a research program, a political identity, or a criminal path is seldom the result of one sovereign act. It is usually the result of many reinforcing moves. Understanding this helps people intervene earlier, at lower cost, with greater effect.

A person who understands cumulative effects will think in terms of compounding. A person who understands constraints will invest in removing bottlenecks. A person who understands environment-shaping will stop relying exclusively on brute resolve. A person who understands incentives will pay attention to how systems reward certain behaviors and penalize others. A person who understands opportunity sets will seek to widen them rather than merely “choose better” within a narrow and degraded field. The quality of a choice depends partly on the quality of the option set. Sometimes the smartest move is not choosing among current options but changing the field so that better options become available later.

This is true at the personal level. Someone trying to improve their life may need less advice about “making the right choice” and more about expanding capacities, reducing friction, accessing institutions, building buffers, and entering environments with better return curves. The issue is not simply stronger will. It is strategic repositioning. It is also true at the organizational and policy level. If you want better decisions from employees, students, patients, drivers, consumers, or citizens, it is often not enough to exhort them morally. You design better systems: clearer incentives, less noise, better defaults, better feedback, lower administrative burdens, more transparent pathways, fewer perverse tradeoffs. Better outcomes often arise not from metaphysical freedom but from intelligently structured action spaces. This is why unflattening agency can be empowering rather than diminishing. It gives people more levers. They can shape their environments, alter inentives, sequence commitments, build capabilities, change peers, reduce exposure, increase slack, improve information, protect attention, expand options, and plan for feedback loops. That is a much more actionable conception of freedom.

It also yields a more mature understanding of responsibility. Responsibility need not mean pretending that each moment is born from nowhere. It can mean becoming skillful at recognizing the conditions under which one acts badly or well, and then learning to govern those conditions. In that sense, acknowledging influence is not a retreat from responsibility. It is often the beginning of real responsibility.

Someone who says, “My environment affects me, my habits affect me, my incentives affect me, my prior commitments affect me,” can use that knowledge to create better patterns. Someone who says, “Only my raw choice matters,” may actually become less effective, because they are blind to the mechanisms through which choices are stabilized or undermined.

This reframing also improves humility. If you understand your own success partly as a function of favorable scaffolding, you are less likely to worship your own judgment and more likely to maintain or recreate the conditions that support good decisions. You become less theatrical about willpower and more attentive to maintenance. You ask:

• What conditions help me think clearly?
• What pushes me into short-termism?
• When do I predictably make poor judgments?
• What recurring environments narrow my option set?
• How can I design my life so that the better path is easier to take?

There is also a broader ethical gain. People often fear that once you admit situational influence, personal agency will evaporate. But in practice the opposite can happen. When influence is acknowledged accurately, interventions become imaginable. You no longer oscillate between blame and helplessness. You can identify sites of leverage. That is true for individuals and for institutions. For example, with crime, the unflattened view would show that some people are dangerous and accountable, but dangerous conduct also emerges from developmental, situational, and institutional patterns; therefore serious response includes sanction, prevention, redesign, and trajectory interruption. Likewise with education, health, and work the richer view says behavior is shaped but shapeable, constrained but revisable, and therefore the task is to improve both personal practice and structural conditions.

The self-serving side of libertarian free will is not just morally unattractive; it is cognitively impoverished. It mistakes a flattering story for a useful model. And because it is a poor model, it often produces poorer decision-making and poorer institutions. It narrows causal imagination. It hides leverage points. It encourages blame where redesign is needed. It encourages pride where gratitude and realism would be healthier. A fuller model of agency does two things at once. First, it checks arrogance and attribution error by showing that outcomes are rarely simple readouts of isolated will. Second, it enlarges practical intelligence by showing that better choices often depend on shaping the pathways, environments, and option sets within which future choices will occur. That is why “unflattening” is such a strong word here. It is not simply adding complexity for its own sake. It is restoring the actual topography of human action: uneven, layered, cumulative, constrained, revisable, scaffolded, and pattern-forming. From that perspective, freedom is less like a mysterious spark and more like an achieved condition. It is not the absence of causes. It is the presence of usable capacities, intelligible alternatives, supportive structures, temporal foresight, and room to maneuver within a shaped world. The more those are developed, the better the choices tend to become.

