George Ellis responds to my criticism of his argument for free will

Yesterday I posted a critique of an Aeon article by physicist George Ellis, arguing that science itself gives evidence for true libertarian free will. This rests on his claim that psychology exerts a “top-down” effect on molecules, and those top-down effects, because they stem from our thinking, our experiences, and our personalities (all subsumed under “our psychology”), constitute libertarian free will. (He didn’t say exactly how the top-down stuff gives a non-physical “agency” to people, but merely suggests that it’s a way to think about it.)

For once most readers agreed on my take, mainly because those readers who aren’t hard determinists like me still accept the laws of physics, while Ellis seems to argue that the “top down” influence on our molecules, and hence our behavior and “choices”, cannot be reduced to physics, and in fact is free of the laws of physics.

My response was brief. Your personality and character—the “top”—are formed by changes in your brain induced by your genes, your environment, and all the experiences you have. And those changes are ultimately molecular changes that affect neurons.  And those changes obey the laws of physics. There is no “top” free from the laws of physics. (I add here that Ellis won the Templeton prize for harmonizing science and religion, which may go some way towards his promotion of a free will that seems quasi-religious, but certainly seems dualistic.)

Reader Steve commented on that post, saying “I asked Ellis to read your post and reply. Here’s what he said: https://aeon.co/essays/heres-why-so-many-physicists-are-wrong-about-free-will?comment_id=33240″

I consider this an inadequate answer, but I won’t engage with Ellis’s ad hominem argument that “it’s a typical Jerry Coyne response.” I’ll address his comment on the “core issue.” And that is Ellis’s claim that physical things like electrons and “psychological” things like feelings, emotions, and behaviors, are two different things; in fact, they are two entirely different things, and you can’t understand “psychology” using molecules.

Perhaps we’re not at the stage where we can predict the effects that the environment or brain molecules have on behavior, but we’re not completely clueless, either. Should Ellis doubt this, ask him to imbibe a few stiff bourbons and see if there aren’t predictable results. Or give him a course of testosterone and see how it affects his behavior. As many people, including me, have indicated, there are plenty of experiments showing that one can affect one’s decisions, one’s beliefs,—and, indeed, one’s sense of agency—through physical manipulations of the brain, whether they be by experimenters or disease.

In contrast, as Sean Carroll has emphasized repeatedly (see the article and tweet below) not only that there’s no evidence for physics-free top-down causation, but, indeed, there’s evidence against it from the laws of physics. There is no way we know of for nonphysical thoughts to influence physical processes.

The solution, of course, is the parsimonious and evidenced idea that thoughts and feelings are the results of the laws of physics, combined, of course, with the evolution that helped program our brain to (usually) behave adaptively.  When Ellis says “the true statement is that electrons interacting allow and enable the thoughts to take place at the psychological level,” he might as well say, “the electrons (and other particles and molecules) are what make the thoughts take place at the psychological level.” Then the influential “top” goes away.

Here are some writings by Sean Carroll on the intellectual vacuity of downward causation. Click on screenshot below.

 

46 thoughts on “George Ellis responds to my criticism of his argument for free will

  1. His last paragraph seems to kill his own argument. Correct, electron flows enable higher-scale algorithms to run, and algorithms in turn modify electron flows. But this doesn’t create free will in a computer, so why would it in us?

    There’s nothing magical about feedback loops between different scale phenomena that allow them to avoid determinism.

  2. His discussion of downward causation in the paper he linked (which is written by him, of course), seems to simply assert that downward causation exists.

    His claim that particle physics “must also be shaped in a downward way by engineering and biological variables” necessarily implies that particle physics is wrong, and I believe Sean Carroll’s argument in his blog post on the immortality of the soul would apply equally to this. How do these “engineering and biological variables” alter the standard model Lagrangian? How does an electron know when it is being affected by “engineering and biological variables” and when it is not, so they may behave accordingly? If it is the trivial claim that free electrons behave differently from electrons in molecules, then I think it hardly deserves the moniker of “downwa causation”.

    -Ryan

    1. It does count as such (re: electrons in molecules) – the boundary conditions are as important as the laws. (See my other post.)

      But that doesn’t allow spooky-stuff to exist either.

      1. That can be explained entirely by electrons being affected by the electromagnetic field of atomic nuclei. I don’t see how it is molecules that cause electrons to behave differently rather than the configuration of the electromagnetic field, and that isn’t downward causation.

        -Ryan

        1. Yes and no – if the field is meant as in “everything to do with it”, then *maybe*, but I am not so sure. Consider bond energies: a carbon-hydrogen bond has a different strength if the carbon also is bonded to oxygen. This is sort of a boundary condition to the reactions (say) that might result. The whole electromagnetic field does not play a role, at best a localized one, which I think would render the two descriptions equivalent.

  3. The fact that there is downward causation of sorts appears obvious when we consider the different levels (categories, if you like) we talk about; physics is different to biology is different to psychology.

    What is less obvious to me is how it’s argued that these different levels of ’emergence’ are not entirely caused by the underlying physics. Just because the murmurations of starlings ’cause’ things to happen doesn’t mean the starlings aren’t causing the murmurations. So I’ve never quite been able to see past this conundrum to accept the ‘something not physical is happening’ brigade. There always appears to be a *presumption* that the higher level thing is definitely not physical.

    And if the downward causation is non-physical acting on the physical then why wouldn’t we be able to harness an alternative energy supply? Why don’t the proponents of this non-physical energy build a downward causation power station? They would make a fortune.

    1. But if there were a thought power supply, does it get charged up by not thinking?

      Bada-boom! I will be here all week…

    2. Exactly. Just because the emergent properties provide useful explanations that are *much* more concise than atom-by-atom explanations (“weak emergence” in philosophy speak) doesn’t mean that atoms can suddenly violate their usual rules when they link together into a larger object (“strong emergence”).

      The sad part of Sean Carroll’s commentary is that he ignores the original definition of “downward control” by Roger Sperry, which was only weak emergence. Of course it wasn’t long before the woo-meisters got hold of the term and abused it to mean a strong emergence idea. Perhaps we should speak of “weak downward causation,” I dunno.

      1. A good summary of an important difference.

        Though I wouldn’t pan Carroll for not making the distinction. If he’s responding to woo-meisters who only use ‘downward control’ to mean strong emergence, then it’s reasonable to think he was mirroring their language back to them for purposes of clear communication.

  4. Therefore, the cosmic consciousness is in everything. Why would it not be? I think Ellis is one Chopra from walking off the Deepity end!

  5. This is correct: top-down causation (or rather, at least, determination sensu Bunge) is routine (J. Kim notwithstanding). But it is also not mysterious, and it would be *very* weird if it involved submergence of conservation laws (and we would have noticed by now). *This* is what Carroll means, I think.

    However, what set me off was not the original author’s appeal to downward causation, but was the fact that he talks as if emotions, etc. were somehow different from the activity of neurons and such in animals (in this case). *That* is the mistake. The emergence can occur, and emotions are new processes with new properties not of their components, but they are *made up* of the lower levels. In the case of the multiple realizability – this is a *type* claim, as Paul Churchland pointed out years ago (20 at least – he did this at PSA 2000 by the latest, when I heard him point it out.) In any *given* case there is a realization, not somehow that individual animals have a “non-realized implementation”.

    Also, DC materialist or otherwise doesn’t solve the problem!

    1. *This* is what Carroll means, I think.

      I dunno, he seems very insistent about the “all interactions are local” thing. I think Carroll might look at Ellis’ computer example and point out that the “algorithm” which Ellis treats as a separate higher order thing is just a set of atoms and electrons in a certain configuration, and each responds deterministically to a local perturbation, one after the other.

      I think the snowflake analogy isn’t very intuitive. I’d use “falling dominoes” instead. You can arrange dominoes into massively complex patterns that do all sorts of crazy things. Spell out names, turn on machines, etc. But ultimately, each domino’s behavior is only governed by the one before it hitting it and knocking it down.

    2. This. Unfortunately, while you and I were growing older, a bunch of woo peddlers came and semi-successfully hijacked the term “downward causation”. There are now a bunch of respectable philosophical thinkers (Sean Carroll included) who think that “downward causation” implies strong emergence. So (reasonably enough given that etymological blind spot) they reject it.

  6. I think Ellis is usurping phenomena which are not predictable and delineating them as evidence for free will.

    If we have no capability of either explaining or predicting emergent, complex behavior, like that found in condensed matter systems, then it would appear that those physical states would be indistinguishable from states generated by free will.

    Ellis references a few of Tony Leggett’s papers. One, in particular, ‘On the Nature of Research in Condensed State Physics’ suggests that complex solid state phenomena are, themselves, to be considered fundamental physics. This is in contrast to quantum mechanics or high energy physics, where single particles or single modes are well understood. But I don’t think the gives Ellis anything but a reasonable argument that complex systems are really important, if not more fundamental than simple conservative systems, like quantum mechanical interactions.

    Bottom line: Ellis can’t escape the physical universe, no matter how complex or unpredictable and regardless of what he chooses.

  7. “Downward causation enables algorithms to control electron flows.”

    I suggest that Ellis is now confused over what “downward causation” means.

    Take electrons in a computer running a program. If you took a complete low-level description of the system, in terms of electrons etc (and ignored the high-level description in terms of human-readable computer code) would that low-level description be sufficient to predict the state of the system at time-plus-1?

    I think the answer is yes (and if it’s not yes then Ellis needs to overturn a heck of a lot of physics). But if the answer is yes then there is no downward causation.

  8. I think of the ‘downward causation’ argument as another ‘Great Chain of Being’ razzle dazzle.

    I contend that there are no ‘levels’ in nature although I accept that we use the concept of levels to make organised thinking possible by throwing away vast amounts of data. But some philosophers (and some physicists!) are so in love with the idea of ‘reason’ as the topmost level they reify our thinking tools into real structures in nature. And it’s a short step from there to requiring the ‘top levels’ to have agency and causation on ‘lower levels’. Teleology by the back door.

  9. Leaving aside the free will debate and the mind-body problem, there’s nothing spooky or anti-naturalist about what Ellis is saying here about what he calls downward causation. An algorithm (his example) is a “real pattern” (Dennett) that can be admitted into our ontology since it’s just as real as the electrons that constitute its physical instantiation. It isn’t just “a way of talking,” as Sean Carroll might put it. The algorithm explains the flow of electrons, so is downwardly causal in that sense, and it’s not usefully described or understood at the electron level, plus it could have other physical bases.

    There’s no non-physical substance in play, and nothing mysterious going on – everything obeys the laws of physics. It’s just that those laws aren’t all there is to perspicuous understandings of either algorithms or psychology. Higher level patterns, laws and regularities emerge non-mysteriously in nature and technology, and play their own (possibly ineliminable) role in causal explanations of what’s going on at lower levels, e.g., electron flows, neural spike trains.

    1. “There’s no non-physical substance in play, and nothing mysterious going on – everything obeys the laws of physics.”

      If everything obeys the laws of physics, then a low-physics description of the physics of the system is enough to predict the behaviour of the system. If that is the case then there is no downward causation.

      “Higher level patterns, laws and regularities emerge non-mysteriously in nature and technology, and play their own (possibly ineliminable) role in causal explanations of what’s going on at lower levels, …”

      If they play ineliminable causal roles then that means we need to overturn swathes of physics. Because it tells us that predicting the low-level behaviour using current physics would not work.

      1. “If everything obeys the laws of physics, then a low-physics description of the physics of the system is enough to predict the behaviour of the system.”

        I’m not sure this has been formally proven, has it? (my question in yesterday’s thread on Ellis)

        “If they play ineliminable causal roles then that means we need to overturn swathes of physics. Because it tells us that predicting the low-level behaviour using current physics would not work.”

