On a recent Youtube episode of Theories of Everything with Curt Jaimungal Daniel Dennett, author and Professor of Philosophy at Tufts University in Massachusetts, discusses his views of Robert Sapolsky's recent work around free will. Sapolsky has been doing the rounds of webcasters to promote his new book, Determined: a Science of Life Without Free Will amd is Professor of biology, neurology, neurological sciences, and neurosurgery at Stanford University.

Dennett states (00:00:10):

"Robert makes one big mistake and it's right at the centre of his work...What he's missed is the difference between control an causation".

Dennett then describes a boulder crashing down a mountain. He states the boulder is caused to do this by "...wind pressure, bumps on the way, material, etc, etc".

"Now, if we rewound the tape of life and replayed the tape of that exact circumstance, it would roll in exactly the same way. That's determinism".

He then compares the boulder's journey with that of a skier who skis down the mountain. The skier differs from the boulder, claims Dennett, in that she is in control.

"Now the path of the boulder is determined. The path of the skier is determined. The difference is that the skier's path is controlled. The rock's path is not".

"If you make the distinction between control and causation, then Sapolsky's argument simply falls apart".

[To give some insight into what Sapolsky's position is - and I hope I'm doing him justice here - he invokes a determinism that extends back to the Big Bang in order to deny free will].

Dennett continues:

"...it doesn't go back to the Big Bang. Because once life gets started...life generates controllers. It generates self-controlled little agents. Things that are alive".

He then goes on to focus on the notion of predictability and how "indeterminism is not required for unpredictability". He states that being unpredictable - whilst still being determined - gives us freedom as agents in the sense that it protects us from other agents. He also states that "...determinism doesn't prevent you from acting locally, and making your own choices and not being controlled by the past".

This is where he confuses me. He is asserting control from determined agents. [I have a feeling this what he came to loggerheads with Sam Harris about in their pub debate following Harris's release of his book 'Free Will').

How can there be control within a determined sequence of events? Isn't this merely the illusion of control?.

"Evolution pseudo-randomly makes a lot of mutations and the ones that survive survive because they do things that the competition doesn't do, and this", he says, "is what gives them a kind of control; a kind of reliability".

Jaimungal asks Dennett what Sapolsky's reaction to his basic position was. Dennett paraphrases Sapolsky as saying, "No, no. You're still determined by the events going back to before your birth".

He then goes on to give a 'parable of the bathtub' as a means of describing that we "...are reliable in spite of variations in our past". I can only assume that he is referring obliquely to the patterns of behaviour we form as identities that have evolved over time, but I still don't see how this gets us any closer to free will.

Am I correct in concluding that Dennett and Sapolsky are at odds in relation to very different views of free will? Ie: That Sapolsky is denying the free will that would somehow overcome determinism whereas Dennett is happy with a free will that bestows an illusion of control?

  • I don't think that humans have words and concepts sufficient to ever resolve the question to everyone's satisfaction. I actually don't think that humans ever will, so it becomes yet another pointless debate. Nonduality resolves it nicely.
    – Scott Rowe
    Jan 1 at 13:29
  • 1
    @scottrowe. Maybe. Maybe not. It might make things more difficult if the leading minds can't agree on terms. Then again, maybe the diversity of approaches might make it more likely that we'll stumble upon the trutg. Jan 1 at 13:33
  • 1
    I seem to recall Sapolsky saying basically that we have no free will but we have to do all the things we are doing either way, for example: criminals can't help committing crimes, so they don't 'deserve' to be imprisoned, yet we can't have them out there harming people. If a concept leads to the same result whether it is true or false, don't we have a word for that? inoperative or something? Let's start using that word. A lot.
    – Scott Rowe
    Jan 1 at 13:48
  • To give some insight into what Sapolsky's position is - and I hope I'm doing him justice here - he invokes a determinism that extends back to the Big Bang in order to deny free will - WHAT?! I must be misunderstanding something here, because that claim is blatantly false. It presupposes that classical mechanics can explain the entire Universe, which was disproved a long time ago. Randomness entailed by quantum mechanics might not allow free will - still they introduce nondeterminism?! One cannot say that all events have been determined since the Big Bang!
    – gaazkam
    Jan 2 at 14:01
  • 1
    It's possible that no small part of the disagreement is just semantics
    – Him
    Jan 2 at 14:59

4 Answers 4


Great question, well posed.

