A Understanding Barandes' microscopic theory of causality

[pines-demon]

This post is a spin-off of the original post that discussed Barandes theory, A new realistic stochastic interpretation of Quantum Mechanics, for any details about the interpretation in general PLEASE look up for an answer there.

Now I want this post to focus on this pre-print:

• J. A. Barandes, "New Prospects for a Causally Local Formulation of Quantum Theory", arXiv 2402.16935 (2024)

My main concerns are that Barandes thinks this deflates the anti-classical Bell's theorem. In Barandes, words:

By invoking this microphysical notion of causation, one can formulate a more straightforward criterion for causal locality than Bell’s principle of local causality. As this paper has shown, quantum theory, regarded as a theory of unistochastic processes, satisfies this improved criterion, and is therefore arguably a causally local theory. Remarkably, one therefore arrives at what appears to be a causally local hidden-variables formulation of quantum theory, despite many decades of skepticism that such a theory could exist

See Conclusion section.

I do not know how Barandes can claim this and walk away with it. Bell's theorem has been discussed many times and finding a way to bypass it could be either a major breakthrough or a big NO to Barandes theory. I have reviewed popular videos by Barandes in the previous thread and I do not think he has said a word about it aside from pointing to this pre-print.

Other things to say:

• Barandes overview of the history of Bell theorem is on-point. He clearly seems to understand the evolution of the theorem. He has made some less-nuanced claims about Reichenbach principle and Bell but he has commented on that mistake. See this post.
• Barandes seems to distinguish and focus on causal locality (the idea that faster than light influences are not possible), instead of Bell's local causality. He redefines the terms but it makes me wonder if he implicitly is just proving the no-signaling theorem and calling it a day.
• My main concerns are sections V to VII. In this section he tries to see causal locality in a Bayesian network analogy. I would like to understand some version of it.
• His new microscopic principle of causality is defined as:
  

  A theory with microphysical directed conditional probabilities is causally local if any pair of localized systems $Q$ and $R$ that remain at spacelike separation for the duration of a given physical process do not exert causal influences on each other during that process, in the sense that the directed conditional probabilities for $Q$ are independent of $R$, and vice versa.

How can we understand entanglement under Barandes interpretation? How is this different from Bell's? What are the implications, can he say that now everything is deterministic and "locally causal" as in Bell's terminology? Or is he violating one of Bell's assumptions?

 

[javisot]

The following is my interpretation of what Barandes means.

We can try to understand QM in terms of locality and Einsteinian causality (which doesn't end well), or,
we can say that those concepts aren't fundamental. The correct and fundamental notions of those concepts arise from QM, not from GR. Barandes doesn't even bother to translate "entanglement" into this new language, since entanglement is something that arises when we classically observe QM.

I have the feeling that describing the universe from Barandes's interpretation would produce a very exotic universe from the point of view of GR.

 

[pines-demon]

The following is my interpretation of what Barandes means.

We can try to understand QM in terms of locality and Einsteinian causality (which doesn't end well), or,
we can say that those concepts aren't fundamental. The correct and fundamental notions of those concepts arise from QM, not from GR. Barandes doesn't even bother to translate "entanglement" into this new language, since entanglement is something that arises when we classically observe QM.

Just dropping some terminology does not mean that it bypasses the weirdness of quantum mechanics or anything that is already established. Also it does not mean that now QM more classical because of that.

I have the feeling that describing the universe from Barandes's interpretation would produce a very exotic universe from the point of view of GR.

Then let's understand it better and not just call it a day.

 

[javisot]

Then let's understand it better and now just call it a day.

So, the only one who can answer your question is Barandes. Have you tried calling him?

 

[pines-demon]

So, the only one who can answer your question is Barandes. Have you tried calling him?

I am waiting a bit, I am trying to get a more precise picture instead of sending him a message saying "what's all this??".

 

[DrChinese]

• Barandes’ new microscopic principle of causality is defined as:
  

  A theory with microphysical directed conditional probabilities is causally local if any pair of localized systems and that remain at spacelike separation for the duration of a given physical process do not exert causal influences on each other during that process, in the sense that the directed conditional probabilities for are independent of , and vice versa.

Pines Demon:
Yes, this is precisely the spot at which it becomes clear as you say: let’s redefine Signal Locality as Locality and call it a day.

Of course that is meaningless to the real Bell. And yields nothing of benefit.

When there are random outcomes at a single location, there can be no visible change in conditional probability without knowing more information. Otherwise you could signal FTL.

Nonlocality in QM expresses itself differently. For spin correlation, it is a sin/cos and possibly a square function. Cos^2(theta) for example for photons. The key thing is that the only variable is theta. Theta being *future* settings of distant measurement devices. Nothing else.

If you didn’t know any better: You’d think the future affects the past. And you would never think c was a factor. Notice also that there can be no difference between the predictions of QM vs. QFT.

Funny that… Of course we all know better, right?

 

[Fra]

• My main concerns are sections V to VII. In this section he tries to see causal locality in a Bayesian network analogy. I would like to understand some version of it.
• His new microscopic principle of causality is defined as:

How can we understand entanglement under Barandes interpretation? How is this different from Bell's? What are the implications, can he say that now everything is deterministic and "locally causal" as in Bell's terminology? Or is he violating one of Bell's assumptions?

I think think is a good focus, but it has proven diffculty to explain. It does IMO involve a paradigm shift, which may require pushing things beyond Barandes paper. I'll reread his papers and see if i can explain it better than in previous threads. But IMHO, Baranders TYPE of hidden variables are not of the type in Bells' theorem. So I would say he violates the implicit assumption of "divisibility" into an objective beable. It can for at at least be best understood as a network or a population of bayesian systems(agents=subsystems).

I know you didn't associate well to this as I wrote about it in previous threads. I'll try to see if there is a way to express it more clearly. But understanding the principles of the paradigm is one think, and seeing it EXPLICITLY as applied to a model of particcle physics requires nothing less than reformulation one SIDE of baranders correspondence, into something else - which does not exists yet[and implies a new theory]. But I think one can understand the principles without the missing details, but only with an open mind! If one goes into this, trying to shoot down his ideas, it will be very difficulty to appreciate it. Do we want to understand, or do we want to shoot down?

I'll be back after thinking on how to convey this better.

/Fredrik

 

[Fra]

What are the implications, can he say that now everything is deterministic

No. Baranders formulation retains the irreducible randomness as QM. The purpose of the interpretation is not to restore determinism.

"it will be important to be keep in mind the distinction between deterministic hidden-variables theories and stochastic hidden-variables theories"
-- p22, https://arxiv.org/abs/2302.10778

and "locally causal" as in Bell's terminology?

We still have the kind of non-causality as Bells defines it. But I think Barandes explains why this is not problematic - by introducing a "better" definition of "causal locality".

I don't see it as wordplay or just redefining terms, I happen to find Barandes notion to be the one more useful in my own interpretation.

"A theory with microphysical directed conditional probabilities is causally local if any pair of localized systems Q and R that remain at spacelike separation for the duration of a given physical process do not exert causal influences on each other during that process, in the sense that the directed conditional probabilities for Q are independent"
-- p11, arXiv 2402.16935

For me to say that the conditional probabilities of two subsystems are independent, is simply the same as to say that the stochastic behaviour of "system Q (~ Agent Q)" depends on it's own configuration. (subjective beables), and NOT on the beables (configuration) of R (~Agent R). This is exactly what I would expect from an "intrinsic perspective".

I would say that each subsystems stochastic evolution is represented by a bayesian network; at least differentially (to ignore large time evolution).

But quantum weirdness appears when we observer tow such "stochastic subsystems" interacting. In particular when they have an inteaction history, and can have correlated hidden variable. But I don't see this this hidden variable is not an objective beable, so it is not the lamba in bells ansatz. It would not make sense to partition the transition probability of the composite system, by this subjective beable.

As I see the the correlation between the two systems (in entanglement) in Baranders picture, does NOT predetermined the outcomes when they are inteacting with additional systems (such as detectors), but their behavioural responses to the environment will be pre-correlated to explain the correlation of results, for any detector settings.

This is how i understand it conceptually, in the new paradigm. But note, that Barandes does not provive a DETAILED model, or first principle proof of this, in the "unistochastic" picture constrained by the time evolution of the transition matrices - this is an open problem IMO. This is where I want to see more. But I can appreciate the picture, even without this beeing on the table. I think others may not.

Or is he violating one of Bell's assumptions?

The ansatz in bells theorem, is not valid for the "subjective beables", as they can't be used to paritition or divide the transitions. That ansatz is in as far as I understand this, only valid for "objective beables" that has an element of objective reality (though hidden). I don't see the how the collection of all configuration spaces of every subsytems are objective beables.

To argue in detail for this, I find hard, without the revised theory. But I think Barandes himself does not have all answers, and en enourages people to explore with this "corresponende" means for certain systems.

/Fredrik

 

[PeterDonis]

Barandes thinks this deflates the anti-classical Bell's theorem.

It doesn't. What Barandes calls "a causally local hidden-variables formulation of quantum theory" violates the Bell inequalities (in the paper he uses the CHSH inequalities instead, but it amounts to the same thing) and therefore violates at least one of the premises of Bell's theorem. Bell's theorem is a mathematical theorem: anything that satisfies its premises must satisfy its conclusion.

In other words, a model that is "causally local" in Barandes's sense is "nonlocal" in the sense of violating the Bell inequalities under appropriate conditions. Which has nothing whatever to do with physics; it's just playing with words. Nothing in this wordplay makes the experimentally confirmed behavior of entangled systems any less mysterious.

 

[PeterDonis]

he tries to see causal locality in a Bayesian network analogy

All this seems to me like a dodge if its intent is to claim that this is some kind of fundamental breakthrough. He's not offering an actual model of "what's really happening" at a fundamental level. All of the probabilities are epistemic--i.e., he is using (though he doesn't emphasize this) the standard ignorance interpretation of probabilities, where they arise from our lack of knowledge of the underlying dynamics. "Stochastic dynamics" is just another way of saying the same thing--we don't actually know what's going on at a fundamental level, but we can describe it, on some level, using these stochastic equations. "Bayesian network" is just another way of describing the same thing, i.e., of saying we don't actually know the underlying dynamics. And stopping there, as Barandes seems to want to do, looks to me like just another form of the Copenhagen interpretation: we can never know "what's really going on", the most we can ever know is a description that is equivalent to the standard QM description and tells us probabilities of measurement results, but no more.

 

[Fra]

It doesn't. What Barandes calls "a causally local hidden-variables formulation of quantum theory" violates the Bell inequalities (in the paper he uses the CHSH inequalities instead, but it amounts to the same thing) and therefore violates at least one of the premises of Bell's theorem. Bell's theorem is a mathematical theorem: anything that satisfies its premises must satisfy its conclusion.

In other words, a model that is "causally local" in Barandes's sense is "nonlocal" in the sense of violating the Bell inequalities.

I agree.

Nothing in this wordplay makes the experimentally confirmed behavior of entangled systems any less mysterious.

But , the mystery seems to originate from that we seem to lack a detailed causal mechanism for how the physical interactions actually work; in particular in the entanglment experiments. And as Baranders also noted in some talks, the normal dynamical law, working on an objective state space giving rise to the block universe picture, is problematic for even making a sensible notion of causation, becase the future is determined already.