A simplistic belief in libertarian free will can feed self-congratulation, attribution error, and essentialist judgments about others by obscuring the situational and structural conditions of action. By contrast, a richer account of agency that includes path dependency, incentives, environment, and opportunity sets not only improves our understanding of others but can also improve our own decision-making, because it reveals more of the real levers through which better choices are made possible.

Enriching the Conceptual Space

What Wittgenstein helps expose is not only that some formulations of the free-will debate are confused, but that the confusion blocks access to a richer vocabulary. The debate does not merely ask a bad question. It also monopolizes conceptual space. It takes a sprawling field of phenomena — deliberation, constraint, habit, adaptation, coordination, self-regulation, institutional shaping, temporal dependence, feedback, learning, drift, lock-in, reversibility, irreversibility — and forces them into one crude frame: “free” or “determined.” Once that frame dominates, many of the concepts that would actually help us speak carefully about action either get flattened or never enter the discussion at all.

A Wittgensteinian move does not just dissolve a pseudo-problem; it decomposes an overcompressed bundle of loosely coupled concepts. It lets us separate concepts that had been rhetorically glued together. “Choice,” “agency,” “control,” “causation,” “responsibility,” “constraint,” “predictability,” “voluntariness,” “path dependence,” and “determinism” are not interchangeable. But the classical free-will debate often treats them as if they all stand or fall together. Wittgenstein’s method is useful here because it asks: what jobs are these words doing? In what contexts do they make sense? What distinctions do they mark in ordinary and specialized practice? Where are we being misled by a grammatical picture into assuming there must be one hidden essence underneath them all?

That last point matters especially with causation. The free-will debate often inherits a very simple picture of cause: one thing pushes another in a linear chain, and if we trace the chain back far enough, everything must reduce to one continuous source of necessitation. Then the question becomes: where could “freedom” fit into that chain? Either nowhere, in which case all is determined, or somewhere outside it, in which case freedom must be a kind of causal exception.

This is exactly the sort of picture Wittgenstein would invite us to distrust. Not because causal language is meaningless, but because a specific image of causality is silently governing the debate without being examined. It is treated as obvious that explanation must have the form of a single linear sequence, and that agency must therefore either be absorbed into the sequence or break it. But that is already a philosophical picture doing hidden work. Once that picture loosens, a much broader landscape appears.

Feedback loops are a good example. Human action is often not best understood as an isolated output caused by a prior input. It is part of a circular and recursive process. Beliefs affect actions; actions alter environments; environments alter future beliefs and incentives; those changes then reshape the next round of action. A person’s confidence affects performance, performance affects social response, social response affects confidence. Institutions react to conduct, and those reactions change the conduct they later observe. Expectations can become self-fulfilling or self-canceling. That kind of process does not fit comfortably into the cartoon image of one-directional causation that often underlies free-will talk. But it is exactly the sort of process we need if we want to understand actual agency. Agency is often not a punctual event but a pattern sustained and modified across recursive interactions. The same goes for systems dynamics more broadly. Once we stop asking whether a person is an uncaused causer, we can start asking much more productive questions:

• What variables are endogenous here?
• What is being reinforced?
• What equilibria or traps exist?
• Where are the thresholds and tipping points?
• What delays make local choice shortsighted?
• What processes amplify small initial differences?
• What processes dampen shocks?
• What kind of adaptation is occurring?

Those are questions about agency in real systems rather than metaphysical agency in abstraction.

We can also construct an analogy to time series analysis. A lot of human conduct is temporally structured in ways the standard free-will debate barely notices. The usual metaphysical framing treats each decision as a self-contained moment. But in many domains, the present is saturated with lagged effects from prior states. Habits, norms, prior commitments, depletion, learning, reputation, trauma, trust, sunk costs, and institutional history all function like structured carryover. Current action is not merely caused by an immediate predecessor; it is shaped by distributed temporal dependence. That makes Auto-Regressive-style thinking a useful analogy. Prior states do not necessarily determine the present in a crude all-or-nothing way, but they do leave a decaying or persistent imprint. Some effects taper quickly, others slowly. Some shocks wash out, others become embedded. Some processes are mean-reverting, others drift. Human agency often unfolds within such temporally layered dynamics. To force all this into the question “could the agent have done otherwise at time t?” is to miss much of the real structure.

