Thanks for reading! The best way to make sure that you don’t miss any of these conversations is to subscribe to The Good Fight on your favorite podcast app.
If you are already a paying subscriber to Persuasion or Yascha Mounk’s Substack, this will give you ad-free access to the full conversation with Kevin, plus all full episodes and bonus episodes we have in the works! If you aren’t, you can set up the free, limited version of the feed—or, better still, support the podcast by becoming a subscriber today!
And if you are having a problem setting up the full podcast feed on a third-party app, please email our podcast team at leonora.barclay@persuasion.community
Kevin Mitchell is Associate Professor of Genetics and Neuroscience at Trinity College Dublin. He studies the interplay between genes, brains, and minds. He is the author of Innate: How the Wiring of Our Brains Shapes Who We Are and Free Agents: How Evolution Gave Us Free Will.
In this week’s conversation, Yascha Mounk and Kevin Mitchell discuss the arguments against free will, if evolution supports free will, and how much control we have over our decisions.
This transcript has been condensed and lightly edited for clarity.
Yascha Mounk: So you have a book which engages with one of the oldest debates in philosophy and in theology, which is whether or not humans have free will. But you claim to be making a novel argument, which is hard to do in a debate that is that old, and in which so much ink has been spilled. Why do you think that to understand the ways in which humans do in fact have free will, we have to think about biological processes like evolution? And how is that different from how we've thought about free will so far?
Kevin Mitchell: If you look at the philosophical or theological literature, there’s a lot of armchair thinking, trying to divine from logical postulates how we could have free will given a particular supposed state of the universe and so on. My own feeling is that we don’t have to think about this issue in these really abstract terms. We can actually get quite concrete. If we’re asking, Do we really make decisions? or Are we in control when we make decisions?—those are actually biological questions. We can get into the neuroscience of decision making, and the biology of control more generally, and explore how these kinds of systems could have evolved.
How could it be that living things can act in the world in ways that non-living physical things can’t? There are some deep metaphysical questions there, but you can get a handle on them by really getting into those biological details and making the discussion a lot more concrete. That’s what I was trying to do.
Mounk: Before we delve into your argument, perhaps you could lay out for listeners the 101 of this conceptual space? What are the main traditions? What kinds of answers do they give to this fundamental question of whether we actually, in a meaningful way, have agency in the world?
Mitchell: First of all, the phenomenon that we observe is that we feel like we make decisions. We're walking around, we're seeing other people make decisions. Not only us, but it also seems that other animals are capable of goal-directed behavior—we can see them figuring something out. They might be conflicted about something and then they decide. So there's a sort of a natural view that we are capable of making decisions.
But then there are these challenges to it. The main one comes from what's called “determinism,” and that has various flavors to it. There's a really hardcore determinism, which basically just says, look, it's all just physics. You're made of atoms and molecules. Atoms and molecules obey the laws of physics. Those laws are deterministic in the sense that it's just the physical state of the system plus the laws of physics which guarantee the next state of the system. There's no leeway, there's no branching paths, there's no open future, it's just one kind of timeline forever.
Mounk: Part of the intuition here is that if the world is purely physical, then the moment that the Big Bang happened—or if you're religious, the moment that God put the world in place—it is actually predetermined what's going to happen today. You might look like you're hesitating about whether to order the salmon or the steak for lunch, but in fact, if you're obeying the laws of physics, there is an objective answer to whether or not you're going to order salmon or steak for lunch before you make that decision. And if that's the case—if somebody with perfect knowledge about the laws of physics and about the particular atoms and neurons in your brain could already predict what’s going to happen—then surely the idea that we're choosing is an illusion. That's the idea here.
Mitchell: That's exactly the framing. And the way that you said it, it's absurd on its face, frankly. There's no evidence for it in physics. But that's a very ancient question. It goes back to the ancient Greeks.The early atomists, people like Democritus, proposed rightly that everything in the world is composed of atoms, things that couldn't be divided any further down. But Democritus had a very deterministic view of how things happen. His idea was that atoms—to use their phrasing—always fall in straight lines. But another Greek philosopher, Epicurus, absolutely realized that if that were true, then everything that you just said would follow from that. Everything would be deterministic. Two things would follow from that.
First of all, all of your actions would be predetermined. But secondly—and this is more subtle—there wouldn't be any room for you to do anything. All of the causes would be at the low level of atoms and molecules interacting with each other and anything else, your desires, your intentions, your thoughts or feelings would just be kind of epiphenomena. They wouldn't actually be causal because all the causes are already exhausted by what's going on at the low level. There's no room in that scenario.
So that's the most extreme version of determinism, but there are a few other varieties. One comes from neuroscience, and basically just says, look, yeah, you're having thoughts and feelings and desires and so on, but how could a thought possibly push physical stuff around in your brain? How could it make neurons fire when a thought is immaterial and neurons firing is a real physical thing, and isn't it just that your brain is doing things? You're not really doing anything. Your brain is in charge. That's a kind of weird way of thinking because it separates you from your brain as if there's some immaterial you that you would want to be in charge as opposed to the idea that actually when you're making a decision and your brain circuits are active, that's you using your brain to make a decision, as opposed to your brain making it.
Then there's another version which says, yes, you can think about things, you can have desires, you can build up intentions of what to do, you can do things for reasons, but those reasons completely derive from your past. They don't involve you in the moment. It's everything that has occurred to you, it's your genetics, it's all of human evolution, it's your experiences. Everything that has configured your brain and your mind—really your psychology up to this very moment—is going to be what determines what intentions you build up when you encounter some new kind of scenario. Again, that kind of view removes you from the equation. You are just not involved in real-time in decision-making. It's a sort of a stimulus/response way of seeing what the brain is doing, which is advanced by people like Robert Sapolsky, for example, who I know you've spoken to before. Of course, there are very strong influences on our behavior from our psychology and predispositions and so on, even from our genetics. But those are influences. They're not completely predetermining everything and they leave lots of scope for decisions to be made by you as a holistic entity.