Isn't Baranders really suggesting that dynamical law, at some point would be relace by stochastic interactions between parts? This is a different paradigm to me. But importantly, not a stochstic process happening in objective beable space, its independent process of spacelike separated parts.

"Interestingly, this connection between the directedness of a Bayesian network’s basic conditional probabilities and the asymmetry of cause-and-effect also sheds light on why causal language is so fraught in the context of theories that are based on microphysical laws that are deterministic and reversible. In a deterministically reversible theory, if a value a of a variable A implies a corresponding value b of another variable B, then p(b|a) = 1, and, in addition, any contingent standalone probability p(a) assigned to a will necessarily equal the contingent standalone probability p(b) assigned to b. It follows immediately
from Bayes’ theorem that p(a|b) = p(b|a) = 1, so these conditional probabilities are not directed, and
the asymmetry of cause-and-effect relationships is lost."
-- p11, arXiv 2402.16935

/Fredrik

 

[pines-demon]

In other words, a model that is "causally local" in Barandes's sense is "nonlocal" in the sense of violating the Bell inequalities under appropriate conditions. Which has nothing whatever to do with physics; it's just playing with words. Nothing in this wordplay makes the experimentally confirmed behavior of entangled systems any less mysterious.

Barandes has an interpretation, its interpretation violates Bell inequalities, so his interpretation violates one of the assumptions of Bell's theorem, the question is now which is it?

If Barandes is just showing that quantum mechanics (AND his interpretation) does not violate some other property (causal locality) then what is its interest for other interpretations?

 

[pines-demon]

No. Baranders formulation retains the irreducible randomness as QM. The purpose of the interpretation is not to restore determinism.

"it will be important to be keep in mind the distinction between deterministic hidden-variables theories and stochastic hidden-variables theories"
-- p22, https://arxiv.org/abs/2302.10778


We still have the kind of non-causality as Bells defines it. But I think Barandes explains why this is not problematic - by introducing a "better" definition of "causal locality".

I don't see it as wordplay or just redefining terms, I happen to find Barandes notion to be the one more useful in my own interpretation.

"A theory with microphysical directed conditional probabilities is causally local if any pair of localized systems Q and R that remain at spacelike separation for the duration of a given physical process do not exert causal influences on each other during that process, in the sense that the directed conditional probabilities for Q are independent"
-- p11, arXiv 2402.16935

For me to say that the conditional probabilities of two subsystems are independent, is simply the same as to say that the stochastic behaviour of "system Q (~ Agent Q)" depends on it's own configuration. (subjective beables), and NOT on the beables (configuration) of R (~Agent R). This is exactly what I would expect from an "intrinsic perspective".

Ok, so no determinism. Would it be safe to say he has nonlocal stochastic hidden variables? This is where I block, maybe Barandes wants to discuss everything in other categories but that does not exclude his interpretation from being discussed in existing categories.

 

[pines-demon]

All this seems to me like a dodge if its intent is to claim that this is some kind of fundamental breakthrough. He's not offering an actual model of "what's really happening" at a fundamental level. All of the probabilities are epistemic--i.e., he is using (though he doesn't emphasize this) the standard ignorance interpretation of probabilities, where they arise from our lack of knowledge of the underlying dynamics. "Stochastic dynamics" is just another way of saying the same thing--we don't actually know what's going on at a fundamental level, but we can describe it, on some level, using these stochastic equations. "Bayesian network" is just another way of describing the same thing, i.e., of saying we don't actually know the underlying dynamics. And stopping there, as Barandes seems to want to do, looks to me like just another form of the Copenhagen interpretation: we can never know "what's really going on", the most we can ever know is a description that is equivalent to the standard QM description and tells us probabilities of measurement results, but no more.

Epistemic or not he has claimed to have a "real" interpretation in the sense that there is some kind of classical mechanics going on in the background.

 

[javisot]

Ok, so no determinism. Would it be safe to say he has nonlocal stochastic hidden variables? This is where I block, maybe Barandes wants to discuss everything in other categories but that does not exclude his interpretation from being discussed in existing categories.

From "the stochastic-quantum correspondence", section H "Entanglement":

Barandes- "This analysis precisely captures the quantum-theoretic notion of entanglement. Systems that interact with each other start to exhibit what appears to be a nonlocal kind of stochastic dynamics, even if the systems are moved far apart in physical space, and decoherence by the environment effectively causes a breakdown in that apparent dynamical nonlocality. This stochastic picture of entanglement and nonlocality provides a new way to understand why they occur in
the first place. The indivisible nature of generic stochastic dynamics could be viewed as a form of nonlocality in time, which then leads to an apparent nonlocality across space. A division event leads to an instantaneous restoration of locality in time, which then leads to a momentary restoration of manifest locality across space."

 

[javisot]

"Systems that interact with each other start to exhibit what appears to be a nonlocal kind of stochastic dynamics", I suppose by this he "translate" the strangeness of QM.

 

[pines-demon]

From "the stochastic-quantum correspondence", section H "Entanglement":

Barandes- "This analysis precisely captures the quantum-theoretic notion of entanglement. Systems that interact with each other start to exhibit what appears to be a nonlocal kind of stochastic dynamics, even if the systems are moved far apart in physical space, and decoherence by the environment effectively causes a breakdown in that apparent dynamical nonlocality. This stochastic picture of entanglement and nonlocality provides a new way to understand why they occur in
the first place. The indivisible nature of generic stochastic dynamics could be viewed as a form of nonlocality in time, which then leads to an apparent nonlocality across space. A division event leads to an instantaneous restoration of locality in time, which then leads to a momentary restoration of manifest locality across space."