Non-stationarity is another powerful concept here. A person is not a fixed system generating choices from a stable underlying distribution. Preferences shift, capacities develop or erode, environments change, meanings change, institutions change, and interpretations of prior events change. The “same person” at one point in life is often operating under different conditions, with different sensitivities and different feasible sets, than at another point. If so, then a lot of philosophical debate about “what this agent would do” quietly assumes a stability that human life often lacks. A Wittgensteinian would not necessarily talk in statistical language, but the spirit is compatible: do not assume conceptual uniformity where actual practice reveals heterogeneity and change. The meaning of agency language partly depends on the forms of life in which it is applied, and those forms are dynamic.

Non-ergodicity is especially relevant. The free-will debate often assumes that what is true “on average” about agents can stand in for what matters in lived trajectories. But many human processes are path dependent in the sense that the sequence matters, not just the aggregate. Early luck, early deprivation, early institutional contact, early trust or mistrust, early educational tracking, early social labeling — these can permanently alter later possibilities. Two people exposed to the same general rules may end up in radically different states because their trajectories diverged early and the divergence compounded. That means agency is not well captured by static or population-level abstractions alone. It has to be thought through in relation to histories. A decision is often intelligible only in light of the path by which the decision-maker arrived there. And once one sees this, the old binary of “free” versus “determined” starts to look hopelessly underdescriptive.

Irreversibility matters for similar reasons. Many actions are not just nodes in a reversible chain. They close doors, create obligations, alter reputations, produce institutional records, change identity, trigger biological or psychological processes, and reorganize future options. A person who commits a crime, has a child, becomes addicted, leaves a job, migrates, enters debt, joins a movement, or undergoes incarceration is not simply making one more selection from a menu. They may be crossing a threshold into a new state space. The future now has a different structure.

This is one reason the free-will debate can be so misleading. It asks whether the act was freely chosen, but often the more important fact is that the act was state-transforming. Its significance lies not only in what preceded it but in how it reorganizes what comes after. A richer discussion of agency needs concepts for lock-in, irreversibility, hysteresis, and regime change, not just concepts for isolated willing.

Wittgenstein opens a space for more rigorous discussion of agency. In fact, one way to put the point is this: Wittgenstein does not give us a substantive systems theory of action, but he clears away a conceptual blockage that makes systems-level thinking harder. He does not supply the mathematics of feedback, lag structure, nonlinearity, or path dependence. What he does is remove the bewitchment that keeps us demanding that every discussion of agency first answer a malformed metaphysical question. That is the enabling role. He frees the discussion from false necessities.

Without that pressure, “agency” can be analyzed more carefully across several dimensions.

First, agency can be treated as temporally extended rather than instantaneous. A person’s agency is not exhausted by what happens in one moment of selection. It includes self-formation, habit cultivation, strategic delay, anticipation, revision, and the management of future option sets. Second, agency can be treated as environmentally scaffolded rather than purely interior. It depends on tools, institutions, norms, interfaces, resources, peers, and constraints. This does not erase agency; it locates it in a richer ecology. Third, agency can be treated as graded rather than all-or-nothing. There are degrees of control, degrees of self-knowledge, degrees of coercion, degrees of reversibility, degrees of responsiveness to reasons. Fourth, agency can be treated as dynamically interactive rather than merely reactive. People do not just respond to causes; they alter environments that later alter them. They create routines, institutions, commitments, and feedback loops that shape later conduct. Fifth, agency can be treated as historically conditioned without being reduced away. A person can be deeply shaped by prior states and still be a meaningful locus of evaluation, planning, and intervention.

That last point is crucial. One fear people have is that once you bring in systems language, time dependence, or structural causation, agency disappears into mechanism. But the better conclusion is the opposite. A richer causal vocabulary often helps us identify where agency actually resides. Not in some magical gap in the chain, but in capacities for regulation, anticipation, adaptation, self-modification, participation in norms, and sensitivity to reasons within a structured process. In that sense, we are after a more rigorous concept of agency, not a weaker one.

