Free Will Is Real (with Kevin Mitchell)
What this covers
Kevin Mitchell and Russ Roberts work through the foundations of human agency in a conversation that reframes rather than resolves the free-will debate. Mitchell, a neuroscientist, argues that the entire dispute has been built on a false premise: that physics dictates determinism. He contends instead that physics contains genuine indeterminacy at both quantum and classical levels, and that organisms use their own organization to constrain possibilities and act for their own purposes. The discussion moves across evolutionary biology, neuroscience, and philosophy, grounding abstract metaphysics in how organisms actually behave—from bacteria making all-things-considered judgments to humans exercising executive function and meta-volition.
The conversation spans several distinct intellectual territories. Mitchell traces As-If Agency to Real Agency, beginning with the purposive behavior of simple organisms and culminating in human metacognition. He explains how informational causation emerged with life itself, introducing meaning and purpose absent from non-living systems. On constraints, Mitchell inverts a common objection: rather than negating freedom, Chosen Constraints are exercises of agency. The pair examine how character is built through a lifetime of active choices rather than merely inherited, and how emotions like regret, shame, and guilt operate both backward-lookingly and as forward-guiding parameters in decision-making. They also consider what neuroscience appears to show about decisions forming before conscious awareness, and why that finding does not entail that you do not decide—only that you decide using your brain rather than apart from it. Throughout, Mitchell challenges the Ghost in the Machine framing that opposes mind to body, and defends Self as Distributed Entity persisting through time as pattern rather than as static physical stuff.
Mitchell argues that free will is real but the entire free-will-versus-determinism debate rests on a false premise: physics is indeterministic, not deterministic, and agency is an evolved biological capacity by which organisms use their own organization to constrain possibilities and act for their own purposes.
- Physics does not entail determinism; there is genuine indeterminacy at both quantum and classical levels, opening causal slack for organisms to exert macroscopic control.
- Agency evolved gradually from 'as if' purposiveness in simple organisms to genuine, metacognitive agency in humans.
- Constraints (history, habits, biology) do not negate free will; an unconstrained self would be a meaningless random behavior generator.
Free will is an evolved capacity to control impulses, resolve conflicts, and deliberate over time.
- What philosophers call free will in humans may be essentially equivalent to what psychologists call executive function—the ability to control impulses, hold multiple things in mind, switch tasks as conditions change, and resolve conflicting goals over multiple timescales—and these are partly innate traits but also learned skills that some people possess and exercise more than others.
“what we call free will in humans on the philosophical front is basically equivalent to what psychologists call executive function, which is the ability to control impulses, to operate with multiple things in your mind at the same time, to keep something in working memory, to switch between tasks if conditions change or payoffs change, and to resolve conflicting goals over multiple timescales.”
- Rather than choosing between unconstrained metaphysical free will or robotic determinism, free will should be understood as an evolved suite of biological capacities that people possess to varying extents, exercise to varying extents, and that can be constrained to varying extents by circumstances—an understanding the law already reflects in recognizing diminished responsibility for mental illness, addiction, intoxication, or childhood.
“free will is the term that we use for an evolved suite of biological capacities that people may have to varying extents, that they exercise to varying extents, that may be constrained to varying extents in different circumstances and conditions.”
- Humans' cognitive superpower is an open-ended future: unlike other animals we can think days, millennia, or beyond our own lifetimes ahead, so far-future goals become part of our decision-making calculus, enabling us to weigh short-term hedonic drives against long-term conflicting goals like health.
“human's cognitive superpower is that the future for us is open-ended... We can think millennia in the future if we want to, even after our own lifetimes. So, that ability means that what--now, far future goals become part of the calculus of our decision-making.”
- In genuinely conflicted decisions where you have good reasons for both options, the outcome is open at the start and decided only at the end; you are not running parameters through a preset algorithm with fixed weights—rather, the act of deciding is the process of figuring out what those weights should be, which on the psychological, neural, and philosophical levels just is the agent deciding.
“it's just not the case that you're running the parameters of a scenario through a preset algorithm where the weights of these different factors are already set. What you're doing in deciding is figuring out what those weights should be.”
- Executive function and meta-volition are not faculties that emerge from nowhere but learned skills: child-rearing deliberately inculcates self-control, delayed gratification, and pro-social behavior through teaching, and over a lifetime these admonitions become internalized into one's cognitive machinery as a conscience, ultimately manifesting as morality as an emergent suite of psychological faculties on which civilization depends.
“Executive function is not just a bunch of faculties that emerge from nowhere. They are skills that we learn... we go to quite some effort to tell them, 'No, you can't have everything you want the second that you want it.'”
- The determinist claim that intent simply pops into your mind and you merely carry it out is wrong; the evidence shows you are actively involved in making the decision, and the intent is the endpoint of a deliberative process you are engaged in rather than something that just appears.
“There's lots of evidence that the intent doesn't just pop into your mind. That, in fact, you are actively involved in making the decision about what to do. The intent is the end point of that product--sorry, of that process--that you are actively engaged in.”
Life introduced informational causation and purposive agency absent from non-living matter.
- When a receptor binds a food molecule, there is no transfer of energy or momentum pushing anything around—only an informational change in the protein's conformation that acts as a signal—so life introduced a kind of informational causation, along with purpose, meaning, and value, that does not exist in any non-living part of the universe.
“When a receptor binds a food molecule like a sugar on the outside of the cell, there isn't any transfer of energy or momentum that's pushing anything around. There's just an informational change... we have a kind of causation that, again, is not at play in any non-living part of the universe.”
- Even a bacterium moving toward food is acting for a reason relative to its purpose of persisting; the purpose is real and indispensable for understanding the organism's configuration, even though there is no awareness—this 'as if' agency is genuinely purposive and baked in by evolution.
“even, like, a bacterium that has a tendency to move towards a food source, for example, is acting for a reason relative to its purpose, and its purpose is to persist. And, it's not aware of that.”
- Life is not a state but a process: you are not just alive, you are living, an activity each cell performs; you are not the physical matter making up your body now but that pattern persisting through time as atoms and molecules are continually exchanged with the outside world, like a storm or flame that is the ongoing process organizing molecules into a higher-order pattern—except that life, unlike a flame, goes on.
“Life is not a state; it is a process. You are not just alive, you are living... You are more than the pattern of physical matter that makes up your body right now: you are that pattern persisting through time.”
- Life arose because self-reinforcing patterns of chemical reactions—interlocking cycles where each reaction's product is the next's substrate—are better at persisting than unstable ones, so by a tautological dynamic stable patterns persist and become the substrate for further evolution, transitioning from geochemistry to biochemistry to membrane-bound organisms doing thermodynamic work to maintain themselves.
“It's a tautological kind of dynamic where things that are good at persisting tend to persist. They are then the substrate for further evolution, and so on.”
- Before life, the universe contained only happenings, not doings; life introduced entities that are causal agents acting in the world, and this agency—the fact that living things do things—is the core element of life, a feature often overlooked in biology.
“before life emerged, there weren't any doings in the universe. There were just happenings. Things were happening. And then, at some point, these entities emerged that seemed to be able to do things.”
The self is a real, causally efficacious continuity of pattern and commitments through time.
- The Buddhist and Humean observation that you cannot find the self anywhere inside the brain is true but trades on an equivocation in the word 'illusion': it shows the self is not what naive folk psychology thinks it is, not that the self does not exist at all—just as gravity is not what we think yet remains real—so the self is a real, causally efficacious bundle of memories, interests, goals, and commitments persisting through time.
“there's a kind of a sleight of raising here that happens when people say, 'The self is an illusion,' because illusion can be taken in two different ways... Gravity is not what we think it is, but I'm not going to jump out the window.”
- Being constrained to act like a human by millions of years of evolution and to act like yourself by your entire past is not a limitation on selfhood but precisely what it means to be you; acting badly out of character is often treated as a pathology or sign of mental illness, confirming that consistency with one's biography and psychology is the norm of selfhood.
“I'm constrained to act like a human being by millions of years of evolution. And I'm constrained to act like me by all of my past. And, me acting like me is just what it means to be me. There isn't anything else.”
- The self can be understood as a centralized locus of control that must be centralized because we can productively do only one thing at a time, requiring an all-things-considered judgment that competitively executes one option—and just as a corporation like Coca-Cola is a real holistic entity exerting control in the marketplace despite no single internal part being its essence, the self is a real holistic entity not reducible to any internal locus.
“the self is present in the way that there's a sort of a centralized locus of control... it has to be centralized, because we can only do one thing productively at a time; and so we need a system that actually can make an all-things-considered judgment”
- Neuroscientific evidence that brain activity accompanies decisions can be read two ways: either your brain makes the decision and not you, or you make the decision using your brain—and the latter is correct, since the framing of a 'you' separate from the brain machinery is a confused ghost-in-the-machine notion.
“One is to say: Clearly, your brain is making the decision, not you. And the other is to say you are making the decision using your brain. Like, what else would you use?”
- The self is not the physical organization of a thing at a single moment—that is just physical stuff—but rather the continuity of that pattern through time, including all biographical information, knowledge, habits, and committed future projects; thus the self has temporal parts just as it has spatial parts, extending into past and future.
“The self is the continuity of that pattern through time. And, for us to be a self means not just maintaining this pattern of physical processes, it means maintaining our psychological selves, our biographical information, all of the knowledge that we've learned, all of the habits that we've picked up, plus all of the projects that we've committed to that extend into the future.”
The absolutist free-will conception is self-defeating and not what agency actually means.
- The core arguments on both sides of the free-will debate predate modern science—William James gave his answer in 1888 and Epicurus arguably had it right 2,500 years ago—so the determinists were arguing for determinism long before science existed, and arguably little fundamental progress has been made since.
“to be honest, I think Epicurus had it right, and that was 2500 years ago.”
- Patients with neurological damage, and even healthy people in contrived experiments, will make up post-hoc stories to explain behavior they performed for reasons they cannot consciously access; the determinist extreme overextrapolates from this to claim that all human reason-giving is confabulation.
“they make up a story. It's called confabulating... they make up a story post hoc. So, this is an example where people have, I think, overextrapolated from that kind of data.”
- A recurring methodological error in the free-will literature is extrapolating from what happens in one kind of scenario (e.g., subliminal experiments or neurological patients confabulating) to every kind of scenario, when in fact intent does sometimes just appear but usually does not.
“one of the key things here is you shouldn't extrapolate from what happens in one kind of scenario to every kind of scenario. And I think that's a mistake that people have made across this literature, generally.”
- Major physical theories like Laplace's deterministic mechanics and Einstein's block universe contain no definition of the present moment—nothing distinguishes present from past and future—which is a striking wrinkle given how central the present is to human experience.
“he has this idea of a block universe where it's very much the same ideas: that everything is sort of fixed, and there's no--there's actually no definition of the present moment within that. Nothing distinguishes the present from the past and the future.”
- The free-will debate wrongly assumes determinism is dictated by physics; in fact physics does not say the world is deterministic—there is inherent indeterminacy at the quantum level (well known) and also at the classical level (less well known), so the real challenge is explaining where organismal control comes from, not where freedom comes from.
“physics doesn't say at all that the world is like that. So it's really weird that the entire debate in free will--when you're talking about free will, people just assume that it's free will versus determinism. In fact, you can make a very good argument from physics that there's inherent indeterminacy in the way that things evolve in physical systems.”
- The absolutist conception of free will—acting with no constraint, no reason, no history, no memory, no learning—is self-defeating because every such feature is a constraint, so under it the self evaporates and all that remains is a random behavior generator; therefore that conception is absurd and not what free will means.
“where I can act for no constraint--which means for no reason, with no history, I'm not allowed to have any memory of anything... it just becomes a meaningless thing where actually what happens in that scenario is the self evaporates. There's nothing left. There's just a random behavior generator.”
Organisms including bacteria exhibit holistic agency optimizing multiple goals simultaneously.
- A bacterium does not always move toward food: in naturalistic conditions it integrates many factors—pH, metabolic state, whether it is dividing—to make an all-things-considered judgment, so zooming out reveals a proactive holistic agent optimizing over multiple conflicting goals and timescales rather than a mechanism pushed around by its parts, directly analogous to bounded rationality in economics.
“when you take a more wholistic view, what you get, rather than this mechanistic, reductive thing being pushed around by its parts, is you get an organism that's encountering the world in a proactive way... it kind of makes an all-things-considered judgment that's optimal given multiple goals and multiple parameters that it's trying to satisfy.”
- If you have no agency or autonomy at all, you are not a person—personhood is defined by having enough autonomy to be afforded its dignity, and it means nothing to be a self without some autonomous control.
“if you think that actually you have no agency and autonomy at all, you are not a human being. You're not a person. I mean, the definition of a person is someone who has enough autonomy to be afforded the dignity of being thought of with personhood.”