        Things can simultaneously obey both the laws of physics and other higher-level laws, so I don’t see that downward causation requires overturning any physics, and I don’t think Ellis says or implies this.

          1. Jerry: “Which ‘higher level laws’ are you talking about? Please be specific.”

            Here’s one you cited: “..the laws of physics, combined, of course, with the evolution that helped program our brain to (usually) behave adaptively.” There are higher-level regularities (you know them well) that characterize natural selection, and these contribute to causal explanations of organisms and their behavior. Can physics alone explain and (in principle) predict speciation and the resulting diversity of organic design? (honest question)

            Coel:

            “If the particles are obeying the laws of physics, then the account in terms of higher-level causes is eliminable, since the laws of physics would be sufficient to tell you what the particles are doing.”

            I’m not sure how you know this to be true. Has it been formally shown? The laws of physics can certainly explain and in principle predict the behavior of say the micro-components of neurons and logic gates and perhaps their local interactions. But it isn’t obvious (to me) that higher-level regularities involved in biology and computational design aren’t needed to explain the arrangements of neurons and logic gates in various systems, and the resulting behavior of the systems as a function of those arrangements. Such behavior might well affect the status (e.g., spatial location, charge) of the micro-components.

            Perhaps, contra Ellis, it *has* been shown that none of the higher level patterning (entities, laws) has any causal autonomy, in which case the laws of physics are in principle all that’s needed to predict the location of a specific molecule in my left hand in a half hour. I don’t know if it’s been proven or not.

            1. This is my last response. Evolution is not a law, it’s the description of a process. There is no “law of evolution” like there are laws of physics. In answer to your honest question, I’ve already answered it: prediction is different from pure determinism, and insofar as evolution and speciation depend on quantum phenomena (e.g. mutation), then evolution and speciation may not be perfectly predictable. But they’re certainly consonant with the laws of physics, and in fact result from them. So no, you’re wrong in saying that evolution is a “higher order law”. It isn’t.

            2. The assumption is that fundamental physics is “causally closed”. You can take it as a metaphysical stance, since there is no proof for it, just overwhelming empirical evidence that it is the case. As Sean Carroll would say, finding out that current physics is not causally closed would be an outstanding discovery

              1. I’m not sure that the plausible assumption of causal closure is equivalent to the claim that in principle only the laws of physics are needed to account for higher-level behavior. Those laws obviously hold for the micro-constituents of complex systems like ourselves, but it isn’t obvious that one needn’t invoke higher-level biological and psychological regularities to account for the status (e.g., location) of those micro-constituents, since that depends on the behavior of the system as well as the laws of physics. It’s all physical and causal, of course, any quantum indeterminism aside.

              2. If you use the words “need to invoke”, you don’t have causal closure. You could say that you can “also invoke“ the other causes, in which case you are advocating for “causal completeness”, which is a weaker form of causal closure (see discussion at https://www.iep.utm.edu/causal-e/)

        1. “Things can simultaneously obey both the laws of physics and other higher-level laws, …”

          If the particles are obeying the laws of physics, then the account in terms of higher-level causes is eliminable, since the laws of physics would be sufficient to tell you what the particles are doing.

          1. Coel,

            Not to disagree with any of the above, but how do you deal with dualities in physics? Example: the Ads/CFT correspondence. Either of the dual theories can be eliminated, in favor of the other. So the fact that an account is eliminable … doesn’t prove quite as much as one might think.

            1. Physics can quite often describe something in two different ways, with both descriptions being true. In such cases, one can pick one of them, and use that alone (the other is then “eliminable”, though that doesn’t stop it being true or useful; you could equally adopt that one and eliminate the other).

              An ineliminable description would means that the other description cannot suffice. It would mean that one cannot use physics to predict the state of a physical system at time t+1. That would overturn swathes of physics.

              1. Exactly. Or that one could not use physics to predict the state at t-1 – which would also overturn some well-confirmed rules of physics.