The old metaphysical debate often makes “agency” mysterious by asking it to be impossible. It must either float above causality or be dismissed as an illusion. But once that demand is dropped, agency can be studied as a real phenomenon at the level where persons deliberate, learn, commit, interpret, coordinate, and reshape their situations. This also helps explain why reduction to a “single source in a causal chain” is often so distorting. That picture encourages us to think that explanation must terminate in one privileged level. But many systems are multiply determined, recursively organized, and level-dependent in their explanations. In such cases, it is not intellectually superior to say “really, it all reduces to X.” Often that is just a category mistake. The relevant explanatory question is: what level of description captures the stable patterns, intervention points, and meaningful distinctions for this phenomenon?

For agency, the answer is rarely microphysics alone, but neither is it a supernatural faculty. It is a middle space: persons in contexts, acting through time, under constraints, within institutions, through learned repertoires, across feedback loops. That is why decoupling these concepts is so important. It lets us say, for example, that predictability is not the same as coercion, causal dependence is not the same as lack of agency, constraint is not the same as inevitability, path dependence is not the same as fatalism, statistical regularity is not the same as personal irrelevance, and physical realization is not the same as explanatory reduction. Those distinctions are exactly what the free-will debate often blurs.

A Wittgensteinian approach is helpful because it suspects in advance that many grand philosophical disputes persist by illicitly merging distinct uses and contexts. Once those are separated, the apparent necessity of the big dispute weakens. Then other disciplines — economics, psychology, control theory, systems theory, sociology, decision theory, time-series thinking, legal theory — can enter without being forced to answer a metaphysical riddle they were never designed to solve. Wittgenstein does not directly provide a theory of dynamic agency, but his critique of philosophical confusion makes such a theory easier to articulate. By dissolving the false binary between “free” and “determined,” he allows the discussion to shift toward the real grammar of action: temporality, feedback, path dependence, constraint, adaptation, irreversibility, and structured opportunity. This does not weaken agency; it makes the concept more precise, more rigorous, and more useful.

And there is one further payoff. This richer account is not just descriptively better. It is also practically better. Once agency is understood through dynamic structure rather than metaphysical isolation, we can ask better intervention questions:

• How do we widen future option sets?
• How do we interrupt destructive loops?
• How do we create positive feedback for good habits?
• How do we reduce lock-in to harmful trajectories?
• How do we design institutions that increase reflective control?
• How do we distinguish reversible from irreversible choices?
• How do we understand responsibility in systems that both shape and are shaped by agents?

Enhancing Discussions about Human Agency

A lot of the classical free will debate operates with a strangely thinned-out picture of agency. It treats an agent as though the essence of agency lies in one special feature: a mysterious power to originate a choice independently of causes, or alternatively, a point in the causal chain where “the real decision” occurs. But that is already a distortion. It strips away most of what actually makes an agent an agent.

An agent is not just a thing that “chooses” at a moment. An agent is a system with capacities: to perceive, model, anticipate, interpret, remember, compare, simulate possible futures, monitor error, revise strategy, infer norms, regulate impulses, form commitments, coordinate with others, and adapt over time. Agency is composed out of these capacities and their organization. That matters because once agency is understood compositionally rather than metaphysically, the free will debate starts to look badly framed. The question is no longer “is there an uncaused chooser inside the person?” but “what capacities constitute agency, how are they organized, how do they function, and under what conditions do they operate well or badly?”

One of the key mistakes in the standard debate is that it treats agency as if it had to be simple. Either the act is “up to the agent” in some pure and ultimate sense, or it is not. But actual agency is layered. At one level, there is perception: what the person notices, fails to notice, misreads, or attends to. At another, there is interpretation: how the person understands the situation, what they take to be salient, threatening, shameful, rewarding, possible, forbidden, expected, or meaningful. At another, there is simulation: the ability to imagine consequences, rehearse scenarios, compare options, estimate risks, and project oneself into future states. At another, there is self-modeling: the ability to know one’s own tendencies, limitations, temptations, likely mistakes, strengths, and vulnerabilities. At another, there is regulation: delaying response, suppressing impulses, reframing goals, shifting attention, and aligning short-term behavior with long-term plans. At another, there is social navigation: reading other people, anticipating reactions, recognizing norms, maintaining reputation, coordinating with allies, and adjusting to institutions. At another, there is temporal integration: connecting present action to past commitments and future trajectories. These are not peripheral to agency. They are agency.

And once this becomes clear, a lot of free will language starts to look oddly primitive. The debate often asks whether the person “could have done otherwise” as if the core of agency were exhausted by branching at an instant. But many of the capacities that matter most are not about branching in a moment. They are about building and managing the conditions under which later choices become more intelligent, more coherent, and more effective. A serious account of agency must include the fact that agents are not passive endpoints of causal streams. They are organisms that form representations, however imperfectly, of themselves and their surroundings. They forecast. They update. They learn. They anticipate consequences. They navigate under uncertainty. They use models of the world to guide action.

Even when those models are incomplete or biased, the modeling itself is central. It means that human action is not well described either as pure spontaneity or as mechanical output. It is mediated by cognition, interpretation, memory, expectation, and strategy. This is one place where the standard free will debate often misfires. It tends to imagine two options: either the agent is some transcendent originator outside causal order, or the agent is just a passive product of prior forces. But an agent with predictive, interpretive, and self-regulating capacities fits neither picture. Such an agent is causally embedded, yes, but also actively engaged in shaping behavior through internal models, feedback, and future-oriented control. There is no need for magic, but there is also no reason to collapse the person into a mere billiard ball.

This helps clarify a deeper point: agency is not the absence of causation. Agency is a distinctive form of organized causation. That is, an agent is a being whose behavior is guided by representations, evaluations, goals, norms, and anticipations. The person is not outside the world’s causal structure; rather, the person is a particular kind of causally efficacious organization within it. One that tracks information, processes significance, and modifies conduct in light of possible futures. That is why the old opposition between “caused” and “agentic” is so misleading. Being caused does not negate agency, because agency itself is a mode of causal organization. What matters is not whether the person has causes, but what kind of system the person is.

Agents also do not operate in a vacuum. Human agency is socially and institutionally situated/embedded from the start. A person’s capacities for interpretation, prediction, self-understanding, and choice are all shaped by language, upbringing, roles, institutions, technologies, incentives, norms, and network position. We do not first become fully formed isolated choosers and then enter society. We become agents through socialization, participation, and exposure to structured environments.

Language itself is central here, and that is one reason this fits Wittgenstein so well. Our ability to interpret the world, describe reasons, follow rules, understand expectations, and form self-conceptions is inseparable from social practices. Human agency is not merely housed in an individual organism; it is partly constituted through shared forms of life. The very concepts through which we deliberate — duty, risk, success, shame, fairness, betrayal, prudence, status, loyalty — are socially inherited and socially sustained.

Agents are embedded in social networks, institutions, and environments; that is not just an external add-on but part of what agency is. The individual agent interacts with other agents who have their own models and expectations, institutions that reward, punish, constrain, classify, and channel action, norms that shape what counts as reasonable or deviant, technologies that alter attention and available options, environments that present friction, affordances, and hazards, and historical trajectories that determine which opportunities are live. This makes agency relational and ecological as well as cognitive.

That is a major corrective to the standard debate. The free will discussion often imagines agency as a mysterious inward property. But in practice agency depends on external supports and contexts. Someone with education, time, legal protection, social trust, stable housing, and access to information can exercise certain agentic capacities much more effectively than someone under chronic stress, surveillance, precarity, threat, or disorganization. It is not that one has “real free will” and the other does not. It is that agency is scaffolded, hindered, amplified, and damaged by conditions. That point has big consequences. It means we should often ask not “does this person have free will?” but:

• How well can this person model consequences?
• What options can they actually perceive?
• How distorted is the informational environment?
• How much time pressure or stress are they under?
• What skills of self-regulation do they possess?
• What institutional constraints are acting on them?
• What feedback do they receive from the environment?
• How reversible are their choices?
• How much slack or margin do they have?
• What social supports help them correct errors?

Those are all agency questions. In many cases they are better agency questions than the old metaphysical ones. It also means that agency is developmental. It is not simply given. People can become more agentic or less agentic in particular respects. They can learn to forecast better, regulate better, interpret social signals better, manage incentives better, understand themselves better, and structure their environment better. Or they can deteriorate in those capacities through trauma, addiction, exhaustion, manipulation, or institutional degradation.