Thanks for the quote. Assuming this means that his interpretation is nonlocal, then it is no much different to Bohmian mechanics, in this sense it could be an improvement as that there is a more powerful mathematical equivalence and hopefully no preferred basis (right?). But then, if I am not being too reductive, I can think of Barandes interpretation as replacing the pilot wave by a nonlocal permeating stochastic force field that is all over space and acts instantaneously. This is to me not deflationary at all.

 

[javisot]

Thanks for the quote. Assuming this means that his interpretation is nonlocal, then it is no much different to Bohmian mechanics, in this sense it could be an improvement as that there is a more powerful mathematical equivalence and hopefully no preferred basis (right?). But then, if I am not being too reductive, I can think of Barandes interpretation as replacing the pilot wave by a nonlocal permeating stochastic force field that is all over space and acts instantaneously. This is to me not deflationary at all.

I completely agree with what Peter Donis says, even if what Barandes proposes is true and his interpretation can be considered "canonical", it does not discover anything new nor does it help to better explain anything old.

 

[PeterDonis]

the mystery seems to originate from that we seem to lack a detailed causal mechanism for how the physical interactions actually work

I think that's one way of stating it, yes.

Isn't Baranders really suggesting that dynamical law, at some point would be relace by stochastic interactions between parts?

That's just another way of saying what you said in the first quote above--"stochastic interactions between parts" isn't a "detailed causal mechanism", it's just a description of our ignorance about such a thing.

 

[PeterDonis]

I can think of Barandes interpretation as replacing the pilot wave by a nonlocal permeating stochastic force field that is all over space and acts instantaneously

But a "stochastic force" is just another way of saying "we don't understand what's actually going on". The pilot wave is deterministic; probabilities arise because of our ignorance of the actual microphysical state (the particle positions in the usual formulation of Bohmian mechanics). Saying that probabilties are fundamental, that the only dynamics we can have is "stochastic", is saying we can't know what's actually going on--in other words, another form of Copenhagen, as I said in an earlier post.

 

[javisot]

Is there any relationship between stochastic dynamics and the uncertainty principle?

I mean, Barandes assumes that particles have defined properties (the uncertainty principle doesn't directly affect their properties), but Barandes also claims that his interpretation is "canonical" in the sense of "being a Copenhagen," so the uncertainty principle should be expressed in some way in stochastic dynamics (is that correct?)

From "the stochastic-quantum correspondence", "introduction":

"Taking a more foundational perspective, this paper also uses the stochastic-quantum correspondence to show that physical models based on configuration spaces combined with stochastic dynamics can replicate all the empirical predictions of textbook quantum theory—including interference, decoherence, entanglement, noncommutative observables, and wave-function collapse—"


I'm looking for an analogy to show what Barandes is proposing. Suppose we want to explain QM using golf as an analogy, we'll say that the ball has no defined properties, and when you hit it...etc.

Barandes, on the other hand, tells us, "Imagine that the golf ball has defined properties, but the club with which we hit the ball has no defined properties."

 

[Fra]

Would it be safe to say he has nonlocal stochastic hidden variables?

No, not as I think you mean it. As Barandes even notes, Bell himself in the 1975 paper generalized the argument from deterministic to stochastic HV, so instead of thinking that the outcomes are determined by the hidden variable, a probability distribution is determined. But it's important to note that the "stochastics" is still relative to the objective beables.

Bell's picture is about marginalizing over the HV. This is not how Barandes picture works. The HV in Barandes view is the configurations of the subsystems. These are beables for each subsytem. They are not objective beables.

Bell implicitly assume that the hidden varible is a classical probability space with divisible markovian dynanamics. Barandes "generalized stochastic" is not divisible, this is where the ansatz in bells theorm does not apply to Barandes picture.

This is where I block, maybe Barandes wants to discuss everything in other categories

I think he does for a reason, that in different categories or paradigms, the picture may be more clear. After all his motivation is that the hilber picture is indeed mystic or hard to understand. Not necessarily hard to USE in the pragmatic sense, but hard to conceptaully make sense of.

I think Barandes pictures may be harder to understand, if you insist on discussing it in terms of inappropriate categories.

For me, it makes most sense in terms of "bayesian networks" associated to agents, and I associate any subsystem = an agent, and then these subsystems may interact. And this is when "quantum weirdness" happens - when described from a third external observer, because then the "configurations" of the two subsystems are no longer beables, toi the external observer.

/Fredrik

 

[Fra]

That's just another way of saying what you said in the first quote above--"stochastic interactions between parts" isn't a "detailed causal mechanism", it's just a description of our ignorance about such a thing.

No, that's exacatly what I think is a misinterpretation of Barandes suggestion.

The whole point with the indivisbility is that the the stochastic process in Barandes view is NOT an "ignorance". "ignorance" is what Bells' theorem is about.

I think the stochastics in Baranders view is irreducible, and the explicit manifestation of that, is that it is not divisible in terms of "ignorance" of objective beables.

The best way I could describe this is that, in Barandes view, we do not have ONE stochastic process. There is an "independent" stocahstic process going on at each subsystem. This can not be described in terms of "ignorance" of an external observer.

/Fredrik

 

[PeterDonis]

The whole point with the indivisbility is that the the stochastic process in Barandes view is NOT an "ignorance".

Quibble. The key point is here:

I think the stochastics in Baranders view is irreducible

And that means that there is no "detailed causal mechanism" at all. It does not mean that, well, "stochastics" is good enough for a detailed causal mechanism so we don't need to look any further. It means looking any further is pointless because it is impossible to find what we're looking for. We simpl have to accept that there is no "detailed causal mechanism" there at all for us to find.