Dennett is the clearest champion of Compatibilism about free-will. Many scientists take from the fact human actions must be reducible to atoms and the void, that we only need the latter for a full account, and any appearance of causal power by agents is so subsumed by the causal narrative of atoms that we can call free will an illusion. This is like saying literature is an illusion, only letters and words are real.

Biology is constituted of vast chains of contingencies. And we can consider genes as an encoding, an image, of a lineages evolutionary niche. Similary when a large-language-model predicts a next word, it does so not from words and letters, but from a low resolution image of their use across the entire internet. You need to know the abstraction in the system, which shaped the control algorithm, to predict the system.

David Krakauer is Professor of Complex Systems at the Santa Fe Institute. I feel his term 'teleonomic matter' can most clearly reveal why these kind of systems are different. He uses it to describe what the domain of Complex Systems Theory is: that it's systems which get information about their environment, record it, and change their dynamics in some way as a result.

This can provide a truly abstract account of what subjectivity. And crucially, of what an observer is in quantum mechanics, where an observation makes the observed part of the observers wave function, the information about which state quantum system is in goes from being isolated from the observer, into the observer's system. The mixed or coherent state goes from being in the wavefunction of the observed quantum system, effectively into the 'wavefunction of the universe'.

It's an over simplistic view of a deterministic universe, to think that the information that fixes or determines future outcomes is distributed evenly everywhere. Quantum systems make it very clear that is not the case. In the same way, the information to predict the skier is disproportionately localised in their control systems, not the shape of the mountain.

A way to picture the physical significance of an increasingly more accurate control algorithm, is to think how the best abstraction of physics could allow energy life and the technology tree to be recovered even the poorest resources, like a patch of interstellar dust, using only very minimal initial resources like data, a fusion device, and a Von-Neuman machine.

I would point to the similarity to high-sensitivity-to-initial-conditions of classical deterministic systems like say three-body systems, which requires measurements of initial conditions that become increasingly difficult, and ultimately can go below the Planck scale, the limit imposed on investigation by the point where any probe would generate a blackhole. See: Gargantuan chaotic gravitational three-body systems and their irreversibility to the Planck length. Similarly improving control algorithms, can amplify sensitivity to initial conditions, to the point where a patch of cosmic dust could do incredibly improbable things.

  • Right, the image of the skier is kind of cool (having skied in the past, I can feel in to it). The motions they make might be more related to music they are listening to or recalling than anything about the external world. I remember singing sometimes when I worked on an assembly line. What about the environment made me sing one song rather than another, or sing at all for that matter? "I know why the caged bird sings"
    – Scott Rowe
    Jan 1 at 22:36
  • @ScottRowe: They might be focused on completing the slalom faster, rather than simply finding a stable equilibrium on route to maximising entropy. For a stone all we need is the latter to predict it. For the former, goals & intentions & an implementation of control towards attaining them are a far better predictor, even though as we know for the total system the 2nd law of Thermodynamics is not being broken.
    – CriglCragl
    Jan 1 at 22:47
  • 1
    "Let the 2nd law of Thermodynamics / Remain Unbroken / By and by, Lord / By and by..." It's got a beat, you can dance to it, I'll give it a +1
    – Scott Rowe
    Jan 1 at 23:12
  • "Many scientists take from the fact human actions must be reducible to atoms and the void, that we only need the latter for a full account, and any appearance of causal power by agents is so subsumed by the causal narrative of atoms that we can call free will an illusion." I find this statement dubious, unless by "many" you mean "more than five".
    – Him
    Jan 20 at 10:35
  • @Him: Specifically here I meant Sapolsky, Einstein, & the many other scientists who explicitly deny the reality of free will. Not exactly amateurs...
    – CriglCragl
    Jan 20 at 11:53

Both debaters are wrong. Our universe is not determined. PARTS of it may be mostly determined, but QM is indeterminate, and chaos theory shows how uncertainties can leverage into the macro scale.

Dennett's conception of a controlling skier is partly correct. Once we have an idea in our heads of what to do, then the guiding of our bodies IS deterministic.

But the idea itself, if one is a physicalist, AND committed to the random/determined dichotomy, is most plausibly seen as the result of a semi chaotic process, where QM effects in the interaction of our neural system produce random events, which then get sorted by the inherent structure of our neurology into poor, good or better choices. MAYBE this is what Dennett means. But then he would be a semi=randomist, not a compatibilist determinist.

