A Understanding Barandes' microscopic theory of causality

[Fra]

"the assumption of AOE implies that, in each run of the experiment (...) there exists a well-defined value for the outcome observed by each observer, that is, for a, b, c and d. Formally, this implies that there exists a theoretical joint probability distribution P(abcd|xy) from which the empirical probability ℘(ab|xy) can be obtained."

The issue here is from my perspective fully analogous to the issue I have with with bell ansats. And what Barandes calls the divisibility assumption in bells ansats. In LF the corresponding "issue" is the assumption
AOE (i) on page 4 of arxiv:1907.05607.

This is what tries to make the beables of the friends objective, so they can be objectively marginalized by "classical probability" (without interference). But this is the part of hte AOE definition that does not make sense; thus AOE fails.

One might still say that there are well defined outcomes for each observer, but they are not objective beables in the sense of Bell. They are subjective beables, relative to each subsystem. So the problem isn't that there aren't well defined outcomes, the problem is they they are hidden from causal influence since they aren't inferrable by the subsystems its interacting with. this is why the premise of both Bells theorem and the AOE (i) on page 4 seem invalid premises. Again the problem isnt the theorem, but the premises.

I mist admit that conceptually this is they very same thing as the core issue of Bells inqeuality, it's just that the example is different and puts hte observers in the system. But at it's core, it is to me at least, it's abstraction is the SAME problem.

/Fredrik

 

[Fra]

Where does he talk about the even more general stochastic process?

The Stochastic-Quantum Theorem (SQT?) is an embedding theorem saying that a general stochastic system GSS can always mathematically be seen as a subsystem in and unistochastic embedding system. This in itself isn't connected to QM. Eq 65 on p9 in arxiv:2309.03085

Then and only then, when we have a (SQC?) correspondence between unistochastic systems and hilber space picture.

This is a key for me personally as it means that the SQC may not be valid universally but applies only when the embedding guaranteed by SQT makes physical sense. I think one should not mistake mathematical possibilities with physical possibilities. I try to understand this form the ABM, perspective, which would suggest that the SQT embedding means that there must be "room" for the emebedding in the capcity of any agent=subsystem, where the GSS can be modelled as a subsystme of a unistochastic system. This imposes some sort of relative complexity balance between the observer and observed. It in this sense I view QM as "emergent". This is why I think things that are often dismissed by FAPP style arguments, often causes use FAPP motivated "constraints", to make deductions too far away from the validated domain. An example is that the apparent "mathematical patterns" vaild for all subatomic physics, are things we can apply to cosmology or other things.

I have nto tracking but in one of the maney youtube interview that was posted here and there, I recall Barandes also speaking about that a subsystem that is to qualify as a sort of meareuemnt device bust qualify as beeing "sufficiently complex/big" and it mest be well equilibrated with the embedding. All these are somehow constinrats on the information flow, and interaction strenght between subsystems and large subsystems vs the embedding environment. In this sense I also hint from Barandes that the correspondence needs som qualifying "limit" in this sense to make sense.

The embedding requirement in Barandes view, is no less problematic than in other interprefations, this is one problem that I see remains unresolved. But the correspondence to me, aligns alot better with my thinking, by isolating the issue to more managable places.

/Fredrik

 

[Sambuco]

One might still say that there are well defined outcomes for each observer, but they are not objective beables in the sense of Bell. They are subjective beables, relative to each subsystem.

I think this is what it means to reject absoluteness of observed events.

Lucas.

 

[Morbert]

One might still say that there are well defined outcomes for each observer, but they are not objective beables in the sense of Bell. They are subjective beables, relative to each subsystem.

I think this is what it means to reject absoluteness of observed events.

Remember though that in this formalism, the configuration a system or subsystem is in is not indexed to any observer. In which case if the above is a rejection of AOE, then this formalism does not reject AOE.

 

[Fra]

Remember though that in this formalism, the configuration a system or subsystem is in is not indexed to any observer.

My issue is this:

I agree it's not indexed by an explicit observer, but i see it as indexed by an implicit observer.