 

[PeterDonis]

The best way I could describe this is that, in Barandes view, we do not have ONE stochastic process. There is an "independent" stocahstic process going on at each subsystem. This can not be described in terms of "ignorance" of an external observer.

Sure it can--the "observer" is just another subsystem. There is nothing logically impossible about that subsystem not having access to details of what's going on in other subsystems that would be needed to do better than "stochastics" at predicting what other subsystems will do.

What Barandes is really claiming, as you describe it, is that those other details do not exist. If "ignorance" is not a good word to describe Barandes's interpretation, that is the reason--that it's not that those details are there and we just don't know them, it's that they're not there at all. It has nothing to do with things like using Bayesian networks to describe the probabilities instead of something else. It's a fundamental claim about what simply doesn't exist in "reality", independent of any of our descriptions.

 

[javisot]

If there's no causal mechanism behind entanglement, why are some particles entangled and others not? If there's no causal mechanism, they could all be entangled.

But we know that's not true; not all particles are entangled.

(MoE?, what's MoE for Barandes?)

 

[Fra]

And that means that there is no "detailed causal mechanism" at all. It does not mean that, well, "stochastics" is good enough for a detailed causal mechanism so we don't need to look any further. It means looking any further is pointless because it is impossible to find what we're looking for. We simpl have to accept that there is no "detailed causal mechanism" there at all for us to find.

I think we lost the focus a bit. Let me explain. Barandes indeed does not suggest a "mechanism" that restores the randomess of certain things. We agree there. (Except for that the transition matrix, certainly begs for an "explanation" which is missing from his picture - but its because he has a correspondence only - not yet a full first-principle reconstruction, it is indeed irreducible - from the perspective of the subsystems stochastics.)

But this was not the "mechanism" I think we discussed in the thread. The "mechanism" I refers to is to exaplain, not the outcomes, but ONLY the correlation of outcomes. In Bells paradigm, that requires predicting the outcomes as well, or a distribution of them. I see the Baranders pictures suggest a way for a "mechanism" for correlations, while keeping the irreducible interderminism. This is different from bohmian mechanics.

But to explain it convincingly to doubters - I think one needs to find a first principle construction(until that is in place, it is admittdely some handwaving, this is why we are left discussing conceptual princiiples of paradigm and not explicit models, so i think that is exactly what the "interpretational" things are) - not just rely on the "correspondence". It is suggestive, the the mission is not completed yet, so in this sense I agree that hte correspondence alone is not the final solution, but I think unlike others that it is a good step towards a new perspective, that helps.

/Fredrik

 

[jbergman]

This post is a spin-off of the original post that discussed Barandes theory, A new realistic stochastic interpretation of Quantum Mechanics, for any details about the interpretation in general PLEASE look up for an answer there.

Now I want this post to focus on this pre-print:

• J. A. Barandes, "New Prospects for a Causally Local Formulation of Quantum Theory", arXiv 2402.16935 (2024)

My main concerns are that Barandes thinks this deflates the anti-classical Bell's theorem. In Barandes, words:

See Conclusion section.

I do not know how Barandes can claim this and walk away with it. Bell's theorem has been discussed many times and finding a way to bypass it could be either a major breakthrough or a big NO to Barandes theory. I have reviewed popular videos by Barandes in the previous thread and I do not think he has said a word about it aside from pointing to this pre-print.

Other things to say:

• Barandes overview of the history of Bell theorem is on-point. He clearly seems to understand the evolution of the theorem. He has made some less-nuanced claims about Reichenbach principle and Bell but he has commented on that mistake. See this post.
• Barandes seems to distinguish and focus on causal locality (the idea that faster than light influences are not possible), instead of Bell's local causality. He redefines the terms but it makes me wonder if he implicitly is just proving the no-signaling theorem and calling it a day.
• My main concerns are sections V to VII. In this section he tries to see causal locality in a Bayesian network analogy. I would like to understand some version of it.
• His new microscopic principle of causality is defined as:

How can we understand entanglement under Barandes interpretation? How is this different from Bell's? What are the implications, can he say that now everything is deterministic and "locally causal" as in Bell's terminology? Or is he violating one of Bell's assumptions?

I should dig into this paper but don't have the time. I did want to say that Bayesian networks have very strong markovian properties which his unistochastic processes don't have, so I am not sure what the analogy is here.

 

[PeterDonis]

Barandes indeed does not suggest a "mechanism" that restores the randomess of certain things. We agree there.

Which means that, by your own admission, Barandes does not solve what you yourself said was the mystery.

The "mechanism" I refers to is to exaplain, not the outcomes, but ONLY the correlation of outcomes.

But, as I've already said, "something something stochastic something something" is not a mechanism. It's just a description. It doesn't explain the correlations. It just describes them.

 

[Morbert]

Re/ Barandes and Copenhagen:

The Copenhagen interpretation (to the extent that it is a single interpretation) does not have a microscale ontology. You cannot assert existing-but-unknown microscopic properties. Instead, it asserts macroscopic preparation protocols, tests, and responses.

Barandes's unistochastic interpretation has a microscale ontology. You can assert existing-but-unknown microscopic properties on which measurement outcomes supervene.

Re/ Barandes and Bell:

Barandes does not dispute Bell's theorem. He instead rejects it as a good basis for a theory of local microphysical causation, and offers an alternative (and, Barandes has remarked, more complete) theory.

-

Ultimately the novelty to Barandes's approach seems to be to move inseparability from states to dynamics. This lets him recover both a microscale ontology and a consistent, unambiguous model of microphysical causality and a principle of local causality.

 

[Fra]

This is hard to discuss. I admit that part of my confidence is from other directions, that happens to merge with Barandes picture at the intersection.