If one is a spiritual dualist, or an agent emergentist, then the origin of ideas, OR the selection between them, can have a spiritual or emergent plane component too, which would matrix with the random then bounded/sorted chaotic neurology.

  • Dennett explicitly rejects this kind of 'free will of the gaps'. Whether addressing deterministic or random systems, they obey laws, which can account for all the causation. Dennett is a Compatibilist, who accepts this but also says we can meaningfully talk about agents with control systems as parts of the physicalist-materialist causal processes.
    – CriglCragl
    Jan 2 at 13:07
  • 1
    "QM is indeterminate" this is only the Copenhagen interpretation. There are other interpretations of QM that are deterministic.
    – Him
    Jan 2 at 14:57
  • @CriglCragl -- a semi-randomist can accept bounded randomness that satisfies physics in chaotic systems that our bounded neuro-inclinations then react to deterministically. I don't know any other way to argue for a skier being any different from a rock.
    – Dcleve
    Jan 2 at 14:59
  • @Him -- the only interpretation with any appreciable support that is deterministic is Bohmian Mechanics -- and that has been failing several tests cases vs Coopenhagen. Everett isn't deterministic -- what outcome you will see for a quantum event is still random, even if other "you"s will see other outcomes. Everett is just Coopenhagen with other worlds added too.
    – Dcleve
    Jan 2 at 15:05
  • 1
    @Him Link: settheory.net/Bohm Also philosophy.stackexchange.com/questions/68224/… Bohm has a lot of problems, most notably that it is theoretically uncalculable, as every particle in the universe has causal influence on every reaction, but much of our universe is outside our light cone, so we have no idea what particles are even there. Your Physics SE link was mostly referencing Wikipedia, which is an unreliable reference.
    – Dcleve
    Jan 2 at 16:40

"...determinism doesn't prevent you from acting locally, and making your own choices and not being controlled by the past".

This statement makes no sense. Determinism means that everything is caused/controlled by the past. There are no agents capable of self-determining their own actions.

How can there be control within a determined sequence of events?

You have reached the main problem with Dennett's compatibilism: How an action can be determined by both the agent herself and the past events? I don't know how he explains this, but it is quite obvious that he must redefine both free will (=agent determines) and determinism (=past events determine) beyond recognition.

There cannot be any control within a determined sequence of events. Nor can there be any evolution or anything random (pseudo- or true).

The problem with both gentlemen is that they are both trying to shoehorn determinism into reality.

  • Yes, and the shoe doesn't fit, because determinism is too small. Cinderella has big feet.
    – Scott Rowe
    Jan 1 at 22:39
  • 1
    Why can there not be anything pseudo-random? (I would define that as deterministic, but sufficiently obscure as to appear random to an observer unaware of the algorithm, but perhaps you are using a different definition.) Jan 2 at 0:50
  • @Oddthinking Pseudo-randomness is not deterministic as pseudo-randomness is literally fake randomness. Determinism cannot fake anything, faking requires free will. Jan 2 at 4:19
  • I disagree on both counts. Pseudo-randomness is fake randomness precisely because it is deterministic (but doesn't appear to be). More importantly, if faking implies free will, and fake things exist, then there must be free will, and the entire issue is resolved! Jan 2 at 7:48
  • @Oddthinking You don't seem to understand the concept of randomness. In a deterministic system there is no randomness at all, fake or true, everything is determined with absolute accuracy. True randomness requires a truly stochastic process. Fake randomness requires someone to deliberately decide the values in order to give an impression of randomness. Jan 2 at 9:06

Dennett's thoughts are consistent with the overall philosophical landscape and with free will too, as we understand it that is. Sapolsky has a point too and his skiing example is apposite.

At times like these I suggest we employ our knowledge or wisdom, if possible, and, let's not forget, imagination. There's a boulder, there's a skier; both on a hilltop. What follows has a structure to it and we conduct, either a gedanken experiment or carry out a scientific experiment.

  • I think there is a good point to be unpacked there. The skier has a mental landscape of possible futures, the processing of which according to control systems like 'skill acquired at skiing' will shape the actions of the skiier. A rock will fall in different ways depending on how similar we can make the initial conditions & the landscape; but it will be a far smaller landscape of variations than repeated runs of the skiier.
    – CriglCragl
    Jan 2 at 13:13
  • @CriglCragl, bingo! We're in good hands, oui monsieur? Jan 2 at 16:51

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .