Free Will and Sam Harris’ Puppet Analogy: A Reply to Iona Italia

I’ve been exchanging letters with Jon Rosen about free will and determinism on the platform Letter. A few days ago, I was happy to read Iona Italia’s thoughtful article about the exchange. I concur with Rosen, who took to Twitter to thank Iona for her “superb coverage,” which “impressively covered much of the material (some of which was quite tricky)” and introduced new insights to the exchange. To find out more about Letter, you can read Iona’s description of its aims here or sign up here.

At the end of this otherwise excellent essay, Italia represents an important compatibilist argument in a way that (to my mind) misses the mark. In what follows, I’ll make that case and—I hope—shed some light on the larger debate.

Determinism and Compatibilism

Determinism is the claim that what actually happens is the only thing that could have happened. More precisely, the kind of determinism at issue in the philosophical debate about free will is causal determinism. Causal determinism is the claim that, for any time T1 and any later time T2, the total physical state of the universe at T2 is the only way it could have been, given the total physical state of the universe at T1 and the relevant physical laws governing the chains of cause and effect.

Imagine a supercomputer, which can simultaneously track the current position and trajectory of every single molecule in the universe and which knows everything about the laws of physics—not just everything humans currently know, but everything there is to know. If determinism is true, such a supercomputer would be able to feed current positions and trajectories into its physical equations and use them to predict with 100% certainty where all the molecules would be in five minutes, five years or five thousand years.

Next, imagine that, five thousand years in the future, one man stabs another to death with a kitchen knife in a jealous rage. Let’s call the murderer Mark and his victim Warren. Mark, Warren and the knife used to commit the murder are all physical objects composed of molecules. (Even if you believe that Mark has some sort of non-physical soul, the hand gripping the knife is made of molecules.) Thus, if determinism is true, a hypothetical supercomputer would know what Mark was going to do to Warren thousands of years in advance.

Many people argue that, if Mark’s crime is the unavoidable (determined) result of chains of cause and effect that were set in motion long before he or Warren existed, it follows that Mark wasn’t exercising the kind of control over his actions (free will) that would be necessary for Warren’s death to be Mark’s fault (i.e. for Mark to be morally responsible for the action). But why not?

One obvious argument goes like this:

Premise One: If causal determinism is true, no one is free to do otherwise (i.e. no one has the ability to do anything they don’t in fact end up doing).

Premise Two: If no one is free to do otherwise, no one is in control of their actions in the right way to be morally responsible for them.

Conclusion: If causal determinism is true, no one is in control of their actions in the right way to be responsible for them.

People who endorse this argument are sometimes called hard determinists. Remember, though, that not every determinist is a hard determinist. Some determinists (i.e. people who think that causal determinism is true) think that free will and determinism are compatible. These are compatibilists.

A Quantum Loophole?

Before delving into how compatibilism works, it’s important to note that, according to the Copenhagen interpretation of quantum physics, the antecedent that causal determinism is true is (at least at the level of subatomic particles) false. The Copenhagen Interpretation holds that the basic physical laws governing the cause-and-effect behavior of particles are probabilistic, not deterministic. Albert Einstein famously rejected this claim with horror, declaring that, “God does not play dice with the universe.” But what if he does? If determinism is false, doesn’t that render the need for a compatibilist defense of free will moot?

It doesn’t—for at least three reasons. The first is that, as a philosophy professor points out in an early scene of Richard Linklater’s animated film Waking Life, it’s not clear that this kind of indeterminism helps with free will. If you’re ultimately a physical being whose actions are jerked around by random physical processes outside your control, are you any freer than one whose actions are smoothly guided by deterministic processes? As Jerry Coyne puts it more bluntly in a recent article in Quillette, “we can’t use our will to move electrons.”

Secondly, there’s a logical leap from the behavior of protons and electrons is ultimately indeterministic to the behavior of human beings composed of protons and electrons is ultimately indeterministic. This is a form of what logicians call the composition fallacy—it’s like saying that because none of the molecules that make up the Brooklyn Bridge are visible to the naked eye, the Brooklyn Bridge itself isn’t visible to the naked eye. Even if the Copenhagen Interpretation is correct, there’s no evidence that quantum indeterminacy filters up to the level of human behavior.