This again makes the classical free will framework look too static. It imagines a fixed power of will that one either has or does not have. But actual agency varies across time, domain, and context. Someone may be highly agentic in long-range planning but weak in emotional regulation. Strong in professional strategy but poor in intimate relationships. Excellent at modeling institutions but bad at modeling themselves. This variability is exactly what one would expect if agency is composed of multiple capacities rather than a single metaphysical faculty.

And that composition matters morally and politically.

If agency is composite and embedded, then improving agency is not merely a matter of moral exhortation. It can involve education, better information, better institutions, better tools for self-monitoring, better social conditions, better defaults, more stable environments, and more room for reflection. In other words, a serious account of agency does not just evaluate people; it asks what conditions foster better agency. The metaphysical debate prevents us from doing practical analysis. Instead of debating about transcendent causation, we can ask:

• Did they understand the situation?
• Could they imagine alternatives?
• Did they accurately model consequences?
• Were they able to regulate themselves?
• What norms were operative?
• What institutional structures shaped the field?
• How much of the environment was legible to them?
• What social feedback were they responding to?
• What kind of self-model did they have?
• How did prior decisions structure this moment?

That is a far more rigorous vocabulary.

It also gives a better picture of responsibility. Responsibility need not depend on a mythical power of absolute origination. It can depend on the extent to which a person is functioning as a reflective, interpretive, self-regulating participant in shared normative life. That is already a rich standard. It allows for gradations, impairments, and contextual nuance without erasing accountability.

This is very much in line with a Wittgensteinian spirit because it resists the temptation to look for the hidden essence of “free choice” behind all our ordinary practices. Instead it examines the actual grammar of our agency-talk: intending, deciding, hesitating, regretting, deliberating, being pressured, acting knowingly, acting recklessly, being responsible, being manipulated, being competent, being in control. Those expressions belong to a complicated human practice, not to a single metaphysical problem.

The old debate is extremely unsatisfying. It treats the most impoverished notion of agency as foundational. It forgets that agency already includes world-modeling, self-modeling, interpretation, simulation, regulation, anticipation, and social embedding. Once those are restored, the philosophical puzzle shifts. It is no longer about locating an impossible pure chooser. It is about understanding how these capacities hang together and how they interact with environments and institutions.

The free will debate often assumes an artificially thin notion of agency, as though agency were exhausted by a moment of choosing. But real agency is a composite capacity involving prediction, interpretation, self-modeling, simulation, regulation, and social coordination. Because agents are embedded in networks, institutions, and environments, agency is not an isolated metaphysical spark but a dynamic, situated mode of intelligent participation in the world.

Revisiting Kevin Mitchells "Free Agents"

Mitchell’s argument in Free Agents is mainly naturalistic and biological: he tries to show how organisms capable of modeling, predicting, simulating, introspecting, and shaping behavior could have emerged through evolution. The book’s core framing is that nervous systems gave animals the capacity to learn about the world and to “model, predict, and simulate,” and that these capacities reach a high point in humans through imagination, introspection, and in-the-moment reasoning. Mitchell appears to be doing, at the empirical level, what the Wittgensteinian reframing is doing at the conceptual level. He is trying to replace the thin picture of agency — the picture of a passive machine or a single metaphysical “will-event” — with a richer picture in which agency is composed of capacities like prediction, simulation, metacognition, and purposive control. His own public summaries of the book emphasize exactly those capacities, along with the idea that this matters for how agency can be “enhanced or infringed.”

He still applies the label "Free Will". In my initial blogpost, I hevaily referenced his work in favor of my seemingly self-evident position that we do indeed have free will. After considering the Wittgensteinian approach, I think it's best to drop that label all together; based on this enriched view of agency and the naturalistic account of its origins given my Mitchell. Mitchell is not primarily defending the folk-metaphysical fantasy that a person escapes causation. He is trying to defend the reality of agents as biologically evolved systems that evaluate situations, simulate alternatives, and guide action purposefully. That is much closer to the richer conception of agency we have been developing than to the cartoon free-will debate.