I'll try again in this way, not to fill in the missing parts, but to try to dress what I think are the keys

• My main concerns are sections V to VII. In this section he tries to see causal locality in a Bayesian network analogy. I would like to understand some version of it.
• His new microscopic principle of causality is defined as:

How can we understand entanglement under Barandes interpretation?

I should dig into this paper but don't have the time. I did want to say that Bayesian networks have very strong markovian properties which his unistochastic processes don't have, so I am not sure what the analogy is here.

They key lies on composition of subsystems, namely in what a constraint implicit in the composition imply.

If you consider two independent bayesian networks, but then combine them either "as is" as two independnent parts, or you combine then under some constraint, then this constraint implies a dependence between the two previously independent parts - in different ways, depending on what the constrain is of course. But the general observations has nothing todo with those details.

The magic lines in understanding the constraints, what they mean, conceptually. Barandes only has a correspondence, which via the time dependent stochastic matrices implements this constraint. But he offers no first principle explanation. This is missing, and I agree.

So I think the generalization is to consider not one bayesian network, but a system of bayesian networks that are dependent via constraints; and during certain circumstances this generalisation is a unistochastic process.

The reason for my confidence in the ideas is that that my own interpretation contains alterantive mechanisms here, that fill in the missing parts at least for me.

The most natural interpretation of the constraint is to see the two indenepdent bayesian networks as encoding predictions of the future; but from non-commutative perspectives, or different basis. And sometimes this may offer better predictive power per storage. So a global memory constraint would be a natural interpretation of a constraint. Or energy constraint if that makes more sense. I mean, now matter HOW you divide a ssytem into parts, there is some sort of conservation usually, thats implicit in the division. What that conservation is can vary.

In this sense this can all be seem as generalized inference system, where you instead of having a "simple" bayesian network, entertain parallell networks representing different encodings and consider encoding their composition. But this implies via a constraint that can mean different things in different contexts, that the parts become dependent via a top-down constraint. What properties does this system have? If its unistochastic, then Barandes showed that such systems always will exhibit quantum behaviour.

So what possible intuition do we have for such top-down constraint? There are apparently many, from a model perspectrive, but which ones would be suitable for nature and why? This is strongly related to I think information theoretic interpreations of QM, which is why i like Barandes correspondence so I feel I ought to defend the idea when some suggest that it adds nothing, which i think is not very fair.

He's view at least offers a handle to a new paradigm, to be modest. But it's up to us, what todo with this handle. Build onto it, or consider it a useless appendix

There are many other interpretations or formalisms i would consider useless before Barandes.

I would say non of this is "weird", but its rejecting realism, in the sense that it is a strong information theoretic approach. But if you take information processing as REAL, when done not by humans, but by physical subsystems, doing "stochastic" basic processing. Then it is abstract, but still "real". There is nothing surrealistic in this this. For me it is a "realistic" explanation, but not in the seense of how Bell uses the word.

/Fredrik

 

[javisot]

Barandes does not dispute Bell's theorem. He instead rejects it as a good basis for a theory of local microphysical causation, and offers an alternative (and, Barandes has remarked, more complete) theory.

Regarding Bell's theorem, Barandes comments the following in the paper shared by @pines-demon,

"There are several incorrect ways to read the stronger 1975 version of Bell’s theorem. One is that the theorem rules out hidden variables altogether. Another false reading is that one can avoid violating Bell’s principle of local causality merely by avoiding the introduction of hidden variables—but this reading confuses the weaker 1964 version of Bell’s theorem with the stronger 1975 version, which applies even to theories that do not include hidden variables at all, like textbook quantum theory itself. The correct reading of Bell’s theorem is to stay close to what Bell himself wrote and conclude that his principle of local causality is violated by all empirically adequate formulations of quantum theory, including the textbook version of the theory, again putting aside various potential loopholes.

It is far from clear, however, that the principle of local causality that Bell used to prove the stronger version of his theorem was the correct way to formulate the more basic condition of causal locality in the first place. Bell himself warned against taking his principle of local causality too seriously. Indeed, in a 1990 lecture [15], he cautioned that his principle “should be viewed with the utmost suspicion.”

Bell had good reasons for being skeptical of his own theorem’s premises, due to his history with an older theorem proved by John von Neumann decades before. That earlier theorem had been widely viewed as completely ruling out the possibility of hidden variables [34–36]. Already in 1935, Grete Hermann had determined that von Neumann’s theorem depended on an assumption about expectation values that was too narrow [37, 38]. Bell essentially discovered the same flaw in von Neumann’s proof decades later [39]. (For an excellent historical discussion of von Neumann’s theorem, its shortcomings, and its critics, see [40].)"

But I still don't understand. Barandes comments in another paper: "Systems that interact with each other start to exhibit what appears to be a nonlocal kind of stochastic dynamics."

This means that stochastic dynamics also contains nonlocality (or am I misunderstanding this?)

 

[Morbert]

This means that stochastic dynamics also contains nonlocality (or am I misunderstanding this?)

Barandes says two systems are entangled if the stochastic process describing their joint evolution cannot be factorized. (See equation 60 https://arxiv.org/pdf/2402.16935 ). However, he shows that this does not enable a violation of his proposed principle of causal locality (55, same paper) by considering observer systems Alice and Bob intervening on the entangled systems (section VIII, same paper).

 

[pines-demon]

Barandes says two systems are entangled if the stochastic process describing their joint evolution cannot be factorized. (See equation 60 https://arxiv.org/pdf/2402.16935 ). However, he shows that this does not enable a violation of his proposed principle of causal locality (55, same paper) by considering observer systems Alice and Bob intervening on the entangled systems (section VIII, same paper).

But this is kind of pointless, no entangled state in any interpretation will violate his causal locality principle right?

 

[DrChinese]

MoE Monogamy of Entanglement

But we know that's not true; not all particles are entangled.