Unlike qbism and other things, Barandes indeed does not entertain the notion of "observer" or "agent", but the only way Barandes picture makes sense to me is to identify "sufficiently complex/large" systems or subsystems with an angen or "measurement device". So for me, the configuraiton - including the decomposition into subsystems when appropriate, defines the "agent subsystem". Thus when you say the configuration is not indexed by any observer, this to me means it is simply representing the "current state" of the "observer" associated to the sufficiently large subsystem (large here means the same as in normal QM, large enough to FAPP qualify as a macroscopic device).

So the embedding configuration that you say isn't indexed to an observer, to me, DEFINES the maximal superobserver". I don't think Barandes ever put it this way, so perhaps he would disagree but to me it is the only way to make sense of it.

What makes me think this is consistent with Barandes thinking is that on p12 of arxiv:2402.16935 he associates the subsystems that Alice and Bob "has access to". And I read this to mean, that an observer has access to it's own configuration space, and nothing else.

So I think - anything "relative to a configuration" - means when using "observer terms" - it's relative to an "observer". Namely the "observer microstructure" defined by the configuration.

So the "objective events" you refer to, IMO refers to the implicit maximal superobserver associated to the embedding unistochastic system, which is isolated - at least effectively.

I think we may be interpreting Barandes interpretation here, but this is my best understanding of this, but I think it's true that this extends a bit beyond the mathematical correspondence. But if we aren't to reflect from differnt angles, there might not be anything to understand at all. It just is what it is.

/Fredrik

 

[Morbert]

My issue is this:

I agree it's not indexed by an explicit observer, but i see it as indexed by an implicit observer.

Unlike qbism and other things, Barandes indeed does not entertain the notion of "observer" or "agent", but the only way Barandes picture makes sense to me is to identify "sufficiently complex/large" systems or subsystems with an angen or "measurement device". So for me, the configuraiton - including the decomposition into subsystems when appropriate, defines the "agent subsystem". Thus when you say the configuration is not indexed by any observer, this to me means it is simply representing the "current state" of the "observer" associated to the sufficiently large subsystem (large here means the same as in normal QM, large enough to FAPP qualify as a macroscopic device).

So the embedding configuration that you say isn't indexed to an observer, to me, DEFINES the maximal superobserver". I don't think Barandes ever put it this way, so perhaps he would disagree but to me it is the only way to make sense of it.

What makes me think this is consistent with Barandes thinking is that on p12 of arxiv:2402.16935 he associates the subsystems that Alice and Bob "has access to". And I read this to mean, that an observer has access to it's own configuration space, and nothing else.

So I think - anything "relative to a configuration" - means when using "observer terms" - it's relative to an "observer". Namely the "observer microstructure" defined by the configuration.

So the "objective events" you refer to, IMO refers to the implicit maximal superobserver associated to the embedding unistochastic system, which is isolated - at least effectively.

I think we may be interpreting Barandes interpretation here, but this is my best understanding of this, but I think it's true that this extends a bit beyond the mathematical correspondence. But if we aren't to reflect from differnt angles, there might not be anything to understand at all. It just is what it is.

/Fredrik

The only relevance systems that partition into a measured system and a measurement device/environment have is they are the systems we are typically interested in. The formalism can be applied in principle to the universe as a whole, or to a single isolated hydrogen atom, never to be measured.

Nor is the configuration of a system at any moment contingent on whether we partition it into subsystems or how we partition it into subsystems.

Nor is configuration contingent on access to configuration information. AOE is about the ontic, actually existing state of affairs rather than observer epistemics. If the entire universe consisted of the hydrogen atom mentioned above, a unistochastic account would mean the atom has a definite configuration at all times even when nothing exists to measure it.

 

[Fra]

The only relevance systems that partition into a measured system and a measurement device/environment have is they are the systems we are typically interested in.

If that is as far Barandes ever wants to take it, I think we are missing out making the most of the idea he started.

I personally see a different reason, as I have an interential perspective (because making inferences about reality is to me what "we" do). Here I see ontological and epistemological perspectives are not in constrast, I see them as different perspectives that supports each other.

The ontology of the inferrer; constrains the epistemology. And the epistemology is what supports, explains the ontology. So I need both. Ontology without epistemological perspective is just as silly, as a pure epistemologicla perspective without ontological comittements.

The more technical reasons why I insist seeing it this way is that an ontology, other needs to be fine tuned. Ie. to postulate a non trivial ontology, would have very little explanatory value.

The formalism can be applied in principle to the universe as a whole, or to a single isolated hydrogen atom, never to be measured.

Nor is the configuration of a system at any moment contingent on whether we partition it into subsystems or how we partition it into subsystems.

Nor is configuration contingent on access to configuration information. AOE is about the ontic, actually existing state of affairs rather than observer epistemics. If the entire universe consisted of the hydrogen atom mentioned above, a unistochastic account would mean the atom has a definite configuration at all times even when nothing exists to measure it.

Yes I agree with all these. The concept of "observer" does not need constant observations, one can consider the ontological state of an observer in isolation. This is then apparently what Barandes picture suggest. so we have an ontology without measurements. This is fine. But then you have something to finetune. Will this help up solve the open problem in physics? I think of the next step then I would like to think of the ontology having an epistemological handle (that has been temporarily dormant).

So when I mentally think there is an "implicit observer" it doesn't mean that there are active measurements going on, it can be thought of just as the ontic state of an isolated obsever (which then; if further questioned needs to be fine tuned OR put into a bigger context; ie break the isolation).

I don't disagree with you, I just see this as nuances of how to understnad hte configuration. Perhaps Barandes would not like this angle, but for me it's the ongle angle from which I can appreciate his correspondence. I simply cant find a better view. The "fine tuning" view is valid, but seems not as constructive.

/Fredrik

 

[Morbert]

The ontology of the inferrer; constrains the epistemology. And the epistemology is what supports, explains the ontology.

I'm not sure what you mean by this but I don't think I agree, and it would take us too far into philosophy for my comfort.

 

[iste]

Thought these papers might be relevant to realist perspective on Wogner friend:

https://www.mdpi.com/1099-4300/24/7/903

Note how equation (12) and (34) is more or less the same as the kind of (in)divisibility condition you see in Barandes' papers.

https://scholar.google.co.uk/schola...2237082&hl=en&as_sdt=0,5&as_ylo=2021&as_vis=1

Argues that Wigner experiment can be explained by mutually incompatible measurement contexts that can never co-exist. At the end I.believe they are saying this allows an objective resolution which is nonetheless observer-dependent.

I actually haven't read either in sufficiently proper detail yet but look relevant.

 

[Sambuco]

It's not too important. At 2:58:01 he remarks on Wigner's friend scenarios. Even if a Wigner superobserver uses a stochastic map with no division events, Wigner's friend will still be in a definite configuration. The link is youtube<dot>com/watch?v=JsmX3YxiUj0&t=10681s

Thanks for the reference. Barandes is pretty clear about that.

I don't see how the bit in bold is true if dynamics are non-Markovian. [edit] - I.e.:

AOE + Markovian dynamics = joint probability exists
no AOE + Markovian dynamics = joint probability doesn't exist
AOE + non-Markovian dynamics = joint probability doesn't exist

I was trying to understand at what exact point Bong et al. assume a Markovian dynamics. In eq. (3), they write $P(ab|xy) = \sum_{c,d} {P(abcd|xy)} = \sum_{c,d} {P(ab|cdxy) P(cd)}$. In the first equality, they assume AOE, while in the second they assume NSD (non-superdeterminism). However, NSD only implies $P(cd|xy) = P(cd)$, so when applying the second equality, they also to assume that $P(abcd|xy) = P(ab|cdxy) P(cd|xy)$. Could we say that this represents, in some way, the Markovian assumption?

If that is true, Barandes' formulation overcomes the LF no-go theorem not by rejection AOE, NSD, and/or L, but by demonstrating that a fourth implicit assumption was used to derive the theorem.

Lucas.