That becomes especially clear in the way the book is presented. The book’s summaries describe it as a story of how “living beings capable of choice emerged from lifeless matter,” with nervous systems providing capacities to learn, model, predict, and simulate, culminating in human introspection and reasoned action. In a podcast summary of the book, Mitchell’s themes are listed as “informational causation,” “the evolution of decision making,” “agency,” “simulating possible courses of action,” “decision making as a process,” and “constraints and top-down causation.” This is a far better approach to understanding agency than the traditional dichotomy. He is essentially articulating a rich conception of agency under the traditional label; I say let's drop the label.

A Wittgensteinian would say that much philosophical trouble begins when one word — “free will” — absorbs too many different questions. Mitchell’s book appears, in practice, to break that bundle apart even if he keeps the old label. He talks not only about willing, but about the evolution of nervous systems, informational causation, simulation of possibilities, decision-making as a process, constraints, top-down causation, and moral responsibility. That is already a kind of conceptual decoupling I've been referring to above.

Instead of treating agency as a single metaphysical property, he seems to be reconstructing it as an evolved organization of functions and capacities. That is very close to what I have been saying: agency is not exhausted by a mysterious moment of pure choosing. It is composed of predictive, interpretive, simulative, self-modeling, and socially embedded processes.

A number of responses to Mitchell appear to pull him back into the very debate he is trying to escape. For example, a review in Undark frames the issue through the standard opposition between determinism, compatibilism, and illusionism, and notes that Mitchell distinguishes physical predeterminism, causal determinism, and biological determinism while still rejecting compatibilism. Another critical response argues that Mitchell’s scientific picture really supports compatibilism rather than the stronger anti-determinist position he wants. From a Wittgensteinian angle, many of these critiques risk missing the most interesting part of Mitchell’s project. They often seem to hear the phrase “free will,” then immediately slot the book into the inherited metaphysical battlefield: Is he a libertarian? Is he a compatibilist? Does indeterminacy help? Can top-down causation beat determinism? Those are not illegitimate questions, but they may force his argument back into a conceptual arena whose terms are already distorting. If what Mitchell is really giving us is a biologically enriched conception of agency, then a critique that insists on translating everything back into the old metaphysics may be reenacting the very confusion Wittgenstein warned about. In other words, the critique can miss the re-description because it is policing the label.

Using the label increases the chances of falling into the traps envisaged by the Wittgensteinian critique of philosophical language. The trap is this: they hear “free will,” assume the old grammar, and then judge the book according to whether it solves the inherited pseudo-problem, instead of asking whether it has redescribed the terrain in a more fruitful way.

Mitchell’s strongest insights are not in the metaphysical residue attached to the term “free will.” They are in the detailed account of what agents are and how evolved organisms come to possess forms of purposive, model-based, self-modifying control. The book is persuasive precisely because it sounds least like the classical free-will debates. This is why his book resonates with me so much. The themes I have been discussing are implied by his book:

• the move from isolated decisions to decision processes,
• from static causation to dynamic interaction,
• from reduction to organization,
• from single moments to histories,
• from bare choice to opportunity structures,
• from metaphysical freedom to real capacities for modeling and control.

Mitchell should be read less as proving an old metaphysical thesis called free will and more as redescribing human and animal agency in biologically serious terms. The important achievement is not showing that agents transcend causation, but showing that organisms can be genuine agents because they are organized systems that sense, learn, predict, simulate, evaluate, and regulate action in context.

Once “free will” is no longer doing all the work, Mitchell’s themes become easier to connect with systems language: feedback, process, lagged effects, self-modeling, environment-modeling, multi-level causation, path dependence, and the ways agency can be enhanced or infringed by institutional structure. His own summaries explicitly connect the book to how agency may be “enhanced or infringed” and to collective agency in global crises. That is already much closer to our discussion of opportunity sets, constraints, incentives, and environments than to the old “could I have done otherwise in the exact same universe-state?” puzzle.

Conclusions

I think we need to enhance the naturalized concept of agency by moving it away from crude reductionist mechanism and simplistic anti-agency conclusions, without invoking incoherent metaphysical and supernatural doctrines. We need to outgrow "ghost in the machine" explanations of agency. This debate in my view, is simply a non-starter that doesn't give us a richer description and understanding of human agency. The Wittgensteinian analysis shows us that our concepts of human agency can be enriched, and I think Mitchell's empirical work is a solid basis on which to build.