Finally, there’s something deeply strange about the idea that we can’t know whether we’re in control of our actions in the right way for those actions to be our fault unless we know which interpretation of quantum mechanics is correct. Do we really need to bet on the outcome of esoteric disputes between physicists in order to know whether a murderer is morally responsible for his crimes? Compatibilists believe we have free will (or at least the kind of free will that matters for moral responsibility) whether or not determinism is true. As John Martin Fischer likes to put it, they thus have the advantage of being able to explain autonomy and moral responsibility in such a way that these important concepts don’t “hang on a thread.”

Defining Compatibilism

Compatibilism is the position that, even if causal determinism is true, there’s still such a thing as free will. Many people find this position so confusing when they first hear it that their instinct is to reinterpret it in a way they find more reasonable. Thus, I always have students in my introduction to philosophy classes who think that what the compatibilist must be saying is that, even if some or even most human actions are determined, there are exceptions—indeterministic islands of free will, surrounded by a sea of determinism.

This is a misinterpretation. Compatibilists are saying something much more interesting and surprising. The compatibilist’s claim is that, even if everything is determined—including all human thoughts, decisions and actions—some of those (completely determined) human decisions are also free, at least in the sense of free that matters for moral responsibility.

So How Do Compatibilists Respond to the Argument Above?

Compatibilism encompasses a rich and diverse family of views, espoused by figures as philosophically different as David Hume, Gottfied Wilhelm Leibniz, Daniel Dennett and Kadri Vihvelin, but probably the most common contemporary compatibilist strategy is to deny Premise Two. Compatibilists of this kind—sometimes called semi-compatibilists—don’t claim that residents of deterministic universes have free will in the sense that they’re free to do otherwise. Instead, they claim that even people who aren’t free to do otherwise can be free in the sense of exercising the kind of control over their actions that matters for moral responsibility.

In her essay, Italia correctly summarizes the views of semi-compatibilist philosopher John Martin Fischer:

Imagine, he suggests, that a man called Green has had a device implanted in his brain through which neuroscientists can monitor his brain activity and intervene to change his intentions, if necessary. Green is walking past a pond when he spots a drowning child and dives in to save her. However, if he had not done so, the neuroscientists observing him would have flipped a switch that made him decide to go to her rescue. Should Green still be praised for having saved the child? Yes, Fischer answers.

Unfortunately, though, she doesn’t explain why Fischer thinks that Green is morally responsible in a positive sense (praiseworthy) for saving the child. Instead, she immediately switches focus to the views of Fischer’s fellow semi-compatibilist, Harry Frankfurt.

This shift in focus might make sense. The story about Green is a version of the sort of example originally developed by Frankfurt. (Such examples are generally called Frankfurt cases.) Unfortunately, though, Italia doesn’t explain the point of Frankfurt cases. Instead, she skips to explaining Frankfurt’s theory of autonomy, which is that we’re free as long as our actions line up with our higher-order desires (roughly, we’re not just doing what we want to do, we’re doing what we want to want to do) and relates this to Sam Harris’ uncharitable summary of compatibilism: “A puppet is free as long as he loves his strings.”

The Point of Frankfurt Cases

One common objection to compatibilism goes something like this:

Isn’t your idea of freedom just playing with words? You say that there’s a way of understanding the word that makes it possible for an action to be both free and determined. Well, maybe there is. Words and phrases like free and can and able to are a bit ambiguous. (Think about the old joke about the man with his hand in a cast, who asks his doctor whether he’ll be able to play the piano when the cast is taken off.) But I don’t care about those semantic games. I care about the kind of free will I thought I had before the discussion started.

Well, at least one of the ideas that we pre-philosophically have about free will is that we can only be morally responsible for our actions—we can only deserve praise or blame for them—if we performed those actions of our own free will. The point of Frankfurt cases is to show that we can have this kind of free will, even if we don’t have free will in the sense of being free to do otherwise.