Mounk: There are, in some ways, an intuitive set of ideas here. We know that socioeconomic class determines a lot of life outcomes. We know that psychological predispositions determine a lot of outcomes. We know that the level of intelligence you happen to be born with determines a lot of life outcomes. And so it's tempting to say, well, perhaps that's just all there is to it. And that's the argument that Sapolsky takes, I think, quite far in his latest book. As listeners who will go back to that episode will see, I'm quite skeptical of that for a number of reasons, as I believe you are. There's a difference between acknowledging the fact that all of those things influence us, that we don't completely self-create without any of those influences, and saying that therefore you're not morally responsible for anything you do.
One of the things that makes me very concerned about accepting that conclusion is that I think from a certain kind of political point of view—or just from a moralistic point of view—it can seem very appealing because suddenly when somebody has done something bad, they're not really guilty. When somebody has committed a crime, we don't really have to hold them responsible. But the thing that falls out of consideration when you talk like that—and I think T.M. Scanlon, who I've also had on the podcast in the past, makes this point very beautifully in his writing—is that you then can't really appreciate people for good things anymore either. It might be tempting from a certain kind of progressive point of view to say, oh, this poor murderer, it's just his terrible childhood and he was born with the wrong genes and his influences were bad.
But if you're then consistent about that, that also means that you can't appreciate your partner for the kind things they do, your parents for the generosity they may have displayed in raising you, your friend for the wonderful virtues or qualities that make you cherish having them in your life. And in fact, in a certain sense, it means you can't hold yourself responsible. You can't aspire to act in the right kind of way. So there are a lot of good reasons to try and resist that kind of determinism.
So what are the main traditions that try to take on the determinist challenge? Let's start with the compatibilist tradition and then go to the libertarian one.
Mitchell: First of all, we have the free will skeptics who just say, because of determinism, we don't have free will. But then, like you say, there are a couple of groups in the philosophical literature who have thought about this problem and defend the idea of free will or at least defend the idea of moral responsibility. Some of them actually do so without defending free will, which is interesting. So compatibilists, as the name suggests, think that either free will itself or at least moral responsibility is compatible with the idea that the world is completely deterministic.
Now, I don't find those arguments compelling at all. I don't really even find them coherent, to be honest, but it's a major position amongst philosophers and many scientists, and it allows them to keep their general commitment to physicalism or materialism or naturalism—that there's nothing supernatural in the world—so they don't have to posit a soul or a spirit or some magic ghost in the machine. They can square that view with the idea that we can still defend our moral responsibility. The real problem with that is that it doesn't explain how you could have any choice. If you accept the real hard physical determinism, then there aren't any choices there. Things are just happening. And the argument is that, okay, but if you were configured differently, you would have had different reasons for doing something. Therefore, you were the source of the causation of what happened.
So it's kind of an appeal to what's called a counterfactual scenario, this imaginary scenario in which you could have done otherwise if you had wanted to. And yet in the real universe, you never would have because nothing else would ever have been different. So it becomes an odd—to my mind slightly tortuous—exercise in motivated reasoning because it's people trying to defend moral responsibility. They literally just want to defend moral responsibility against determinism, when in fact they don't have to do that, because determinism just isn't true. Physics doesn't say the world is like that.
Mounk: Let me try and steelman that argument for a moment and see how you respond to it. I think that the intuition is the following one, like, how do I feel about my best friend? Why is it that I appreciate them? Why is it that I like them? Why is it that I praise them for their good actions? Why is it that I disdain my nemesis? Why do I think that they're a bad person doing bad things in the world?
If you tell me, it just so happens that this person's character is good and this other person's character is bad, it just so happens that they have certain predispositions which make one person inclined to be really altruistic and loving and really thoughtful about the world, and the other person knee-jerk and reactive and mean… it's not clear to me that that defeats how I feel about them. I mean, to me, the object of moral judgment is the person they are, the character they have.
Now, it may not just be a particular action. Suppose you have a person who, in general, is incredibly kind and cleverly altruistic, but, on a particular occasion, somebody slipped a weird drug into their food, unbeknownst to them, and that drug just has the effect of making them act meanly—I might say, all right, I’m not going to hold them responsible for that particular action because it doesn't flow from their deeper character. But to me, what moral judgment in human affairs is, is to say, do they have a kind of general character, the kind of general dispositions, or the kind of general pattern of action, which is praiseworthy, which is attractive, which is appealing in various ways? And the people who are my good friends are people who I appreciate because I clearly answer that question in the affirmative. Now, you might tell me that if they'd had a different mom and dad, or if the alleles had somehow aligned in a different way when they were an embryo, or if they had gone through horrible experiences as a child that would have permanently psychologically maimed them, then they wouldn't be like that. I'm willing to accept that. I'm willing to accept that there's a good amount of luck in having created the person I appreciate—but the person that I judge in a moral way is the person they are. And even though the person they are was shaped in all kinds of ways by luck and circumstance, that feels to me like the right object of moral judgment. That, to me, is a strength of the compatibilist insight.
