Sunday, October 6, 2013

Against Physicalism

Are humans more than just a complicated physical machine? Physicalism is the idea that the physical is "all there is" - everything that exists is either physical or is reducible to something that is physical. Thanks to Ed Feser's blog, I've come across an interesting argument from James Ross that physicalism cannot be true. Ross takes an approach that draws on Kripke, Goodman, and Quine to build a rather astonishing claim about physical systems. Ross's argument is very subtle and worth a close look. If it succeeds, it's truly an astounding accomplishment: one of the great philosophical debates of all times, solved. I don't think it succeeds, and I'm going to try to suggest why not.

Feser summarizes Ross's argument like this:

All formal thinking is determinate.
No physical process is determinate.
Thus, no formal thinking is a physical process.

Specifically, Ross refers to "pure functions" that humans can define but that cannot be implemented by any purely physical system. He gives examples like adding, squaring a number, and the modus ponens of logic.

Now, what makes Ross think that a physical system cannot add? Of course he knows that mechanical devices and computers are capable of performing sums, but he says they are only simulating addition, not truly adding. He writes:

 Whatever the discriminable features of a physical process may be, there will always be a pair of incompatible predicates, each as empirically adequate as the other, to name a function the exhibited data or process "satisfies." That condition holds for any finite actual
"outputs," no matter how many. That is a feature of physical process itself, of change. There is nothing about a physical process, or any repetitions of it, to block it from being a case of incompossible forms ("functions"), if it could be a case of any pure form at all. That is because the differentiating point, the point where the behavioral outputs diverge to manifest different functions, can lie beyond the actual, even if the actual should be infinite; e.g., it could lie in what the thing would have done, had things been otherwise in certain ways. For instance, if the function is x(*)y = (x + y, if y < 10^40 years, = x + y + 1, otherwise), the differentiating output would lie beyond the conjectured life of the universe.

Now, I can go along with Ross as far as the epistemological aspect of his conclusion: no matter how many input-output pairs we examine, we can never know what function is being computed. But Ross claims much more: he says physical systems are not just epistemologically indeterminate but "physically and logically" indeterminate, too. That is, it's not just that we can't know what function the machine is computing, but there really is no fact of the matter about what the output will be until it actually happens.

The problem is, the argument Ross gives is not up to the task of proving that claim.

First of all, what does Ross mean by "empirically adequate"? He is not using this in the sense of van Fraasen, for whom empirical adequacy means agreement, not just with all past observations, but with all possible observations. For Ross explicitly mentions a "differentiation point", possibly at some remote future time, at which the outcomes disagree. Nor does he mean "agreement with all future observations", for the same reason. So he must mean merely "agreement with all past observations."

But having two hypotheses that agree with all past observations is not enough to tell us that the physical system is actually (physically) indeterminate. It only says that our information is insufficient to distinguish between the two.

Another example Ross gives is the problem of determining a function, knowing only a finite number of data points. He (correctly) points out that there is an infinity of curves that will agree on those finite data. But this just says we don't know what the function is that produced the data. It doesn't follow that there is no such function at all. But that's what Ross needs for his conclusion that physical systems are not just epistemically indeterminate, but physically indeterminate.  

Ross's other arguments draw on Goodman's and Quine's work. These, too, also only reach as far as epistemology. Goodman's grue problem suggests that we can never know for sure whether we are inducting on the right categories. But it is a long way from that epistemological claim to the claim that there are no correct categories for induction to work on. Duhem's claim about undertermination says only that we can't know what part of whole complex of assumptions, theories, and practices is at fault when an experiment disagrees with theory. Again, this is only an epistemological claim. True, Quine tried to extend this uncertainty to the whole realm of human knowledge - but this extension hardly helps Ross's claim that humans can add, employ modus ponens, etc. Thus,none of Ross's arguments, either in the article or in his book, Thought and World, take us beyond epistemological indeterminacy.

I have to say it is exceedingly odd to see Feser defending physical indeterminism here. In our discussions of quantum mechanics and causation he argued strenuously that there is no such thing as physical indeterminism - not even in the case of quantum mechanics (where nearly all physicists accept fundamental indeterminism). So I'm wondering how Feser can square real, physical and logical indeterminism with the principle of causality.


  1. Well...

    I can't say what Ross or Feser establish (I haven't read the paper). But it does appear to me reasonable and indeed novel to argue that physical indeterminism implies that "there is no such function at all" - that the output isn't produced by some determinate algorithm at all. Since the real "algorithm" that produces the output, the laws of physics, produces variable output while the abstract algorithm does not, it is clearly not the same algorithm that's being run. But equally, it isn't any determinate algorithm. An indeterministic world cannot, in principle, "run" any standard (and therefore, determinate) "pure function" algorithm.

    It should be noted that this is a variant of the more general Argument from Reason. And that as it relies on indeterminism, its rather more limited in its scope than the more general Argument from Reason (which basically says that you can't implement any algorithm in any mechanistic world, since what you're implementing is the rules of physics rather than the algorithm).

    My standard response to the argument from reason is that "close enough" is close enough. No, our brains don't really, at the fundamental level, do the function of "adding" or "modus ponens". But they are built so they effectively do, including our thinking about and defining these concepts. This response applies to this argument as well - it doesn't matter that there is a chance to produce a result not in accordance with the pure function, it's still the case that we're effectively running an algorithm that implements it.

    So - I agree with Feser/Ross that physical systems "are only simulating addition, not truly adding." And I think Ross's argument is novel (to me, at least). I just don't agree with the black and white nature of the Ross/Feser argument. The conclusion should be "To the extent that a physical system is indeterminate, it cannot implement formal thinking". They're missing the (50?) shades of gray.

    I haven't touched the other points raised in your post, but that I think is the core of it. One more general note - "physicalism" is generally taken to include also things that supervene on the physical, even while not being reducible to it. Thus, in many physicalist theories of mind the mental content supervenes on the physical content - meaning that certain physical structures/dynamics imply certain mental feelings / events - but the mental content is not fully reducible to the physical content, it's a separate thing (how a thing feels like rather than how it behaves, for example). This point is important to me since it means I, as a panpsychist (in the philosophical, not religious, sense) still count as a "physicalist". If you only allow theories of mind that actually reduce the mind to the physical to count - which is usually labeled "materialism", BTW - then I would have to wear a different label. :)


  2. Hi, Yair,

    I think your response is a possible route to take, but I wanted to focus on the fact that Ross hasn't established what he needs for his argument. It's not clear to me if your insistence on indeterminism has anything to do with quantum phenomena. Ross's indeterminism doesn't (he never mentions QM in the article, and has only a passing remark on it in the book.) So I would like to ask you, do you think QM makes deterministic machines impossible?

    (I don't recall if we've discussed this point before - apologies if we have. Apologies too for mischaracterizing physicalism.)

    1. I haven't read Ross, so can't say what he's established.

      My own argument does rely on physical indeterminism such as that of QM.

      QM makes making fully indeterministic machines.... difficult to construct. Not impossible, but difficult. You basically have to construct everything to work off deterministic variables, such as the total spin (J^2), rather than the spin along a certain direction (J). You can then ignore the physical indeterminism, having isolated yourself from it. This is exceedingly difficult in practice, however, as things like atoms vibrating, gamma rays passing through your machine, and so on are all probabilistic. I don't know if it's really possible to construct or even design a machine that will be fully deterministic.

      What we're talking about here, however, are thought experiments. Any machine may have a miniscule chance of producing varying results due to quantum indeterminacy. But in practice, this can be so small to be totally negligible. Throw a ball at a wall, and it might have a miniscule chance of tunneling through it by QM; but waiting for this to happen in practice will take much more than the lifetime of the universe. In practice, it's fairly easy to build machines that are so deterministic, they can be treated as ones. Including the machines in our own head, that allow us to think up and discuss deterministic concepts in the first place. That's kinda my point against the argument from reason.

      (And no apologies needed. It's all good. :) I don't think we've discussed things before, and I'm not too worried about the labels I carry :) )


  3. I assume you mean " fully deterministic machines" rather than " fully indeterministic machines".

    I was thinking about electronic computers, in which the level of discrimination between 0 and 1 is far greater than could be changed by a few extra quantum tunnellings. But I suppose you would say there is still SOME probability, even though it might be very small....

    I suppose one could try to run a proof along these lines:

    - Definite eigenstates are an idealization not achievable in the real world.

    - Apart from definite eigenstates there are only probabilistic outcomes.

    - Therefore, no physical system is completely deterministic.

  4. Yes, that seems like a sound argument to me. Of course, it's an argument within physics, not philosophy. The philosophical aspect is that it implies that no physical system implements classical "pure functions", which are deterministic.

    Note that one can easily define pure functions that are indeterministic. And in that case a mirror image of the theorem emerges - these cannot be implemented in a deterministic world. But this line of argument can say nothing about implementing deterministic function in a deterministic world or indeterministic functions in indeterministic worlds.


  5. Do not expect Feser to be consistent.

    He is essentially just an apologist.

    If X being true suits his purpose today, then of course X is true. But tomorrow X being false is what he wants. So X is false.

  6. Thinking more about your OP - I think you're spot on. Assuming you've faithfully representing Feser/Ross's argument (still haven't read it), I couldn't agree more: the Problems of Undertermination that they raise are merely epistemic, saying nothing about whether physical processes actually embody pure functions; and it's important to consider what's plausible, not only what's possible.

    I was focusing on your last comment, which associates them also with physical indeterminism a la QM (not just with undertermination). Hence my above argument. If it isn't included in their work - well, I think they missed a novel argument that they were very close to.



  7. Feser's response to this is interesting in that it undermines his own A-T philosophy. But Ross's proof is easily rejected.

    Suppose Jack approaches Jill in a bar. He makes advances. Jill wonders if this guy is married. But she doesn't want to ask for ID and perform a credit check. So she invents a logical proof:

    1) All married men wear a ring.
    2) No single man wears a ring.
    3) Thus no married man is a single man.

    If you think this proof should set Jill's mind at ease when she sees Jack wears no ring, then you stand with Ross and Feser. For that is the structure of Ross's proof. But for me, this so-called proof is pointless. And that should be obvious.

    But let's look closer anyway.

    We are to assume: "All formal thinking is determinate." So let me ask, what does "1+1=2" mean? What is "determinate" about it? What does 1 stand for? An apple? An orange? What about x+y=z+1? What does that mean? The truth is, we cannot determine any meaning from mere formal language. Numbers and variables are designed to be replaced with not-so-formal items later in the game. The equation, a^2 + b^2 = c^2, is not determinate in the sense Ross and Feser claim it is. It describes not one triangle, but any triangle. And given that equation, it doesn't necessarily describe triangles at all. It might describe tomorrow's weather for all we know. We humans attach our interpretation to those equations every bit as much as we attach them to the output of calculators. Feser wants us to believe meaningful, "determinate' equations are floating in space. We just reach up an grab them and declare, "Wow, this one means energy and mass are relatively the same! But it doesn't work that way. We create the formal language. It was specifically designed to help find and describe meanings in the physical world Feser finds so indeterminate. He's confused about what a tool is and how we use it. Instead he gives our invention a strange detached and deified status.

    We are to assume "No physical process is determinate." This merely begs the question. The very thing in question is this: Is formal thinking a physical process? So how can we assume "no physical process is determinate" when we don't yet know if formal thinking is a physical process? We can't. My assumption is that all thinking, including formal thinking, is a physical process. Begging the question is of no help.

    You would think professional philosophers would be able to see these obvious flaws in their arguments. It's too bad a amateur such as myself has a better grasp on their own subject.