Robert Sapolsky is Wrong
A new book about free will fails to offer an original argument or make a convincing case.
A review of Determined: A Science of Life Without Free Will by Robert Sapolsky, 528 pages, Penguin (October 2023)
In Determined: A Science of Life Without Free Will, Robert Sapolsky argues that free will does not exist, and explores how he thinks society should change in light of that conclusion. Sapolsky is a professor of biology and neurology at Stanford University, known for his studies of hormones and behavior in wild baboons. In this new book, he makes an effort to address several different ways in which others have proposed that free will could be real, despite the laws of nature acting on our physical brains. But in his efforts to cover the bases, Sapolsky fails to offer an original argument supporting his claim that free will is not real. Instead he serves up a partial and rewarmed version of the argument made by Bertrand Russell in the 1940s. Sapolsky ornaments the argument with findings from the biological sciences, which he admits do not prove his intended point.
In 1946, Bertrand Russell claimed that a scientific understanding of human physiology would reveal ordinary physical determinism at work in every human choice. To Russell, where there’s determinism, there can’t be any free will. And determinism seems to work the same way in human physiology as it does anywhere else, without little gaps for magic. So Russell saw no room for free will in a deterministic world informed by science. Sapolsky makes basically the same claim, but he illustrates the point with examples of biological mechanisms. None of the particular examples are either necessary or sufficient for the overall argument, as Sapolsky says: “You can’t disprove free will with a ‘scientific result’ from genetics or any other scientific discipline. But put all the scientific results together, from all the relevant scientific disciplines, and there’s no room for free will.”
What exactly does “no room” mean? And what kind of “space” would free will occupy? We are not told. Nor does Sapolsky explain how determinism would keep free will out of that space. There seems to be an unstated premise: If an event is determined, it can’t be free. But if that’s simply true, then why write a book instead of a sentence? Rather than arguing that determinism does inherently preclude freedom, the book does two things: it reminds readers that biological processes are deterministic, and it tries to show shortcomings in several theories of free will. In order to do that, Sapolsky states his criterion: “What is needed to prove free will: show me that the thing a neuron just did in someone’s brain was unaffected by preceding factors.” This criterion, and Sapolsky’s use of it, are shot through with self-contradiction. Most of his attacks on theories of free will do not make any sense when examined. So, the book neither offers a distinct argument for the hard determinist position, nor a good reason to reject those theories that claim free will is compatible with determinism.
I’ll start an examination of Sapolsky’s self-contradictions with an easy one: the one he recognizes as such himself. The final chapter explores moral and political implications of his anti-free-will determinism: “There is no justifiable ‘deserve.’ The only possible moral conclusion is that you are no more entitled to have your needs and desires met than is any other human. That there is no human who is less worthy than you to have their well-being considered.” But moral declarations like this depend on certain assumptions—for example: that humans deserve to have their well-being considered at all. Sapolsky recognizes this: “Emotional responses feel real. Are real. Pain is painful. … It is logically indefensible, ludicrous, meaningless to believe that something ‘good’ can happen to a machine. Nonetheless, I am certain that it is good if people feel less pain and more happiness.” It might seem like this self-contradiction is limited to the moral realm; the forgivable result of having a big heart while also being an unwavering scientist. But there’s more to it. It’s actually just one instance of a recurring logical error that leads to self-contradiction in the supposed “scientific” aspects of Sapolsky’s thought as well.
Sapolsky recognizes that his care for people is not logically reconcilable with his reductionist determinism. He’s right that the two don’t work together, but he’s wrong about which of the two is ludicrous. If he thought a little more about his observations, he might see why. Pain is real. Pretending otherwise would be the epitome of nonsense. Before moving on to the question of whether or not the pain matters, we already have a contradiction with Sapolsky’s reductionist determinism. If he were consistent in his approach to understanding the human subject, he would say, “If pain exists, show me a neuron that just experienced pain.” After all, that is the only way he allows a function of the mind to be real; it has to happen in a neuron; that’s his criterion. But there is no neuron that experiences pain; there is no neuron that experiences anything.
Experience is a phenomenon that unquestionably exists. Yet it only exists at the level of the person, not the neuron. Neurons are absolutely involved. But it must be neurons (plural); basically all of the neurons in a person’s brain taken together, and only taken together. It would obviously be a mistake to look for experience in a neuron, find none, and conclude that there must not be any experience in general. In the same way, it is a mistake to look for free will in a neuron. It’s easy to see why. Before getting to free will, we should start with just will. Will is a conscious and deliberate intention to do something, usually followed by the act of doing that thing. Like pain, will exists; You do consciously intend to do things and then do them. Whether your will is truly free or not is a further question. For the moment, it's important to note that will exists, whether or not it’s free.
With that in mind, let’s see if Sapolsky’s criterion for free will makes any sense. “What is needed to prove free will: show me that the thing a neuron just did in someone’s brain was unaffected by preceding factors.” He’s looking for free will in a neuron. But he’s gotten ahead of himself. There is no will at all to be found in a neuron, let alone free will. Sapolsky’s criterion for free will is like the following criterion for navigable oceans: show me the navigable ocean inside of any US state, or else there is no such thing as a navigable ocean. If we already know that oceans are real, but not found inside US states, then checking for navigability should not take place within US states. Likewise, if we already know that will is real, but not found within any neuron, then checking for free will should not take place within a neuron.
Sapolsky attempts a sort of reply to others who have made similar points before. He opens a whole chapter dedicated to the task by stating, “Reassuringly, no one thinks that free will lurks in the neuronal equivalent of individual bricks [as opposed to a palace made of bricks].” But later in the same chapter, Sapolsky bafflingly does exactly what he said “no one” does, as if we wouldn’t notice: he sets the criterion for free will at the level of the single neuron. “Individual neurons don’t become causeless causes that defy gravity and help generate free will just because they’re interacting with lots of other neurons.” This is ultimately how he checks whether or not free will could emerge from the whole brain, rather than from its parts severally: he checks the parts severally. I wish I were joking. Sapolsky does not understand the basic point made by those with whom he disagrees. But he admits as much: “A lot of people have linked emergence and free will; I will not consider most of them because, to be frank, I can’t understand what they’re suggesting.” What we’re suggesting is that you can’t tell much about free will or anything else about a whole person by looking at any single neuron. The “mechanism” that produces deliberative choices is the whole person.
So we are the authors of our own decisions. Being the source of one’s own actions is part of having free will, but there’s more to it, which Sapolsky also misunderstands. To fully have free will, we should be able to upend pure predictability. I can say that I’m the master of my fate, but if it’s possible to predict everything I’m going to do before I do it, then my credentials are dubious. So now we come to the issue of predictability. It’s central to the question of free will, and it’s the scene of our next self-contradiction by Sapolsky.
Some proponents of free will have tried to argue that because chaos makes human behavior unpredictable, our choices are actually free. Sapolsky disagrees because unpredictability is not the same thing as indeterminism. He insists that it doesn’t really matter how predictable or unpredictable behavior might be; the real—and completely different question is only about determinism, not predictability. And to Sapolsky, that question is answered by reference to the nature of the small parts that make up the bigger system: “Emergent systems can’t make the bricks that built them stop being brick-ish.” If there were a principle behind Sapolsky’s claim it would be that it doesn’t matter how predictable or unpredictable phenomena seem at a high level of emergent organization. What really matters is whether or not the smaller parts are deterministic or not; that’s where the real action is for Sapolsky.
William Tomos Edwards
Well, at least that’s his approach until two chapters later, when he needs to wave away the indeterminism that exists at the quantum level. He does so by doing exactly what he just decried. Suddenly, it’s predictability at the emergent level that matters to Sapolsky, not the question of whether the building blocks are deterministic or not. He quotes Jesse Hobbs:
The law of large numbers, combined with the sheer number of quantum events occurring in any macro-level object, assure us that the effects of random quantum-level fluctuations are entirely predictable at the macro level, much the way that the profits of casinos are predictable, even though based on millions of “purely chance” events.
Predictability does play a role in answering the question of free will, just not in the arbitrary and ambivalent way Sapolsky uses it. In the philosophy literature, free will is often defined as the ability to have chosen otherwise than one did. Sapolsky also appeals to this criterion. But there’s a complication that neither Sapolsky nor most philosophers have considered. There is a temporal asymmetry in the question of whether you could have done otherwise. The question in its typical formulation is backward-looking. It asks about what could have been in the past, and at first it seems like a coherent question. You did one thing yesterday, and we wonder if you could have done something else.
But what if we wanted to figure out whether or not you’ll have free will tomorrow? From that temporal angle, the question of the ability to do otherwise stops making sense. In a forward-looking sense, the question becomes manifestly nonsensical. Can you do otherwise in the future? “Otherwise?” Other than what? Other than the thing you will do? The question stipulates that you will do a certain thing, and simultaneously asks whether or not you can avoid doing that thing. The stipulation contained within the question makes the answer trivial. No, of course you cannot do something other than the thing defined as the thing you will do. In order for the question to have any meaning, it must be modified. The question cannot directly stipulate that you will do a certain thing. The question must ask whether or not you can do something other than what you could be expected to do, not other than what you will do. So the criterion for free will becomes about predictability if we’re talking about the future.
But this raises another question. In judging a choice to be free or unfree, should we be talking about the future (forward-looking) or should we be talking about the past (backward-looking)? We should use only the forward-looking formulation because a choice by its nature is forward-looking. We don’t deliberate or make choices about the past. Choice is always about something, and that object of choice always lies in the future, thus choice is always forward-looking. At the time when a choice is actually made, there is as of yet no “what” as in “Could have done other than what?” You have not already made the choice, so there is no established action to have done otherwise. So the only coherent way to judge the freedom of any choice is by its predictability.