Note that this isn’t an analogy. The idea isn’t that the neuroscientist waiting in the wings to interfere with Green’s brain chemistry if he needs to—and remember that he doesn’t interfere because he doesn’t need to—is somehow similar to deterministic processes of cause and effect resulting in actions in the normal case. The idea, rather, is that, if you can acknowledge in this hypothetical that Green is free in the sense of performing the action of his own free will and therefore is responsible for it even though he isn’t free in the sense of being free to do otherwise (if he’d been about to do otherwise, the neuroscientists would have intervened), then you’ve acknowledged that these two kinds of freedom are different and that it’s at least possible for them to come apart. Once you’ve acknowledged that, you can’t just help yourself to the assumption that Premise Two is correct when you’re arguing for the incompatibility of determinism and free will.

Frankfurt and Fischer on What Free Will Is

To be a semi-compatibilist is to say that the freedom to do otherwise isn’t the kind of free will that matters for moral responsibility. This is a purely negative claim, though, and it’s compatible with various very different theories about what kind of control does matter for moral responsibility.

We’ve already seen Frankfurt’s answer. That answer has at least a few things going for it. It helps us explain, for example, why we think that adult human beings who aren’t suffering from mental illness or severe cognitive deficiencies are responsible for their actions and dogs, for example, are not. Dogs don’t have the mental complexity required to have higher order desires—desires about desires.

Even so, there’s something quite unsatisfying about Frankfurt’s answer. Remember that everyone—Frankfurt, Fischer, other kinds of compatibilists, incompatibilists, everyone—would agree that Green wouldn’t be morally responsible for his action if the neuroscientist had altered his brain chemistry to make sure he saved the child. Green at that point would be no more praiseworthy than a robot programmed to save drowning children. Frankfurt can explain this by saying that if someone is being manipulated to want to act in a certain way, they’re not acting in the way they want to want to act. But what if the neuroscientist had done a more thorough job and altered Green’s brain chemistry such that he not only wanted to save the child but wanted to want to save the child? If Frankfurt would be forced to say Green is free and responsible in that case, Sam Harris’ puppet analogy might be apt.

Let’s turn then, to Fischer’s theory of free will. Here’s how I summarized this view (and an important twist on it) in a recent article in Quillette:

Fischer’s idea is that, even in a deterministic world where we aren’t free to do otherwise, we have ‘guidance control’ over our actions just so long as we’re appropriately ‘reasons-responsive’—that is to say, we understand reasons for and against possible courses of action and we’re at least somewhat moved by such reasons. Building on Fischer’s idea, Ryan Lake argues in his doctoral dissertation, No Fate But What We Make: A Defense of the Compatibility of Freedom and Causal Determinism, that the reasons that move us must be our reasons in order for us to be exercising the guidance control necessary for moral responsibility.

To get a better sense of Lake’s idea here, think about property rights. If I find a precious stone on a beach that no one has laid claim to, it’s fine for me to take possession of it. If I find the same kind of stone in a friend’s pocket, it would be wrong of me to take it without asking. In both cases, the stone comes from outside of me. I didn’t make it. The difference is that in the first case, I’m the only person with a claim to it. Similarly, if the mindless operation of natural laws playing out through a complex combination of genetic and environmental factors results in my weighing the pros and cons of a possible course of action in a certain way and coming to a particular decision, the reasons that motivated me are my reasons. If a mad scientist’s chip in my head brings about the same result, they are not. In both cases, my decision is determined by factors outside of me, but only in the first case do I have a unique “claim” on my reasoning.

When applied to this version of compatibilism, Harris’ quip simply doesn’t apply. It doesn’t matter whether Green loves his decision. What matters for moral responsibility is whether that decision came about in the right way.

Why All of This Matters

A final point is in order about why it matters whether we have this kind of moral responsibility. After all, one of the reasons (though hardly the only reason) we care about moral responsibility is that it seems to be linked to the justification for criminal punishment. Sam Harris and Jon Rosen, however, have both suggested that we can make sense of punishing dangerous criminals without bringing moral responsibility into the equation. Italia accurately summarizes the views of my sparring partner, Rosen (which closely resemble the views of Sam Harris): “Criminals must be punished, Rosen stresses, in order to protect others and as a social deterrent.” Such considerations have nothing to do with moral responsibility. Indeed, this purely utilitarian way of thinking about criminal punishment may seem more humane and enlightened than one that focuses on criminal retribution.

I’d argue that this appearance is misleading. To see why, consider the case of a town on the edge of a race riot. Members of one group are furious about the death of one of their own in an apparent hate crime. The local authorities conspire to frame and execute an innocent member of the other group in order to prevent a riot in which many more people would have died. If our only concern is harm prevention, this is justified.

To think this would be wrong, you don’t need to go all the way to a purely retributive view of punishment. You can have a two level rights forfeiture view. According to this theory, innocent people have a categorical right not to be punished (even if it would be good for public safety and order if they were punished). When people commit certain kinds of crimes for which they’re morally responsible, though, they can forfeit this right. This doesn’t mean they deserve punishment. It does, however, mean that they’ve forfeited the right not to be locked up for the greater good.

Note, however, that this two-level view relies on a distinction that turns on moral responsibility. As such, this option is closed off to thinkers who deny that we have the kind of free will that matters for such responsibility.

Of course, it doesn’t follow that this is correct. It could be that our enlightened liberal ideas about criminal punishment simply aren’t compatible with the grim metaphysical truth. What I have hoped to show, though, is that the debate has real stakes.

Letter is a digital platform for one-on-one correspondence. It combines the intimacy of letter writing with the convenience of an online format and the added value that conversations can be publicly read and shared. Our subeditor, Iona Italia, works with Dayne and Clyde Rathbone to make this possible. To find out more, go to www.letter.wiki or contact humans@karma.wiki.

If you enjoy our articles, be a part of our growth and help us produce more writing for you:


  1. The important point is that it’s not just moral responsibility that’s at stake here — it’s voluntary behavior of any kind, including that involved in writing this essay, reading it, and responding to its arguments. And there are no “quantum” loopholes — quantum phenomena involves random behavior, but random behavior is no better than determined behavior in getting a handle on voluntary behavior, or agency. The simple fact is that, from a purely empirical orientation, we’re all simply machines, even if stochastic machines. So in terms of what Ben calls the “grim metaphysical truth”, we have a comical (not so grim!) vision of a Ben-machine that can’t help but write this essay and a lot a reader-machines that can’t help but read it, and a few commenter-machines that can’t help but react, all in this absurd Rube-Goldberg perpetual motion.

    All of which suggests that there may be another orientation than the empirical — namely the moral. “Compatibilism” in this sense just comes down to the idea that human social structures require more than one orientation toward experience in order to function.

  2. What is the significance of having or not having free will? Apparently, it’s “Moral Responsibility”. And what is the definition of Moral Responsibility? None is given, but apparently, it’s an excuse to punish someone more severely than the utilitarian principles discussed above (protect society from further offenses from the offender through either rehabilitation or imprisonment, and deterrence of others from similar offenses) would allow. In other words, we don’t need to feel guilty for imposing a cruel punishment!

    The race riot example provided by the author is a terrible, silly straw man. No one has claimed that our only concern is harm prevention. This has nothing to contribute to the discussion about whether there is such a thing as free will.

    Ray Andrews – the claim that “only a moral agent can be reformed” is meaningless. Whether or not we have free will has nothing to do with whether our behavior can be changed by external inputs, such as education and deterrence. And whether or not we can be certain someone will or will not reoffend is not a strategy we SHOULD use to define punishment protocols for large populations of offenders. That’s a straw man as well.

    1. @Chris H

      “Ray Andrews – the claim that “only a moral agent can be reformed” is meaningless. ”

      I experience it’s meaning very clearly. That you do not see something is not proof that it does not exist.

      “Whether or not we have free will has nothing to do with whether our behavior can be changed by external inputs, such as education and deterrence.”

      I agree that unfree mechanisms might be repaired — in the case of organisms they might be conditioned by external inputs, but in the case of someone with moral agency and free will, they might simply *choose* to modify their behavior.

      “And whether or not we can be certain someone will or will not reoffend is not a strategy we SHOULD use to define punishment protocols for large populations of offenders.”

      Then what would you suggest? If not the danger of re-offense, then what should guide parole boards and what is the purpose of imprisonment at all? Should those likely to re-offend be paroled? If not, then what is the basis for a parole board being sure that the mechanism will not re-offend?