(MoE?, what's MoE for Barandes?)

 

[javisot]

MoE Monogamy of Entanglement

I mean, we asked how Barandes understands entanglement, swapping, or delayed choice, but how does Barandes understand the monogamy of entanglement?

 

[Fra]

If there's no causal mechanism behind entanglement

I don't agree with this. It is the exact path chosen in the stochastics that has no causal mechanism, from the intrinsic perspective of transition matrix.

, why are some particles entangled and others not? If there's no causal mechanism, they could all be entangled.

But we know that's not true; not all particles are entangled.

(MoE?, what's MoE for Barandes?)

WHY we have entanglement in the first place is an interesting but deeper question. Barandes however says that the composite system’s transition matrix $$\Gamma_{AB}(t)$$ which encodes cumulative statistical information and therefore correlations to the extent that exists, will cauase fail to factorize between the two subsystems A and B.

But there is not first principle construction of why this ΓAB(t); to answer that would be the ask for an explanation of the internal structure of a system. It is similar to asking, why doe we have this particular hamiltonian operator in QM?

But I think asking these questions in the Hilbert formalism pretty much ends up with the picture that the hamiltonian is simpled inferred from an external observer via tomographic processes and statistics.

IF you seek a more evolutionary or emergent picture, rather than plain parameter fitting, I think Barandes perspective of interacting systsystems is a step in a better direction.

Personally I seek answers of these things in terms of evolutionary stability of subsystems, entanglement of parts can be conceptaully related to compressed encoding, and might be necessary for stability of physical systems. But we need a paradigm that makes these questions not look like a pure "information theoretic" picture, where "information" refers to human information in the lab. Many object to such perspectives, and i to do. I think the information encoding must be take more serious, and it is physical. Not the fiction in the brains of mathematicians. This is how I see Baranders as seeing the observer as any subsystem. This is a progress.

/Fredrik

 

[Morbert]

But this is kind of pointless, no entangled state in any interpretation will violate his causal locality principle right?

My understanding is other* interpretations would not readily have access to random variables on configuration space, and hence don't have access to his causality principle (which explicitly refers to "microphysical directed conditional probabilities"). Barandes's spectrum of magnitudes are akin to standard QM's decompositions of the identity operator on a Hilbert space, which are used to establish no-signalling between spacelike-separated systems. Barandes is using his formalism to posit a (in his view) more appropriate locality principle to conclude more than mere no-signalling.

With that said, I wish, instead of (71), he had derived $$p((q_t,a_t), t | (q_0,r_0,a_0,b_0), 0) = p((q_t,a_t), t | (q_0,r_0,a_0), 0)$$I.e. I wish he would have showed that the observer subsystem B has no causal influence on either Q or A. By focusing only on A, he might still have to worry about a no-signalling loophole. Nothing about section VIII suggests this alternative expression would not hold, but I am not that familiar with such derivations so there might be some subtle point I'm missing.

* [edit] - Bohmian mechanics presumably would. It would be interesting to know what it says about these Bayesian networks on configurations.

 

[Morbert]

I mean, we asked how Barandes understands entanglement, swapping, or delayed choice, but how does Barandes understand the monogamy of entanglement?

I have not found Barandes explicitly mentioning monogamy of entanglement in literature. But I suspect it would mean, given three subsystems A, B, C: If the unistochastic process $\Gamma_{AB}(t)$ is maximally nonfactorizable, then the unistochastic process $\Gamma_{ABC}(t)$ must factorize into $\Gamma_{AB}(t)\otimes\Gamma_C(t)$, at least when no interactions are present. Nonfactorizability would hopefully be quantifiable, in the same way bitartite entanglement is quantifiable in the Coffman–Kundu–Wootters inequality.

[edit] - Tried to get this kind of expression for the GHZ case and failed so this might not be true.

 

[Morbert]

As an aside, people might find this timestamped video relevant.

 

[pines-demon]

This post was recently created:

• Barandes's Unistochastic Refomulation Applied to Entanglement Swapping

It made me wonder, if there a way to use Barandes formalism to check which assumption of Bell's local causality is violated by Barandes interpretation? Probably it is the locality condition, but I wonder if it is statistical independence or something else.

 

[gentzen]

I finally finished watching

i.e. the last video linked in that other overly long thread which is now closed. I guess I just wanted to remove it from my "watch later" list. The reason why it got there is that I started to watch it when the other thread was still open, and I definitely got the impression that Barandes is back to earth now, and focuses more on solid stuff and his huge philosophical and physical background.

I changed my mind on this latest unistochastic paper now, but not with respect to the unistochastic part. If I read this paper as an attempt to get a better grip on causality, and see the unistochastic part just as what motivated Barandes, then it is a valid investigation which could lead to interesting learnings.

Barandes tries to present an alternative to the interventionist conception of causation. I recently saw somebody claiming "correlation + intervention = causation". It guess it was Peter Morgan, but maybe not. And somebody else explained why intervention is not required for causation (here I am quite sure that this was in Sean Carroll's mindscape episode with Barry Loewer).
I was initially underwhelmed by Barandes' enthusiasm for Bayesian networks, because I know that Judea Pearl promoted them exactly in the context of his work on causality. However, what I now learned from that latest video is than John Bell had no good grip on causality, independent of whether the work of Judea Pearl nailed causality or not. So it makes perfect sense to synthesize the current knowledge on causality and then check how well Bell's theorem holds up in that context. And if good old "no signaling" should be good enough to deny the nonlocality conclusion of Bell's theorem then, that would be an interesting learning.

 

[Morbert]

It made me wonder, if there a way to use Barandes formalism to check which assumption of Bell's local causality is violated by Barandes interpretation? Probably it is the locality condition, but I wonder if it is statistical independence or something else.

Barandes argues that Bell's principle of local causality relies on a definition of common cause that is too narrow. Specifically, Bell's principle of local causality assumes common causes must take the form of "Reichenbachian variables". E.g. If two variables A and B are correlated, but not exerting influence on one another, then there must be variables λ that exert a common causal influence on A and B.

Barandes argues there can be non-Reichenbachian common causes that establish the correlations seen in entanglement, like local interactions at a previous time, that Bell's principle miss as they do not take the form of λ, and hence you can have causally local theories that violate Bell's principle of local causality.

 

[Morbert]

I was initially underwhelmed by Barandes' enthusiam for Bayesian networks, because I know that Judea Pearl promoted them exactly in the context of his work on causality. However, what I now learned from that latest video is than John Bell had no good grip on causality, independent of whether the work of Judea Pearl nailed causality or not. So it makes perfect sense to synthesize the current knowledge on causality and then check how well Bell's theorem holds up in that context. And if good old "no signaling" should be good enough to deny the nonlocality conclusion of Bell's theorem then, that would be an interesting learning.

I'm having a hard time identifying a standard procedure for constructing appropriate Bayesian networks from his papers. E.g. Two particles that become entangled via a local interaction will therefore causally influence each other according to Barandes's principle of causal locality, but how can this be expressed via a Bayesian network, as you cannot have two nodes pointing at each other. I assume you would need to split the variables into moments 0, and t (the start and end of the unistochastic process) but I see no mention of this by Barandes so for now it is just a guess on my part.

 

[RUTA]

Quibble. The key point is here:


And that means that there is no "detailed causal mechanism" at all. It does not mean that, well, "stochastics" is good enough for a detailed causal mechanism so we don't need to look any further. It means looking any further is pointless because it is impossible to find what we're looking for. We simpl have to accept that there is no "detailed causal mechanism" there at all for us to find.

Exactly, he doesn't actually deflate Bell's theorem, he ignores it. Quantum information theorists have a way to deflate it for example. They point out that superposition and entanglement are kinematic facts about QM, not dynamical effects due to some nonlocal or superdeterministic or retro causal mechanism. They then supply two principles giving rise to the kinematic (Hilbert space) structure of QM (relativity principle justifies Information Invariance & Continuity) exactly like the relativity principle justifying the light postulate gives you the Lorentz transformations (kinematics) of special relativity. The vast majority of physicists no longer bother looking for a causal mechanism for length contraction and time dilation, they accept those are kinematic facts resulting from the observer-independence of c as justified by the relativity principle, not dynamical effects due to some causal mechanism (like the luminiferous aether). The analogy is completed by showing that Information Invariance & Continuity entails the observer-independence of Planck's constant h. Barandes could simply adopt this view and be done with it rather than engage in meaningless word plays.

 

[DrChinese]

Barandes argues there can be [1] non-Reichenbachian common causes that establish the correlations seen in entanglement, like local interactions at a [2] previous time, that Bell's principle miss as they do not take the form of λ, and hence you can have causally local theories that violate Bell's principle of local causality.

Thanks for this...

[1] This is precisely the kind of hand-waving (not on your part Morbert) that is objectionable. Barandes simply asserts that such type exists so he can "escape" Bell. How does any of this make sense? Perhaps a concrete example of what this might look like?

[2] We already know that delayed entanglement swaps don't feature a local interaction at any previous time. Ma et al 2012, etc...

 

[Morbert]

[1] This is precisely the kind of hand-waving (not on your part Morbert) that is objectionable. Barandes simply asserts that such type exists so he can "escape" Bell. How does any of this make sense? Perhaps a concrete example of what this might look like?

In section VII ( https://arxiv.org/pdf/2402.16935 ) he considers two systems $Q, R$ that interact with each other at time $t'$, such that the time evolution $U_{QR}$ no longer factorizes. Even after the interaction, when the time evolution factorizes again, the transition matrix $\Gamma_{QR}$ remains unfactorizeable. This "unfactorizeability" is entailed by the interaction. Barandes is not the first to challenge Bell's principles along these lines. Barandes quotes Unruh in section III of the same paper.

Re/ the charge of Handwaving. I don't think Barandes is handwaving, but I do think his reformualtion is quite raw, and there is plenty of homework to do to bring the project to maturity.

I'll leave entanglement swapping for the other thread.

 

[Morbert]

Perhaps a concrete example of what this might look like?

You could consider two particles entangled by the appropriate quantum gates. The unitary operator $U\ket{00} = \frac{1}{\sqrt{2}}(\ket{00} + \ket{11})$ would not factorize to operators on the individual systems.

 

[DrChinese]

You could consider two particles entangled by the appropriate quantum gates. The unitary operator $U\ket{00} = \frac{1}{\sqrt{2}}(\ket{00} + \ket{11})$ would not factorize to operators on the individual systems.

I was asking about a concrete example of what you called (while explaining Barandes) non-Reichenbachian common causes. Whatever that is, because I don’t think it is anything. It seems like hand-waving in order to deny Bell.

What you described is nonseparability of components of a normal entangled system.

 

[Morbert]

I was asking about a concrete example of what you called (while explaining Barandes) non-Reichenbachian common causes. Whatever that is, because I don’t think it is anything. It seems like hand-waving in order to deny Bell.

What you described is nonseparability of components of a normal entangled system.

The non-Reichenbachian common cause is the nonseparable $U$.

Notice that this local interaction, despite being the‘common cause’ of the correlations between Q and R, is not the sort of ‘variable’ that can be plugged into theunistochastic theory’s microphysical conditional probabilities. Reichenbach’s principle of common causes (16) therefore does not hold.