Neural correlates of free will

Ken G

I can see where you are coming from on the first two issues, but the basis for the final claim just depends on what one thinks determinism really means. I would say determinism is quite demonstrably an analysis tool, not a description of how things happen (and you might be saying something similar, but then we cannot see that what happens is "deterministic", we can only say what happens admits to useful analysis via determinism). Indeed, when we attempt to use determinism as a literal description of how things happen, it invariably breaks down at some point along the way. Even when we say we are sure an apple will fall when we drop it, we are speaking in terms of probabilities-- it is highly probably the apple will fall, but we don't know that something else could happen we did not expect, like a bomb might go off that blows the apple upward instead. Probability is always an assessment of what you don't know as well as what you do, and determinism is an analysis tool that is intentionally blind to this fact.

Randomness is also an analysis tool-- my point is merely that being able to predict an outcome with high success is an example of the usefulness of the concept of determinism, not an example of a deterministic process. If you decide what movie you will see today, and I know you quite well, I might be able to write in an envelope what you will choose, based on my knowledge of you. That means I can determine your choice with high success rate-- it does not mean you are not exercising free will. The issue of determinism is nothing but predictability, and is quite orthogonal to issues of free will.
Here I would agree with you.

apeiron

My feeling is different having studied precisely this question of how the brain "wills" actions. We already know more than most people could ever want to know.

I would just say pick up Luria's The Working Brain, published in 1973, and read chapter nine. The broad outlines were worked out 50 years ago, and the gaps have been filled in by electrophysiology and animal studies much more than neuroimaging. Read Graybiel on the striatum or Passingham on the frontal lobes for example.

The neural correlates of freewill are one of the "easy problems" even if you are a Chalmer-ite by persuasion. But who really reads neuroscience textbooks?

nismaratwork

I guess wherre you see filled gaps, I see them as bridges to ever widening gaps in our knowledge... we know a lot, but not enough to really explore what the mind is. Well... we can explore, but not in what strikes me as a meaningful way.

Oh, and... nerd that I am, I read them... I read and read them, often for fun. So... that's me... that's a serious bias on my part I guess.

apeiron

Yes, determinism like randomness is in the eye of the beholder . It is how the world looks when it is reduced to its simplest alternatives.

The question then is how do we model complexity. It could be that it is just determinism~randomness made more complicated. Or it could be that in creating the simple model, we left out the "something else" - a story about the global constraints - which is what models of complexity require.

nismaratwork

See... this I agree with completely.

Lievo

Sure, but we can construct models that, by definition, are deterministic, and see what happens. That's exactly what I did: I constructed a deterministic model in which the same kind of problem can arise despite it's neither tied to consciousness nor free will. That says nothing about whether consciousness and free will are or are not determinist. That just shows that determinism is not at the root of the problem while interpreting Libet's finding.

That said, I'm not seeing determinism and randomness as just usefull tricks to guide interpretation. To me this has a precise meaning in terms of theory of computability and theory of complexity. I equate determinism with computabilty, and randomness with BPP class of complexity.

Let's begins with the latter: about everyone thinks that P=BPP, meaning that randomness is unlikely to provide any observable change from a more classical universe (that remains to be proven, however). That's exactly the situation with many-worlds versus Copenhagen interpretation: the first is purely deterministic without randomness, the second uses randomness to a large extent, and it does not make any difference in what we expect to see.

The former is more subtile: yes one will never prove that the universe is computable/determinist. However, the reverse (the universe being uncomputable/non deterministic) is IMHO theorically provable (can you compress most arbitray binary strings? If yes congratulation: you have hypercomputing abilities). So the question, to me, is not whether we can prove that the universe is deterministic. The question is: should we think otherwise when otherwise is such an extraordinary claim? To me extraordinary claims are good to Occamise until we find reasons not to.

apeiron

But BPP assumes determinism (the global constraints are taken to be eternal, definite rather than indefinite or themselves dynamic). So no surprise that the results are pseudo-random and Ockham's razor would see you wanting to lop off ontic randomness.

In the short run view, where global constraints by definition look "eternal", this is very valid and useful as a modelling approach. But it does not answer the larger case of the long run view where global contraints may be presumed to vary over time. Even the laws of physics could have evolved.

Real complexity modelling involves allowing the global constraints to develop, to self-organise. It is this intrinsic holistic dynamism that a strictly localised view, based on the standard dichotomy of random vs determined, misses.

Ken G

Exactly-- you constructed a model in which the same kind of problem can arise. Does that mean it is what happens in free will? Certainly not, your model does not exhibit free will. That is the Catch-22 in your argument-- you say computers are deterministic, so what they model is deterministic, and then you claim that free will has to be deterministic. But by making a deterministic model, you have not demonstrated free will, and you cannot tell that you have modeled free will. That is my point-- free will may have nothing to do with determinism, neither produced by it nor precluded by it. And none of your models answer that issue. I believe apeiron is making a similar point.

All the same, you said that we were talking about a deterministic system when we were talking about the brain. The issue is one of definition-- if by a "deterministic system" one means "a system that we gain limited predictive power by modeling it deterministically", then sure we can say the brain is deterministic. But most people's claims about "deterministic systems" require that the system is deterministic, i.e., it's behaviors are determined in advance, which is a very different claim, and not well substantiated by fact-- any better than fact can substantiate that weather is deterministic. Instead, the most straightforward interpretation of the facts is that it is not-- unless we restrict to the weaker meaning of the term.
Note those are both aspects of models of real systems, not aspects of real systems. The issue here is what evidence you have that your models are successful at modeling free will. What evidence is that?
But it is not an extraordinary claim at all, the more extraordinary claim is that the universe is built to submit to our analysis. More simple is the claim that we tailor our analysis to achieve goals, and the universe is just the universe, and a brain is just a brain. The ultimate irony is when we think that our brains our built to understand how our brains are built.

Lievo

Are you sure you don't mix-up my argument with those of someone else?

Didn't I explicitly said the same things? Again, my analogy says nothing about whether consciousness and free will are or are not determinist. That just shows that none are at the root of the problem while interpreting Libet's finding, because one can explicitly construct the same kind of result while evacuating both free will and determinism.

I'd say it's mathematical definition. Whatever. What's is important is that from these mathematical definitions we can infer whether this or that properties lead to predictions. If an aspect of the model cannot lead to prediction, then you have the mathematical guarantee that this properties is not important to care about. If it allows some prediction, then you can check reality to decide which kind of model can or cannot describe reality: with or without the property?

From the mathematical definition of randomness, an informed guess is that either randomness isn't at the root of free will, or free will can account for nothing. From the mathematical definition of computability, you can infer that either free will is determinist or it allows hypercomputing. So if one find evidence for hypercomputing that'd be evidence against determinism. Notice hypercomputing doesn't mean unpredictability. It means extraordinary abilities. See Penrose for one who defends this line of though, and especially defends that mathematicians have superpowers.

Some would disagree.

nismaratwork

Science is a method, it's no guarantee that the universe is comprehensible.

Q_Goest

Hi Ken G,
I understand what you're getting at, but chaotic systems are clearly defined as deterministic in the literature as I've quoted above. Yes, they are mathematically deterministic. Are they physically deterministic? When looking at the 'weather' or any other fluid system for that matter, we use statistical mechanics to define the fluid's momentum, density, internal energy, etc... at any point and at any time, and to the degree those values are accurate, the model will make accurate predictions. The fact that a fluid's momentum is made up of an aggregate of molecules and those molecules are being lumped together means that we can never be perfectly accurate. But does that really matter? Does it really matter that after an extended period of time, even our most accurate measurement of the macro states won't provide sufficient detail to define the micro states and thus the sensitivity to initial conditions might again cause a deviation from our model? I suppose one could also argue that given sufficient information about the micro states of molecules in the fluid, one could debatably predict the system with even higher accuracy, though I won't go that far. So are you suggesting that physical determinism isn't possible because we can't know the micro states, or are you suggesting that there might be some kind strong emergence and thus a form of downward causation that subordinates local physical laws? Or are you suggesting such systems aren't deterministic for some other reason?

apeiron

Have you found time to read this great paper yet?

http://arxiv.org/abs/0906.3507

You will see that Franks makes the argument that it does not matter whether the microscale is ontically random or ontically deterministic because it is the global constraints (the information preserved at the global scale and which acts top-down) which explains the patterns of nature.

We already knew this of course. You can generate fractals either by deterministic iterative equations or suitable stochastic processes. It looks the same in the end as what matters is the information represented as the global constraints.

But Franks makes this explicit. There is a top-down view which is not reducible to the bottom up. The whole is more than the sum of its parts (whether they be random or determined). And this is true even for simple systems (like those with a gaussian, or even simpler(!) powerlaw, statistics). It is of course obviously true for complex systems like life and mind.

Q_Goest

Hi apeiron,
I’m not familiar with “holonomic” so I did a search:

It sounds like Pattee wants simply wants these macromolecules and genetics to have a stronger causal role in evolution but I'm not sure exactly what he's getting at. Perhaps you could start a new thread regarding Pattee and his contributions to philosophy and science.

Sure, Baranger's paper is pretty basic, but it clearly makes the point that chaotic systems are deterministic given precise initial conditions, which is relevant to the OP. I think it’s important also to separate out chaotic systems that are classical (and separable) in a functional sense, such as Benard cells, from systems that are functionally dependant on quantum scale interactions. Our present day paradigm for neuron interactions is that they are not dependent on quantum scale interactions, so it seems to me one needs to address the issue of how one is to model these “non-holomonic” properties (classical or quantum mechanical influences) and whether or not such a separation should make any difference.

This is a good example of what confuses me about everything you say about this "systems approach". Are you suggesting these "top-down constraints" are somehow influencing and subordinating local causation? That is, are you suggesting that causes found on the local level (such as individual neuron interactions) are somehow being influenced by the top down constraints such that the neurons are influenced not only by local interactions, but also by some kind of overall, global configuration? Or are you merely referring to how boundary conditions act as local causal actions at some 'control surface' such as we use in multi-physics approaches that use FEA to model physical phenomena in engineering and the sciences? Note that FEA and similar approaches are simplified versions of the underlying philosophy surrounding the more conventional “systems approach”, that nonlinear differential control volumes must be in dynamic equilibrium over time. It’s this dynamic equilibrium between local causes that might somehow be misconstrued as there being some kind of genuine downward causation which of course, isn’t a mainstream concept. Being an engineer, I’d readily accept that boundary conditions act on any given system, but the underlying philosophy of how those boundary conditions act on any classically defined, separable system, does not allow for nonlocal causation and thus does not allow for downward causation in any real sense of the term.

After rereading his paper, I’d say that he does in fact try to separate mental states (phenomenal states) from the underlying physical states as you say, but that mental states are epiphenomenal isn’t an unusual position for computationalists. Frank Jackson for example (Epiphenomenal Qualia) is a much cited paper that contends exactly that. So I’d say Farkus is in line with many philosophers on this account. He's suggesting mental states ARE physical states, and it is the mental properties that are "causally irrelevant" and an epiphenomenon (using his words) which I’d say is not unusual in the philosophical community. Not that there aren’t logical problems with that approach. He states for example:
That says to me, he accepts that neurons only interact locally with others but we can also examine interactions at higher levels, those that are defined by large groups of neurons.

There are some areas in his paper I’m not too sure about. Take for example:
If he’s suggesting that this “higher level” is not determined by the interactions of the lower level (their interactions) in a deterministic way based only on the local interactions of neurons, then that sounds like strong downward causation which is clearly false. Certainly, there are people who would contend that something like that would be required for “free will” or any theory of mental causation. But I’m not sure that’s really what he wants.

In another questionable section he states:
In the part emphasized, I’d say he’s trying to suggest that a person is somehow “immediately” and “simultaneously” affected by a “global state” on entering this classroom which I picture as being a zone of influence of some sort per Farkus. Were the same person to enter the same room and was blind and deaf, would these same “global states” immediately and simultaneously also affect that person? Sounds like Farkus wants his readers to believe that also, but that sounds too much like magic to me.