 

[Fra]

I'm not sure what you mean by this but I don't think I agree, and it would take us too far into philosophy for my comfort.

Jusy to explain the "philosophy" what I had in mind is not just word plany, its just related to a computational complexity association. But I hoped the conceptual things came acroess without going into details, beucase the details can change without affecting the abstraction. I find that details sometimes help, sometimes obscures.

I'd not as easy to dig up youtube references but in one of the 1-2 first Barandes interviews that was posted somewhere on here, there was a guy discussing quantum computing, and its superiority over classical computing. It was the first thing that caught my attention. He was talking to someone else (dont remember) about quantum computation and then about his new formulation.

It seems to me Barandes correspondence offerts insight to a "classical understnading" of the power of quantum computers; as a parallell processing if you consider the interacting subsystems, unlike a sequenstial computation based on initial states. (system dynamics vs ABM aslo reflects this).

If you think of time evolution of each subsystem in barandes view, it's a stochastically driven (guided) process and not drive by system dynamical law. So if we can speed up the stochastic rate, it relates to flow of time. This is the interesting angle of barandes when he replaces dynamical LAW but a purely stochastically driven process. This is pretty profound IMO.

ontology ~ computer hardware ~ configuration - the computer is ontic, real independent of whatever code it runs

epistemology ~ algortims, inferences, measurements; requires a hardware to be realized, so it depends on and is constrained by the hardware

intuitively the configuration microstructure, constraints the "stochastic process" (which is the driver of time evolution). To understand how the computational economy supports the hardare, I think one needs to think in these terms https://en.wikipedia.org/wiki/Evolutionary_computation

/Fredrik

 

[Morbert]

Thanks for the reference. Barandes is pretty clear about that.


I was trying to understand at what exact point Bong et al. assume a Markovian dynamics. In eq. (3), they write $P(ab|xy) = \sum_{c,d} {P(abcd|xy)} = \sum_{c,d} {P(ab|cdxy) P(cd)}$. In the first equality, they assume AOE, while in the second they assume NSD (non-superdeterminism). However, NSD only implies $P(cd|xy) = P(cd)$, so when applying the second equality, they also to assume that $P(abcd|xy) = P(ab|cdxy) P(cd|xy)$. Could we say that this represents, in some way, the Markovian assumption?

If that is true, Barandes' formulation overcomes the LF no-go theorem not by rejection AOE, NSD, and/or L, but by demonstrating that a fourth implicit assumption was used to derive the theorem.

Lucas.

I still need to read the paper properly to give a proper response, but in the AOE section they have "ii. P(a|cd,...)" and "iii. P(b|cd,...)". Since c and d are on the righ hand side of the |, I read these as sparse directed conditional probabilities, conditioned at the time when Charlie and Debbie measured their respective particles. I.e. A Markovian division event, which would not exist in the unistochastic matrix used by Alice and Bob.

I didn't get as far as the NSD or L conditions.

 

[Sambuco]

I still need to read the paper properly to give a proper response, but in the AOE section they have "ii. P(a|cd,...)" and "iii. P(b|cd,...)". Since c and d are on the righ hand side of the |, I read these as sparse directed conditional probabilities, conditioned at the time when Charlie and Debbie measured their respective particles. I.e. A Markovian division event, which would not exist in the unistochastic matrix used by Alice and Bob.

That's true, but cases ii and iii correspond to situations where Alice and Bob enter their respective laboratories and ask Charlie and Debbie for their measurement outcomes. In other words, these are cases where Alice and Bob do not erase Charlie and Debbie's measurement results.

Lucas.

 

[Sambuco]

Thought these papers might be relevant to realist perspective on Wogner friend:

https://www.mdpi.com/1099-4300/24/7/903

Note how equation (12) and (34) is more or less the same as the kind of (in)divisibility condition you see in Barandes' papers.

https://scholar.google.co.uk/schola...2237082&hl=en&as_sdt=0,5&as_ylo=2021&as_vis=1

Argues that Wigner experiment can be explained by mutually incompatible measurement contexts that can never co-exist. At the end I.believe they are saying this allows an objective resolution which is nonetheless observer-dependent.