Mitchell: I think that's right. And I think that would be the compatibilist argument there. However, there are a couple of things being mixed up there. So there's the question of free will, which is, I think you could say, at any moment, are there choices available to you in the way the universe could go? And among those choices, do you have control over what actually happens? And those are questions that you can actually ask and answer without talking about moral responsibility, without talking about merit or praise or blame or what we find worthy of praise or blame, which are all to me secondary or tertiary kinds of questions layered on top of that one. And when they get conflated like that—a lot of the free will literature just takes free will and moral responsibility as the same question almost—to me that leads to some confusion, or it leads to people making arguments at one level—like what you find praiseworthy or blameworthy about a person’s character in social settings or relationships—and using that against a premise couched at the level of particle physics. But those two things just feel like a mile away from each other.
Mounk: I think to be fair, that is a conscious move, because I think what people like Scanlon would say is, Who cares about particle physics? Perhaps the determinists are right about particle physics. Perhaps they're not right about particle physics. But the choices we make have a lot of relevance in the world irrespective of that.
The most trivial example given in the literature, as I recall, is to say, if you're a waiter, should you respect the choices that your customers, your diners make for what to order for lunch or dinner? Now, let's say that it's true that the world is deterministic, that an omniscient being would know in advance that somebody is gonna get the salmon. And perhaps it's even true that actually they would enjoy the steak more if it somehow arrived. We still have reason to respect the choice of the person ordering one meal rather than another for all kinds of social reasons. And so, in the same way, at a much deeper level, you then recognize that how I'm going to feel about my friend or how I'm going to feel about my nemesis doesn't actually turn on alleles in the formation of their fetus. So it's not an answer to the first fundamental question, but it's an argument to say, we don't need to bother with that first fundamental question because the thing we care about doesn't turn on it.
Mitchell: You're right. I think if you could imagine a scenario in which there are people going into restaurants, making choices about what to eat, and there are waiters making choices about how to feel about it under determinism, then that's fine. You get to make that move. You get to go from here, all the way to there, and say, I don't care about all the intervening stuff. But what I would say is that, under compatibilism, you're assuming a deterministic universe, so how do you get there? There's no way to get the emergence of beings like us that seem to be decision-making beings with real agency and choices and using the kind of vocabulary that we talk about—choices, taking action, making decisions, and so on. None of that means anything in a deterministic universe, nor do compatibilists ever supply an account of how such agents would emerge. They just take that for granted and then say, even if everything were deterministic, we can still apply these moral arguments as you just did. I just don't think we would get there without some indeterminacy. Thankfully again, it becomes moot because physics just doesn't say that the world is deterministic. It's just a misreading of basic physics actually to think that.
Mounk: Talk us through the libertarian tradition on free will and then I think we'll have done enough setup.
Mitchell: The traditional libertarian view is defending free will, but usually they're incompatibilists. And they say, if the world were deterministic, we wouldn't have free will, but the world isn't deterministic, and that leaves some scope for us to intervene in decisions and decide which way things go. Part of the problem with that sort of traditional libertarian view is that they have developed a kind of view of the way the universe is—which is what I like to call determinism plus randomness—where it's like things are going along mostly deterministically and then every once in a while a radioactive atom decays or you get some quantum event or some weird little random thing happens. And the problem with that kind of view is that it doesn't explain how you are doing anything. It leaves you open to this challenge. Either way, you're screwed. Either everything is determined at the low level and you're not making any decisions, or there are some random things happening at the low level, which are the things that determine what happens, in which case you as a whole entity are still not doing anything.
So that's why libertarianism has always been kind of unsatisfactory, I think, to a lot of people—because it doesn't yet answer the question of how you go from having some indeterminacy at low levels to the emergence of higher order structures like living systems that have control of things at a high level.
Mounk: That's really helpful because I was always somewhat confused about the moral intuition behind libertarianism for exactly the reasons that you outlined. Just to go back, why is it called libertarianism? I assume that it's in contrast to determinism? Whereas the determinists think that the world is unfree in the sense that if you know everything about it at point A, in theory, you could know everything about it at point Z. In the libertarian world, there's some degrees of freedom in that world. There's going to be some random events such that knowing everything at point A doesn't actually tell you what's going to happen at point C. In a way that translates to, well, if you don't know in advance what's going to happen, then perhaps we do have free will because somewhere along the way people can make decisions that are different. You can't predict how people are going to act.
But as you're saying, the problem there is, well, is it the kind of action worth having? Is it the kind of action that actually would make us think that we have free will?
Let's say that you have a self-conception as having made a really reasoned decision about whether or not to have children. You've reflected in a deep way about what your life goals are and what you really want to accomplish. And so you feel proud of your decision to have kids or not to have kids. It really reflects yourself in a deep way. Now, the threat comes in from a determinist and they say, haha, actually, you're deluded. You're an idiot. It was always predetermined that you were going to have kids because it's just a matter of how these atoms hit each other. And we could have known that all along. Well, OK, that seems upsetting. Now here comes the libertarian and they say, Great news. Actually, it wasn't predetermined at point A whether you're going to have kids or not. It's because of quantum physics and quantum mechanics. And actually, there's this one random way in which one atom hit the other in a way that couldn't have been predicted. And that's why you had kids.
And of course, the answer is, how is that any better? How does that make me feel that that decision comes in any sort of more significant way from what seemed to be my reflection about the world, my values, my goals? I mean, if that's the free will it saves, it doesn't seem to be a free will worth having.
Mitchell: That's always been this sort of classic problem. And it's partly why compatibilists say, look, it doesn't matter what the answer to determinism is. Either way, we can just have these discussions at a high level about moral behavior and talk about that. We don't need to think about what's going on at these low levels. So, the problem with the libertarian view is it shows where freedom comes from in the world, in the sense that many different things could happen, just based on physics. But it doesn't answer the question, Where does the control come from? That's the question that's really interesting. How does a living being control what happens? Given that now many things could happen, how do I narrow that possibility space so that the thing that happens is the thing that I want to happen? That's the real challenge.