I suspect that the punchline to all this is that the proposal these folks are after is that higher order levels influence the future higher order levels by influencing lower order levels. That of course is strong downward causation. I don't see any room for a 'medium' causation that somehow doesn't allow a higher level to influence a lower level but still allows higher levels to have some kind of influence. The higher level is made up of lower level constituents, so if there's no change in the lower level constituents caused by the higher level, there's no change.

I think this is a good lead in to strong emergence and strong downward causation which, in one way or another, is necessary for mental causation and free will. The question really is, can the higher physical levels somehow subordinate the local interactions of neurons? And if so, how?

Ken G

I did conflate your argument with Q_Goest, my apologies. Yes, and I agree with you-- Libet's finding really doesn't say much about free will at all, it says something about how we come under the conscious impression of having free will. That might be something quite a bit different from free will, just as the conscious impression of getting burned by a stove is quite a bit different from the process of burning. I should not have taken issue with your comments, I think we are largely in agreement.
Yes, I agree, the purpose of the mathematics is to empower the predictions, not to identify the actual process. In fact, I would say the express purpose of a mathematical model is to replace the actual process with something that fits inside our heads. For some reason, this replacement often gets misconstrued as a complete description, missing the point that the whole purpose was not to provide a complete description. No, this is the point, no mathematical definition can tell you something about free will other than whether or not the mathematical definition is a useful replacement for free will. It certainly can't tell you if free will is determinist, unless one adopts the weak meaning that anything that is usefully replaced by a determinist model is what we mean by "deterministic" when applied to a real thing.

An interesting tack, but all too easy to say, "according to the mathematician." An artist might say that artists have superpowers. My point here is only that there is no need to find evidence against determinism, the responsibility lies squarely on those who claim that determinism has something to do with free will, either for or against, to demonstrate that property.

Ken G

But you see the error there right away, the word "defined" is inconsistent with the word "system." We don't define systems, we notice them. What we define are mathematical models of systems, but a model is never a system. If the literature is being lazy on this point, then it is really missing something important, perhaps along the lines of what apeiron is saying it is missing.

No, systems are not mathematically deterministic, because systems are not mathematics.
That's the issue.
That's indeed the question. Or the follow-on question, does it matter to whom, and in what way? I would say it all depends on the goals. I think those who make models sometimes seem to forget that they are making models for a reason, they have a goal, and that goal is never to describe completely that which they model, for a complete description is not a model at all, it is only the system itself.

I'm suggesting that determinism is itself a construct, a mathematical idea, not necessarily applicable to real systems except that it makes a useful template to hold up to them-- just as all mathematical models of reality are useful templates. That's easy to state, but the issue in regard to free will is that we don't yet know what elements of free will we are even trying to model, so we cannot say whether or not determinism is a useful template to hold up to free will. We already have examples, in weather and in quantum mechanics, where determinism is not always a useful template, though it does have some applicability and some tendency to break down.

Pythagorean

Q_Goest,

"Not quantum" doesn't mean classical. Nonlinear dynamics and complex systems are modern physics; in my undergrad curriculum they are taught in the two-semester modern physics course, after quantum and relativity.

They do make use of classical physics (moreso than QM does, for instance) but they are not constrained by classical physics, especially because they allow for dissipative (and stochastic) processes.

Dissipiative processes in thermodynamics are irreversible. Moving through a conservative force, like gravity, your can completely recover your ground... in the real world we have friction: a dissipative process from which heat and entropy flow.

This all becomes very important in turbulence models, where heat dissipation and entropy are rampant among correlated deterministic behavior (and change the deterministic behavior that is chaotic, so it's hard to predict how small, random changes from heat dissipation can manifest large consequences)

On stochastic non-holonomic systems
N. K. Moshchuk and I. N. Sinitsyn
Journal of Applied Mathematics and Mechanics
Volume 54, Issue 2, 1990, Pages 174-182

CUMULANTS OF STOCHASTIC RESPONSE FOR A CLASS OF
SPECIAL NONHOLONOMIC SYSTEMS
Shang Mei and Zhang Yi
Chinese Physics
Vol 10 No 1, January 2001

nismaratwork

Nonlinar dynamics includes nonlinear optics, right? (just clarifying for me here, not a leading question.)