As a criterion for free will, predictability can’t just refer to our current primitive forecasting abilities. To really mean something for free will, we need to talk about what would in principle be predictable given perfect knowledge of the present universe. The strongest conceivable form of prediction would be that of a being who knows the present masses, positions, shapes, temperatures, velocities, charges, spins, and all other physical properties of all particles in the universe; and has the ability to analyze these data. Such a being was imagined by Pierre Simon Laplace in 1820 as an illustration of his view of determinism. The imagined being has since come to be called Laplace’s demon. The ability to do something not expected by Laplace’s demon is the capacity for free will. This is how determinism is related to predictability and free will.
Sapolsky inadvertently appeals to Laplace’s demon in his attempt to rule out free will. As I mentioned, Sapolsky admits that none of the biology in the book could ever disprove free will, but he believes that a proof emerges from science as a whole. “You can’t disprove free will with a ‘scientific result’ from genetics or any other scientific discipline. But put all the scientific results together, from all the relevant scientific disciplines, and there’s no room for free will.” But no one has ever synthesized all of science, and no one ever will. Most of the relevant scientific results don’t even exist yet.
Here is a little example to illustrate the incompleteness of current science: Neuroscientists have taken precise measurements of the synaptic wiring in C. elegans, a roundworm 1 mm in length, with a total of 302 neurons. With those data, researchers built simulations of the 302-neuron network at the cellular level, including the roundworm’s entire muscular system and representations of its physical environment. Despite the comprehensive measurements and simulations, it remains unclear to researchers how the actions of 302 neurons combine to generate the behaviors observed in the C. elegans roundworm. Human brains have roughly 86 billion neurons. The number of possible relationships between neurons grows exponentially as the number of neurons grows linearly. To fully explain most human traits and behaviors would be impossible right now. So when Sapolsky says that “all of the relevant science” put together leaves “no room for free will,” he’s not talking about the paltry science that’s currently available and fits into his mind; he’s talking about all possible science, synthesized by a purely hypothetical entity. He’s talking about Laplace’s demon.
So what could give us the ability to surprise Laplace’s demon? Computational undecidability. This is a term describing a system that cannot be predicted, given complete knowledge of its present state. This fundamental unpredictability shows up in algorithmic computation, formal mathematical systems, and dynamical systems. Though an unpredictable dynamical system may evoke the concept of chaos, undecidability is a different sort of unpredictability. As described by one of the greatest living information theorists, C.H. Bennett:
For a dynamical system to be chaotic means that it exponentially amplifies ignorance of its initial condition; for it to be undecidable means that essential aspects of its long-term behavior—such as whether a trajectory ever enters a certain region—though determined, are unpredictable even from total knowledge of the initial condition.
If a system exhibits undecidability, then it is unpredictable even to Laplace’s demon, while a system that is merely chaotic is perfectly predictable to the demon. Chaos is only unpredictable because the initial conditions are not perfectly known. So it would be fair to dismiss that kind of unpredictability as mere ignorance—an epistemological issue, not an ontological reality. But the delineation between the epistemic and the ontic falls apart when we talk about what Laplace’s demon can’t know. An issue is “merely” epistemological when there is a fact of the matter, but the fact is unknowable. There actually is no fact about how an undecidable system will behave until it behaves. For a fact to exist, it must be in reference to some aspect of reality. But nothing about present reality could ground a fact about the future behavior of an undecidable system. In contrast, the exact actual state of present reality grounds facts about the future of chaotic systems. We just can’t know the exact actual state of present reality, thus unpredictability is “merely” epistemological in the case of chaos, but not in the case of undecidability.
Arguably, human behavior is undecidable, not just chaotic. And that would mean that human choice is free in exactly the way we’d want it to be; determined—by our own whole selves, with no fact of the matter of what we’ll choose before we choose it. But Sapolsky seems unaware of undecidability as a concept. He mislabels cellular automata as chaotic, rather than recognizing the truth that they exhibit undecidability. This is a major factual error on Sapololsky’s part.
Factual errors, in combination with his miscomprehensions and self-contradictions, make Sapolsky’s attack on free will ineffectual. Rather than forming a coherent understanding of how unpredictability and emergence relate to the question of free will, he denounces their invocations until he needs to invoke them himself. He doesn’t realize that knowing everything in the present would still leave the future unspecified, because he hasn’t heard of the concept of undecidability. Putting “all the scientific results together, from all the relevant scientific disciplines” still leaves a lot of room for free will. Sapolsky’s conclusions about morality and politics stand on nothing beyond his personal tastes. His book was marketed with such authoritative headlines as “Stanford scientist, after decades of study, concludes: We don’t have free will.” In contrast to the hype, Determined is ultimately a collection of partial arguments, conjoined incoherently. And Robert Sapolsky is to blame.