And it turns out that actually that's what being a living thing is all about—making things happen. That's almost the definition of life: living things have causal power as holistic entities. It's not the power of any of their parts. It's the power of the whole thing to make things happen in the world. Generally what they're making happen is themselves. They're making themselves happen. That's how they keep on living. Even a simple bacterium is constantly working against the laws of thermodynamics that say all of its parts should go into equilibrium with the universe and it's taking in energy and working to make sure that it doesn’t.
Mounk: So I think we've gotten people up to speed on some of the basic debates about free will. One of the characteristics of this is that for the most part, these debates have been armchair debates—whether from philosophers or theologians, it was people sort of sitting, thinking about the world in an abstract way and trying to grapple with these incredibly complicated questions. Now, your training, as I understand it, is as a biologist, as a neuroscientist, and so you take a different kind of empirical approach to this. This is not entirely novel. There's a bunch of people who've been trying to apply neuroscience and other disciplines to this debate about free will. But you come up with an answer that is rather different from what I’ve seen before. And that is to say, part of the answer here is evolution, that somehow through evolutionary processes we have emerged with some kind of capacity for a higher level of reasoning. Explain to us first of all what that process looked like, and then I'll push you a little bit on what exactly that means for our ability to reason freely.
Mitchell: To me, if you think about any biological phenomenon, taking an evolutionary perspective is always a good idea because that's just how they came into being and it helps to understand them. And I think in this case in particular, we're having these discussions about human free will and human cognition and consciousness and it's so sophisticated because we have layers of language and sociality and morality on top of it. It's just an incredibly complex kind of scenario to try and understand off the bat. So there's an argument to be made that, actually, let's start with just the very simplest kind of instantiation of something doing something. The simplest living being that we can say is acting in the world. And let's try and understand that and ground some concepts that we're going to need to understand human cognition and then trace across evolution this elaboration from single-cell organisms to multicellular ones to ones with nervous systems to ones with much more complicated levels of cognition and metacognition and so on. That's the arc that I trace in the book. And it really actually starts with the origin of life, what it means to be a living being at all.
And in bacteria, all sorts of chemical metabolic reactions are going on, there is a sort of a regime of interlocking, constraining processes, all these feedback processes that link the whole thing into an ongoing pattern that is the bacterium. That's what the bacterium is, and it's alive because it continues to be that way through time. In order to continue to be that way, it has to do work. It has to take in energy. But the problem is that the world can be a hostile place. It can be a changeable place, which means that for organisms to persist for longer, it helps if they can take in information about the state of the world and then act on that information in a way that's adaptive. So even the simplest organism, you could say, first of all, has a purpose. Something that is just not present in the nonliving world. An atom has no purpose. The sun has no purpose. The whole universe has no purpose, but a bacterium does. And its purpose is to persist through time. To survive and reproduce. Because of that, they get selected for functionalities that help them to survive in a changeable world where they can use information and then act on it.
And so once you get that, when those concepts are grounded, then you can just elaborate. You're off to the races. We've got organisms doing things as whole entities, and from there, you can just get more sophisticated, which is what happened gradually over evolutionary time.
Mounk: It's interesting to think of the bacterium as sort of goal-directive and purposive in this kind of way. And I look forward to hearing what new levels of complexity and therefore perhaps freedom are introduced as we go up the evolutionary chain to more complex beings. But how exactly does this defeat the determinist argument? Presumably a determinist is going to say, at the level of the bacteria, OK, fine. They respond to a stimulus. They'll go left if the sugar is present on the left, and they're not going to go left if the sugar isn't present on the left. That's different from an atom just sort of randomly knocking into things or a tree leaf blowing in the wind. But we know that human beings were capable of those kinds of decisions. We know that if you're starving and somebody puts a steak in front of you, you're going to move towards the steak. That's not news to us. The point is that the laws of the universe predetermined that you were going to be hungry at this point and that then somebody is going to put that steak in front of you and that those two things being the case, you're going to move towards the steak. And surely the same is true of the bacterium.
So how is it that this has changed anything about the determinist account of the world?
Mitchell: There's two sides to that argument. One is going back to the idea of physical predeterminism: that actually everything that was going to happen is just determined by physical laws. I mentioned that libertarians would argue with that. There's no evidence for that from physics, but there's two ways of thinking about that. One is “determinism plus randomness,” where the default state is deterministic, but occasionally you get these incursions of random events. But I don't think that's an accurate way to think about it. I think more generally, it's just that you have a kind of a pervasive indefiniteness to the present state of things. There's just a limit to how precisely the physical parameters of a system are specified. It's nothing to do with observers or uncertainty or anything. It's just that the things themselves have a finite amount of information as a physical system at any point.
So given that there's some fuzziness and some jitteriness at the lowest levels, it's just that many things could happen. It's not that there are branching paths in the future waiting for us to choose. It's just a big fuzzy continuum. So we can say for our bacterium, then, that the whole reason why a bacterium could evolve in the first place is because the causation is not exhausted at those low levels. Many things could happen based on the laws of physics. The way that a system is organized can have some top-down causal influence and constrain the way that a system is. Again, based on the principles of selection in the case of bacteria, we can say the same thing, looking at our computers here—the way that they're physically structured constrains physically what goes on within them. It doesn't change the laws of physics. The electrons are still behaving as electrons. But they're constrained to do what they do because of, in our case, evolutionary selection. So we can put the really hard predeterministic thing aside and say, well, it wasn't predetermined from the dawn of time that the bacterium would be here and the food would be over here.