    2. I’d suggest you read Paul Tourek’s reply to post #4.

      If Ellis means weak emergence, then you are correct there’s nothing anti-naturalist about his position. However in that case, his position also doesn’t support any notion of libertarian free will, which, very clearly, Ellis thinks his own argument does support.

      OTOH, if Ellis means strong emergence, then you are wrong in saying there’s nothing anti-naturalist about his position. But this position would seem to provide a way for his argument to support libertarian free will.

      Long ago, Jerry made a comment that some social scientists use overtly complicated language to try and mix a mundane but true observation with an unusual and surprising observation that was false, to get people to believe they had discovered something unusual and surprising that was true. This is kinda how I feel about this issue. You can’t argue weak emergence -> free will. But you also can’t argue science supports strong emergence. However, with sufficiently opaque language, you might be able to convince yourself and others that emergence (handwave type handwave!) is supported by science and supports free will.

      1. “However in that case, his position also doesn’t support any notion of libertarian free will, which, very clearly, Ellis thinks his own argument does support.”

        Yes, which is why I started my comment with this: “*Leaving aside the free will debate and the mind-body problem*, there’s nothing spooky or anti-naturalist about what Ellis is saying here about what he calls downward causation.”

  10. “Thoughts are not the same as electrons interacting with each other” and “to claim they are identical is wrong” so Ellis as if someone had claimed this and demanded this equation (obviously that is a pure straw man argument).

    In response to these absurd statements from Ellis, it is perhaps best to modify Carrol’s formulation from: “the single molecule has no idea it is part of a snowflake and could not care less”, to : The single electron has no idea it is part of a psychological level or of a thought and couldn’t care less.

  11. “I rest by the science that supports emergence of higher levels with genuine causal powers….

    To claim [electrons interacting and thoughts] are identical is simply wrong (inter alia because of multiple realisation of higher level phenoma at lower levels). Neuroscience texts are rather different than physics texts.”
    —George Ellis

    1. What he “rests by” is a questionable ontological model of nature called strong ontological emergentism, according to which there is a hierarchy of different and irreducible levels or layers of existence/reality with different and irreducible new kinds of higher-level entities (particularly emergent properties and causal powers) that aren’t composed of any lower-level entities.

    2. Physical reductionism can plausibly and successfully deal with the multiple-realizability objection by appealing to domain-specific type-identities or token-identities.

    3. Yes, of course, “neuroscience texts are rather different than physics texts;” but it doesn’t follow that the higher-level entities constituting the subject matter of psychology, neurology, physiology, biology, or chemistry are ontologically noncomposed of and nonreducible to lower-/lowest-level entities belonging to the subject matter of physics.

  12. I thought from yesterday’s post that Ellis was proposing a psychological form of Lamarckism. Except instead of a phenotypic trait like a long neck we end up with free will expression of behaviour, like “sticking your neck out”.

  13. It is somewhat ironic that the biologist argues in favour of (classical) physics being ‘supreme’ in this matter, whereas the physicist against.

    I feel at sea re free will, and just don’t trust much my own opinion, despite making one remark way below.

    I do claim to know enough about physics to state confidently that George Ellis has been a very major positive force in physics/cosmology for 50 years. Back in the 70s, the book with Hawking “Large Scale Structure of Spacetime” was and continues to be very important. Not enough credit is given to the fundamental discoveries of Penrose and Hawking in the 60s, that black holes are not artifices of artificial models of the universe, but actually follow logically from General Relativity and a couple of very reasonable assumptions about energy and about no time-like closed curves violating causality. Penrose deserves far more credit on this than many seem to realize, and that book was good on that point IIRC.

    But it does seem to me that Ellis has let his religious feelings get in the way of better thinking on free will, and maybe other things earlier at times. It’s not just about being 80. He seems still very cogent and energetic.

    And some of the disagreement here about downward causation appears to result from unclear meanings of words, more than real scientific disagreement.

  14. What exactly is the difference between Ellis’s “psychological level” and what religious people call the “soul”?

  15. I find it odd, but interesting, that a physicist would propose a theory of causation that would make all of physics and chemistry total chaos.

Comments are closed.