- When you genuinely have no preference between options but must decide—as when a waiter and others are waiting—there is good evidence that the brain can use the noisiness of its own neural processes as a resource to break the tie.
“there's very good evidence that we can use kind of some of the noisiness of our mental neural processes as a resource to break a tie there, for example, so when you don't know.”
- When the future is genuinely open and multiple things could happen, the way a system is organized can do causal work by constraining which trajectories are possible at a macroscopic level—just as a computer constrains electron paths via hardware and software without violating physics or micromanaging individual quantum events—because the underlying physics is probabilistic and not causally complete, leaving causal slack for higher-level control.
“the computer I'm using right now is constraining the paths and trajectories of electrons within it by virtue of the hardware being organized a certain way, and by virtue of the software that is running. And it does that without violating the laws of physics.”
- Nervous systems can learn from a single experience and generalize it across an entire category—being bitten by one dog makes you wary of all dogs regardless of breed or color—building a complex model of the world populated by categories and causal relations, especially about what you can do and what things can do to you; this generalization is powerful but also produces bias and error.
“If you've ever been bitten by a dog, you can be shy of all dogs in the future because they're the type of thing that can bite you.”
Character emerges through lifetime choices about the basis of choices, not mere innate traits.
- Personality traits like conscientiousness or impulsivity are innate predispositions providing only a baseline tuning, but character is not determined by them; rather, character emerges through a lifetime trajectory of active choices we make—including choices about the basis on which we make choices—so we are co-authors of our character rather than passive recipients of things that merely happen to us.
“There are predispositions, but they're just a baseline. They're just a kind of tuning of some internal parameters that inform how we meet the world”
- Meta-awareness of how one's own mind is configured is genuinely empowering: stepping back from the emotionality of a situation lets one respond rather than merely react, which is the aim of much psychotherapy where learning to notice reactivity allows one to change it; the ancient admonition to know thyself thus confers real power, even if it cannot be wielded at every moment.
“one of the elements of psychotherapy can be: becoming aware that you are reacting to things instead of responding to them... you can learn certain ways to not be like that.”
- Regret is not merely backward-looking; it operates anticipatorily as a parameter in decision-making, since having regretted past actions we generalize to forecast that doing X now would bring consequences—social judgment or self-disappointment—we would regret, and this forward-looking regret informed by others' judgments is crucial to guiding behavior and building character.
“regret is actually guiding our actions all the time, but in an anticipatory way... 'If I did X right now, these would be the consequences. And, some of them I might like, but some of them I would regret because other people would judge me in this way. Or I might even judge myself.'”
- The popular question of what advice you would give your younger self has it backwards: in reality our past selves are continually giving advice to our present self to secure the best possible future—this is what learning is, knowledge from the past extending into the future.
“In reality, what happens is precisely the opposite: our past selves are giving advice to our present self all the time to ensure it has the best possible future.”
Social emotions and moral systems evolved to reinforce pro-social behavior in communities.
- One cannot draw sweeping conclusions about how society, the legal system, or interpersonal norms should be structured from absolutist metaphysical claims about free will; such metaphysics can inform but not dictate, and neuroscientists arrogantly overreach if they assume lawyers, sociologists, and moral philosophers were waiting for them to answer these complex questions.
“you can't really draw sweeping kinds of conclusions about the way society should be structured... based on these metaphysical claims, or arguments, or conclusions. But, they can inform them.”
- Social emotions like shame, guilt, praise, and admiration arise in response to actions judged through the lens of whether they are pro-social, and a species that relies on pro-sociality would evolve such emotions because they functionally keep members acting beneficially, enabling survival in communities; morality is naturally understood as a system reinforcing pro-social communal behavior that gets internalized.
“These are social emotions that arise in response to people's actions, judged through the lens of whether they're pro-social or not... a species that relies on pro-sociality would have these emotions. They have a functional role to play in making sure that we keep each other acting in the way that's most beneficial for everybody.”
Habits efficiently encode past learning and free cognitive resources for novel deliberation.
- Habits get a bad rap as mere reductions of freedom, but they are tremendously useful shortcuts: a habit encodes the conclusion that in a certain familiar situation a particular action is good, so you avoid spending precious cognitive resources and time re-deliberating something you have already thought through many times—an extremely efficient way to exploit learning from experience.
“habits get a bad rap because there's a sense in which having a habit is just reducing your freedom. And, in one sense, that can be true, but in another sense, it's really good.”