However, that leads to a different kind of worry, which is that actually the bacterium, while it is making decisions, is doing it as an automaton. It's basically pre-programmed, it's like a little robot. It's being pushed around by its parts. That's the concern, I think. And it's the concern when neuroscientists look inside you and do an fMRI scan, a neuroimaging scan, and say, you only made that decision because this part of your brain was active. Well, you can say the same thing about a bacterium. The bacterium only moved towards the sugar because these proteins within it got phosphorylated and they linked to these other proteins and they linked to the motor that makes the bacterium move and it's just a stimulus/response machine, the bacterium is not doing anything. And I think there's some truth to that. Of course, a bacterium, while you can say it's doing goal-directed behavior, is not conscious of that. It's not aware of its goals. It can't think about its goals. However, it's actually much more sophisticated than you might think.
When we do our laboratory experiments, we have this tendency to control everything. We control everything that we're not interested in and we just change one thing in the environment at a time. So we have our little bacterium and we either put down some sugar or we don't. And when we put down some sugar, we can see this biochemical pathway acting. We say, look, that's the explanation. But the bacterium never encounters the world like that, it always encounters loads of things at the same time that it has to integrate. And maybe there's a threat in the same place. Maybe there's some bad chemicals where the sugar is. It has to balance those options. It has to integrate with its own metabolic state, with its history, with how crowded the place is. So it's doing a much more holistic kind of a job that justifies thinking of even a simple bacterium as an agent, as opposed to just a machine.
That's where I think we get off the ground with a very basal kind of agency. I'm not saying bacteria have free will. I reserve free will just for humans just to avoid any kind of confusion. But it's a start. It really is the entity as a whole thing. And also, I guess the other thing that's really important there is that it's an entity with continuity through time. That's why we should think of it as a self, as a being, because it continues through time. It's not just a machine, a physical thing with a physical state in an instant that responds to things. The way it responds depends on its history. Now, for bacteria, it's mostly evolutionary history. As things got more complicated, as we get multicellular organisms, especially with nervous systems, then they learned a new trick. Then they could learn as an individual about experiences in the world. They could develop new goals. They're pre-programmed to learn as individuals and then they can develop much more flexible and really genuinely more autonomous behaviors.
Mounk: I love that this conversation is doubling as a remedial lesson, certainly to me, in the biological history of the emergence of higher forms of life. I have a bunch of responses to this idea about the bacteria, but let's follow that chain. And then I'll express where I find what you're saying convincing and where I still have questions. So you get a kind of goal-directed behavior in even a single-cell organism. But as you're saying, there's constraints on the extent to which an organism can reflect on its goals, can act on the basis of its own history rather than just the things that through evolution have been baked into its genetic code for how to behave. What happens once we get multicellular organisms? And what happens once we get animals? And what happens once we get higher-order mammals? Talk us through how all of that happens, but more importantly, how all of that allows something like free will to emerge.
Mitchell: Yeah, so it's interesting that if you think about a bacterium, the challenge that it faces in order to survive is to know what's out in the world and what it should do about it. It needs to have some sensory systems, which are just little proteins on the surface of the cell. It needs to have a motor so that it can move around in the world. But it also needs to have some policies. It needs to have some kind of knowledge about what's the right thing to do given what it senses. Most of those control policies are pre-programmed by evolution. But as you get multicellular organisms that develop more complex bodies, well, first of all, they face the same problems. The world is changing. They need to know what's out there and what they should do about it. So they needed to develop sensory systems, things like sense of smell, things like vision and hearing and touch and so on. They needed to develop motor systems, which in this case become more sophisticated because they're not just locomoting in the world like a bacterium does like a rowboat where the shape of the boat stays the same. The shape of the animal can change. They're not just able to move in the world. They're able to act upon the world. Things with limbs or tentacles or wings or mouth parts or whatever it is, they can manipulate things in the world. So the possibilities for action just became much more open-ended.
That's really important because it's the open-endedness of the possibilities that makes cognition worthwhile. That's why evolution invested in it, because there were so many opportunities there that evolution couldn't pre-state what you should do in every scenario. It just gets too complicated. It's a combinatorial explosion of possibilities. So instead, evolution said, well, you figure it out. You're on your own, go for it. I'll give you some basic motivations and some basic policies, but after that, you need to scaffold on top of that. So, many kinds of animals are capable of that, and along the lineage that leads to humans, you go to vertebrates and then mammals and then primates and then humans.
Those kinds of capacities got more and more sophisticated as the brain got larger. The important thing about that is that actually what you get as the brain gets larger is more internal processing. You get a greater separation between the sensors and the motors. Really simple animals might have either the motors and the sensors might be directly coupled to each other, or they might have like one intervening layer of neurons or two intervening layers of neurons. As you get more complicated animals, you get many, many internal layers. So the brain is processing much much more information. It's building models of the world, it's building models of the self and it's learning from experience. Like, the last time I encountered something like this, I did X and X turned out well, so I'm going to do it again. And that is how we learn from experience. It's called reinforcement learning where rewards or punishments based on our past actions can inform future behaviors.
So in that way, organisms became much more autonomous, much more self-directed, and much more able to plan for distant things in the future. So what they're doing is they're not just acting for reasons, they're accumulating reasons. They're learning through their own experience in the world. And ultimately, along that lineage, what we get is creatures like humans, who are capable of not just having a model of the world and a model of themselves, but also a model of their own mental workings. They developed enough levels of the neural hierarchy that the top levels are actually looking down at the workings of the lower ones and they're internally representing thoughts and beliefs and desires and so on. So there you get to consciousness, you get to what we call metacognition, even meta-volition, where ultimately, we're not just acting for reasons, we're able to think about our reasons, we're able to arrive at new reasons by the act of reasoning. And so we're able to think. And we have to think in all these novel scenarios. It's just not possible to pre-state every action under every conceivable scenario. That's why we have a brain, is to figure that out. There aren't pre-set weights that determine the outcome in every scenario. That's what thinking is for, is to figure out what those weights should be in this new set of combinations of threats and opportunities and so on.
Mounk: So help me understand how exactly the complexity that emerges through evolution relates to the question of free will. I imagine that somebody like a determinist is going to say, well, look, we knew that the human brain is complex to begin with. It's not like that is a novel discovery. All through this armchair debate about free will, we started from the premise that the very puzzle is that humans seem to be these incredibly complex beings and when we're reflecting about things it feels like all of these competing considerations are in our mind and we're reflecting on our purpose and our goals in life and all kinds of things. But in the end, if you think that there is some compatibilist element to the world and perhaps we have to get into that part of the argument in greater detail, then that just all seems to be an illusion because in fact for all of that complexity what you're ultimately going to do was predetermined and so it seems in some ways to be an illusion. Now you're making the argument that actually atoms banging into each other in the world don't have any kind of real agency. Already by the time we get to a bacterium there is a real kind of goal-directedness but not something you would call free will.
Then you get multicellular organisms, more complicated animals, and finally humans, and what's going on in their brain when they're trying to figure out what to do is all of these really complicated considerations. They seem to be determinative of what we do. But determinists are still going to say, OK, sure, this machine is now much more complex. We've gone from a really simple calculator, which perhaps is the equivalent of a bacterium, to the highly sophisticated laptop which I'm using to record this conversation. And there's obviously a huge difference in scale there. But in the end, the pixels that my laptop displays are just as predetermined by things that are outside of its volition as the number that a much more simple calculator comes to when I type in two plus two.
Mitchell: I think there's two ways to look at that in terms of freedom. You can think about it in terms of autonomy from the environment—how much causal power does an agent or an entity have in the world? How much is it pushed around by the world versus how much can it push back? And so what I would say is that we—multicellular animals, mammals, and so on—are more able to push back on the world. It's a vague way of putting it, but you can actually find very concrete ways of measuring that kind of thing. Like for example, how much information can you integrate at once? How many levels of a hierarchy, how much sort of higher order abstraction are you doing to think about things in the world? What's the time horizon that you're thinking over?
I can do things right now that will affect events a year from now or two years from now. And I can consciously do that. Whereas a little worm that's just wriggling around, that doesn't even have eyes, for example, is not going to do that, partly because it has no information about anything that's further away from it than like a centimeter. So they inhabit the here and now, and they behave accordingly. They don't have cognitive systems for planning over long timeframes, whereas what happened over our evolution is that we developed those capacities and what that means is that we need the capacities to be able to adjudicate between goals. Because we have short-term goals and long-term goals. A lot of them will be conflicting with each other. There's opportunity costs. We can't do them all at the same time. So we need to be able to make decisions at any moment and in real time, but also through time, we need to be able to juggle those things, prioritize, optimize our behavior, and then follow through on it. Our behavioral decision-making is not just moment by moment, I can do A or B, and then I can do C or D, and then can do E or F. It's like, I can do A or B, I decided to do A, and now that's going to take me a while. I decided to go to college to get a degree. That's four years I'm gonna have to continue holding that goal in my mind, and that's going to constrain and inform my behavior on a daily basis.
So we have agendas and commitments and policies that we build up through time. And you could say all of that allows us to be more proactive as agents in the world and less immediately reactive. Now the problem with that is, while I think that's a good valid argument, it doesn't meet the challenge that people could say, okay, fine, you're not being coerced by things outside you as much as more reactive organisms are, but you're still being compelled by things inside you. You still have all those psychological things.
Mounk: So I have your book in front of me. I certainly have agency in the way that a bacterium or even a worm does not, in that I can pick up the book, which is an excellent book, and I can sort of hold it into the camera and I can open up the book. And I know that I am able to do all of those things. I can manipulate the book in these ways. The question that bothers the free will skeptics is, but was it in some ways predetermined or was it in some ways inevitably downstream from my inclinations from the way my brain works that I was going to do that? And of course you can extend that argument to something more complex. Like, I want to invest into my individual retirement account. I want to put money into my IRA, but that invitation to go out drinking with my buddies is just too tempting. And by the time that the end of the month comes around, I just don't have the money to do that. Again, in some ways this is due to my choices and so on, but perhaps it's in fact not due to my choices because I just don't have control over myself in the right ways and that's to do with my genes or my upbringing or whatever. So yes, I can manipulate the world, can invest in the IRA and be really rich when I'm old or I can fail to invest in my IRA and have very limited freedom of action when I'm old because I don't have the money to do anything. But the underlying worry isn't that, yes, I can make those choices and they have big consequences in the world 30 years from now possibly, but which of these choices I make seems like it may have been influenced by things that are not as fully under my control as I thought?
Mitchell: I liked the way that you phrased it at the end there, because the way that you phrase it is completely uncontroversial. It's absolutely true that those decisions are influenced by things that are not under as much control as you thought, potentially, if you thought you were in complete control and weren't constrained at all by prior influences. But that's a really, really far cry from saying no one has any control whatsoever at any moment. You're going from, I had less free will than I thought I had to everyone has zero free will. Those are just two extremely different propositions. And I don't think that the more extreme version is defensible while the less extreme version is kind of obvious in the sense that everyone agrees that we have these influences on our behavior. They just have different views of what that means for things like our own kind of meta-control. And this is where metacognition and introspection come in and the question of the emergence of character, because you could say we all have personality predispositions, which is absolutely true—shaping our psychology in broad patterns like how outgoing you are, how conscientious, how neurotic, whatever, is very different from saying those predispositions determine precisely what you're going to do in every exact moment. They don't have enough information to do that. Instead, they inform the emergence of our character through time, the way that we adapt to the world. I've had sets of experiences and I've adapted them to them in my way. Someone else might have adapted to them in their way and that's fine. That's just how we come to be in the world.
Now there's a question, do we ever have any control over that or are we just passive passengers in that process? And some people like Robert Spolsky, for example, would say all of that is just passive. It's just your genes, it's evolution, things in the womb and then the accumulated effect of every experience that has happened to you. That's the way that he phrases it: has happened to you as if you had nothing to do with it. But in fact, if we're making choices in the moment, if there's some freedom and we have some control, then those are things that we chose. A lot of the time we're choosing actions, we're choosing environments, we're selecting, we're creating, we're building our own niches as we go along. We're choosing which friends to go out with, the ones that tempt us out at the end of the month.
So there's a very different view that you can have on that, which is, first of all, that the way that our character emerges is informed by our experiences, which are the result of our own choices throughout our lifetime. And secondly, there's another level on top of that, which is that actually we can very actively ask ourselves, am I making the right kinds of decisions? Are these the right kinds of motivations that I should have going forward? Actually going forward, I want to be more responsible in thinking about my future and I'm going to decide to do that. And even though I know I can recognize that I'm going to be tempted, I'm going to take action by actually putting a hundred euros in my IRA at the start of the month instead of at the end. So, you can build those kinds of things and we do that. It's just clear that we think about our own character. We think about our own volition, in this metacognitive introspective kind of a way. And of course, like some people do it more than others and it can be effortful, but it's possible in principle to do that. And so the idea that I wasn’t involved at all in the configuration of my brain and mind right now, I just don't buy it. I'm just totally not convinced by that because there's so much evidence that in fact that's partly the accumulation of my own choices through time.
Mounk: So I'm going to say something that I think will half make you happy and half make you sad, but it's not deterministic, so we don't know exactly what's going to happen. I find what you're saying very convincing. I have trouble seeing how it is as different as you claim from the tradition that you were quite dismissive of earlier, which is the compatibilist one. It seems to me that what you're describing is the ways that make human life worthwhile are the ways in which we do seem to be free. I can reflect on whether to put the money in the IRA at the beginning or the end of the month. I can reflect on the way my life is going as a whole and say, you know what, I have great friends but I don't spend enough time with them, I should spend more time with them. Or you know what, I actually think that on reflection my friends aren't the right ones, they don't share my values and I should go and find new friends that share my values more closely. And that is what to me makes human agency important.
Now, I don't see how your evolutionary argument tells us that this isn't somehow determined or that this isn't subject to some of the worries that some of the skeptics about free will have. In fact, you might say the fact that this is all downstream from an evolutionary mechanism is one more thing that really constrains how we go into work and how we are in the world, because we certainly didn't choose all of those branches in the evolutionary tree that sort of shaped all of those capacities and so on.
But to me, the strongest argument is, but it shouldn't matter. The point is that you're the kind of higher order being that is capable of reflecting about the world whose actions reflect not just the goal-drivenness of a bacterium, but a much more complex, much more sophisticated sort of goal-drivenness, which does come downstream, among other things, of this really complicated evolutionary process that has given us those faculties. So who cares whether or not in some sense it's deterministic? That's not what matters here. What matters is that my decision to meet up with my friend and to have a glass of wine to many in order to connect and to have a wonderful human experience comes downstream from my values, from my purpose, from my reflections on the world. And whether my values and my purpose, my reflection on what I should do in the world is itself somehow causally predetermined just doesn't seem that important. That to me is a compatibilist argument. And it sounds to me—and you'll explain to me where I'm going wrong—suspiciously similar to what you've been saying.
Mitchell: In terms of moral responsibility, I actually end up in a point that's very similar to compatibilists. It's just that the evolutionary view, and the fact that it relies on some indeterminacy to emerge, allows me to get there. That's the only difference in a sense. But actually at the end, you could say, well, what do we actually judge people morally responsible for? It's their actions, but it's also their character. It's why they did something. It's not just that someone did something. So what we're doing is we're ascribing some ownership, some responsibility for their character. Under determinism, there's no way that they could have any responsibility for their character because they never were actually in control at any moment because the same thing was always going to happen. That's why I find that argument incoherent, but I end up at the same point with respect to moral responsibility because we do hold people responsible for their moral character, not just for their actions. Their actions issue from that character. It's the character that we're actually judging a lot of the time. And I think the view that I've outlined there, which is a naturalistic way that you can have a system that actually has control in the moment and that has meta-control and metacognition, that's a scientifically plausible way that you could get a system like that emerging. And then you end up in that endpoint where we can think about moral responsibility that way.
The other aspect I would say is that some people use this term called ultimate moral responsibility. But so you can't be held responsible if you weren't the ultimate cause of everything that you did, which is to say it's almost a way of saying if there was any prior cause that influenced you, that you weren't involved in, then you can't be held responsible. Like if you weren't responsible personally for all of human evolution. To make it absurd, if you personally didn't institute the Big Bang, then you can't be held responsible for anything that happened since then. It's just kind of silly. We're not after ultimate moral responsibility. We're after proximal, proximal moral responsibility. That's what we want—right now, given all of that, in the choices that are open to you, with the constraints that are there, what did you do? Why did you do it? That's what we care about when we talk about moral responsibility.
Mounk: But that distinction between proximate and ultimate responsibility is exactly what in my mind sounds like those compatibilist arguments. To say that as long as your actions right now reflect your character and your intentions and so on, who cares where that character and those intentions came from in the first place? Now, I understand that I think you want to respond to that by saying, but in fact, it is important to note that your character and your responses and so on are themselves in some sense self-created—not fully self-created, but that you had degrees of freedom in the past actions you took that shaped your character and so on. So let me push on that a little bit. I think here I just haven't yet fully understood your argument. So I'd love for you to explain it to me in more detail, perhaps just to re-explain it, which is how evolution creates a system that is not deterministic.
I want to distinguish here between sort of the levels of complexity that might be involved in a physical system and whether or not we should think of the people within that system as having freedom to act. So the free body problem. We don't really know how to predict how free planets that are orbiting around each other are going to behave. It's too complicated, we can't figure it out. But it would be very strange to say that therefore each of these planets or one of these planets has freedom to decide where to move next. Clearly each of these bodies are just acted upon in a way that pushes them in one direction to another. Given the complexity of the system, we find it impossible to predict which way we're going to go. But that doesn't mean that they can choose which way to go. And isn't that similar to our reflections about what to do? The machine that evolution has created in our brain is so incredibly complex that it does have all of these higher order reflections and all of those neurons doing all of this complicated stuff. And so we may never be able to predict exactly what action that's going to lead to. The fact that we're not able to predict that doesn't mean that we actually have freedom. In the end, like in one of the planets of free body problem, you're either gonna invest in the IRA or not. And perhaps that is in fact downstream from those predetermined things, which doesn't have to worry a compatibilist because it says, but your action reflects the things that you ultimately care about in such a way that that's what matters. But you want to say, no, no, somehow complexity translates into freedom in a way that I haven't quite followed.
Mitchell: There's multiple levels of concern. There's the real sort of basal physical predeterminism, which is to say, it comes back to the question, how does indeterminacy help the argument? How does a control system emerge from this indeterminate thing where many things could happen and what the organism is trying to do is narrow that down and make certain things happen? Now, it's doing that with noisy components. They're just jittering all around the place, even in a bacterium, little proteins binding and unbinding and diffusing around the place. So the problem for the organism is to try and control things as much as it can to the level it needs to. So it doesn't need to control where every little protein diffuses. It can't do that. It just needs to, at a macroscopic scale, control the outcome at a level that is equivalent to its surviving or not surviving, its getting some nutrients or not getting some nutrients. There's a level that it cares about, there's a level that evolution cares about, and below that it doesn't have to micromanage, it just has to constrain.
Control systems emerge in that way because they can. Literally because you could have several different kinds of organization of these things and you're just going to lock into certain ones that tend to persist more and then because they persist more, they'll persist more. And then they're the next stage of evolution and so on. So you get this kind of emergence of control systems that favor the persistence of macroscopic form and pattern. And then you just elaborate on those. And so in us, we have control systems that are, as we've been saying, so much more sophisticated, they're operating over all kinds of information about the world, all kinds of information that we've gleaned from our own experience so we know what to expect. We're making predictions, we're testing them out, we're thinking about the utility of various actions, the reward or punishment that we might get from those. We're weighing that all up as an entity. For our reasons, not for the reasons of our parts. And so we can say we're really not an automaton. We are a holistic macroscopic entity with continuity through time that's acting for individual level reasons. But again, still, the control thing is still a problem. I mean, the funny thing is that the deterministic view for someone like Robert Sapolsky, for example, is almost saying you have so much control that you have no choice, which is just an odd place to land.
Whereas a more, I think, biologically realistic view is that we have some control. That we're trying to exercise as much control as we can, but we've still got noisy components. We're still making decisions with uncertain information and ambiguous beliefs and conflicting desires in a world that's dynamic in the process as we're trying to make a decision. There's just no way that any of that could be completely predeterministic at the level of cognition, unless it were all predeterministic at the level of atoms, and it's not.
But if we think about, even from an economics point of view, decision-making, there's this theory of bounded rationality where we don't make fully rational decisions all the time where there's just one right answer that our brain algorithm has to come to that could in some sense be pre-stated as the optimal outcome because we don't have all the information. Things are changing, and there's conflict. So we have to work through that conflict. We have to figure out what are the weights that are appropriate in this scenario for me. There's some little bit of randomness at play sometimes in that. Sometimes we know what we want to do. Our will is settled. Like if we were in that scenario again, we totally absolutely would do that thing again. It's to be habitual or it's very clear. Other times, we don't know at all and we don't have enough information or we don't care. Do you want a Coke or a Pepsi? I don't care. Like maybe some random noise in my brain kind of, flips it one way or the other. That's fine. That's actually a super good way to make decisions when you don't have enough information. The important thing is that you make a decision. If you're indifferent, it doesn't matter what it is. And then we have a conflict where we're like, should I ask this girl to marry me or should we break up? These are really conflicted and you're going to work through those things with the information that you have and come to a decision and whichever way it goes, it's going to be your decision. And you'll be able to defend the reasons that you had for making that decision. Ultimately, control is what we're aiming for, but the idea that we have so much control that we end up with no choice is just, like I said, a slightly odd position to end in.
In the rest of this conversation, Yascha Mounk and Kevin Mitchell ask if AI could ever have free will, and how different approaches to free will affect criminal justice. This part of the conversation is reserved for paying subscribers…