  3. The debate between free will or determinism is not based so much on what we observe, but on how we can understand, the universe.

    Using very simple examples, we can predict at what angle a pool ball will move if hit in a certain spot. But moving too far beyond this, we operate in the area of “best guess.” For example, if I say tell me the weather over Chicago ten days from today at exactly 3PM … we would be forced into approximations. And the more complex the question or the farther out in time we go, the less reliable our answers become.

    So, though we believe that these things are predictable and deterministic, we cannot precisely determine them ourselves Determinism is a way to understand the world.

    Free-will deals with subjective feelings that originate within us, where their origin seems unknown. Here we have an even greater IN-ability to predict causes and effects.

    So what has occurred is that we have evolved two ways to understand the world.

    One deals with physical world and inorganic outside of ourselves that we understand through cause-and-effect and that we presume to “KNOW” to be deterministic. Even though we cannot decipher everything exactly, it is how we approach external reality.

    And we are also provided a second way to understand the internal world of subjective reality. We call this FREE WILL.

    The concept of free-will simply means that an organism is going to act or react in ways that seemingly cannot be pre-determine. With subjective causes that are and will remain unknown to us and maybe unknown even to them as well.

    These are also how we divide concepts of “thinking” versus “feeling.”

    When we THINK about how we can FEEL we are using our “deterministic” outlook to try to understand the subjective.

    Human moral systems develop to address both circumstances.

    This subject is addressed in a new book titled “Our Human Herds and the Theory of Dual Morality.”

  4. “Indeed, this purely utilitarian way of thinking about criminal punishment may seem more humane and enlightened than one that focuses on criminal retribution.”

    I’m not so sure. If we are mere mechanisms it is true that I have no moral responsibility for what I do, but then again, having no responsibility means that we also can’t be sure I won’t rape another child if I am released from prison, so it follows that I can never be released. Only a moral agent can be reformed, a mechanism cannot. Thus unless the justice system can insure that the mechanism (me) that raped a child is make permanently incapable of doing it again — a tall order — then … well, since we have no souls and are just mechanisms, what do we do with defective equipment that can’t be repaired? We throw it away. And history does show us that enlightened, post religious philosophies like communism are very casual about throwing away those who are considered defective, no? So, ironically, retribution presumes moral guilt which presumes agency which presumes free will and a soul which *can* reform themselves and thus be forgiven.

    1. Morals are social constructs that can be associated with a society even within a deterministic universe. Within such a universe, in a technical sense we aren’t responsible for our actions, but they can still impact others. And the actions of others – including punitive or rehabilitative actions directed towards us for crimes we commit – can influence us to alter our behavior in the future. In a deterministic universe, every action that someone takes is the only action that they possibly could have taken; in that sense it would not be justified to hate someone for his harmful actions, but at the same time it still does make sense for the actions themselves to be condemned and for steps to be taken to correct the future actions of the offender. Induction of moral guilt in the offender could help to achieve this end. In a practical sense, morality operates exactly the same way within a deterministic universe as it would within one having free will.

      1. I’m roughly in agreement. It seems to me that free will may be a social construction, but it is one that we can’t live without. Even hating someone for their actions may simply be a social construction designed to maximize the conditioning of the individual not to do that again — this hating and conditioning also being determined of course so there’s nothing about it that could change.

        But I think there is still a shade of difference: If I, as a moral free agent, give my word that I won’t do some bad thing again, and if the listener also believes I am a free agent who makes choices, then the listener might just believe me. But if I am a mechanism, and I am not in control of myself (my future being already determined), then that guarantee is void ab initio — it *cannot* be given. Thus parole should be given much less often.

        So it is might be best to construct ourselves as free moral agents even if we aren’t because our determined belief in FW does help us to behave better. That is, it is determined that people who do not believe in determinism are less likely to misbehave. As I said, FW tho an illusion is identical to the real thing, so I see no reason to make the distinction. There is no such thing as green, but the illusion of green is exactly like green, so I just call it green — there is no useful dichotomy between green and the illusion of green, they are identical in practice. In practice, that I should right now be considering what I’ll do tomorrow seems necessary and if it’s an illusion it is an unavoidable illusion and that makes it real.

Leave a Reply