I actually haven't read either in sufficiently proper detail yet but look relevant.

Yes, this is really very relevant!

In section 4.1, it says:

"Then, in Equation (3) of [9], it implicitly assumes $p(abcd|xy) = p(ab|cdxy) p(cd|xy)$. Together, they imply $p(ab|xy) =\sum_{c,d} p(ab|cdxy) p(cd|xy)$. However, as discussed in Section 2, the law of total probability does not hold true in quantum theory unless a certain condition is met."

The authors concluded:

"One may argue that the AOE statement is equivalent to invalidity of the law of total probability. However, as discussed earlier, the invalidity of the law of total probability is due to the fact that measurement C (or D) alters the initial quantum state that impacts the probability of outcome for measurement A (or B) since [A, C] 6= 0 (or [B, D] 6= 0). There is a logic gap to equate this reason with the statement “an observed event is not relative to anything or anyone”. Therefore, it appears that the violation of inequalities in [9] just reconfirms the consequence of non-commutative measurements and, therefore, the invalidity of the law of total probability in quantum theory, rather than confirming the invalidity of the AOE statement."

Therefore, it appears to be a matter of interpretation. Given the stochastic-quantum correspondence theorem, the Barandes' interpretation allows us to assume that the observed events are absolute, but nevertheless, the predicted probabilities violate the LF inequalities because it is not possible, from the non-Markovian transition maps, to conclude that $P(ab|xy) = \sum_{c,d} P(ab|cdxy) P(cd)$.

Lucas.

 

[Fra]

the invalidity of the law of total probability in quantum theory, rather than confirming the invalidity of the AOE statement."
...
Therefore, it appears to be a matter of interpretation.

Spot on.

I thini its a matter also of definition of "AOE". As I skimmed the Bong article, this was baked into their definition of AOE? This is why I said I think it's rejected. But definition was theirs, not my choice. IT sort of depends on what you really mean by "absoluteness", just ontic status or that it's inferrable by all. Two different things.

/Fredrik

 

[javisot]

Is Barandés' stochastic "interpretation" really an interpretation?

Something in QM explained using the Copenhagen interpretation can be translated into stochastic terms. Obviously, the translation tells the same story, but in a different way.

It does explain the same experimental data differently, but is it truly an interpretation?

 

[pines-demon]

Is Barandés' stochastic "interpretation" really an interpretation?

Something in QM explained using the Copenhagen interpretation can be translated into stochastic terms. Obviously, the translation tells the same story, but in a different way.

It does explain the same experimental data differently, but is it truly an interpretation?

In previous threads, I have argued that it is not an interpretation in the sense that it does not claim to give an explanatory interpretation of what happens. It is just like some kind of duality.

 

[Morbert]

Is Barandés' stochastic "interpretation" really an interpretation?

Something in QM explained using the Copenhagen interpretation can be translated into stochastic terms. Obviously, the translation tells the same story, but in a different way.

It does explain the same experimental data differently, but is it truly an interpretation?

It offers a reformulation that might prove to be useful beyond foundational/conceptual questions, and there would be no problem in someone adopting the formalism without subscribing to any speculative metaphysical hypotheses. This isn't unusual. E.g. David Wallace makes heavy use of the consistent histories formalism of QM to flesh out the many-worlds interpretation. In this sense, it is not, in and of itself, an interpretation.

But the reformulation does offer up a natural concomitant interpretation of quantum systems as non-Markovian stochastic systems.

 

[gentzen]

E.g. David Wallace makes heavy use of the consistent histories formalism of QM to flesh out the many-worlds interpretation.

I guess you are confusing David Wallace with Wojciech Żurek here.

 

[Morbert]

I guess you are confusing David Wallace with Wojciech Żurek here.

From section 3.9 "The decoherent-histories framework" in "The Emergent Multiverse" by Wallace

Following Gell-Mann and Hartle (1990), we can define the history operator [...] and a decoherence functional [...] The significance of all this formalism is summarized in the following theorem (first stated by Griffiths 1993, so far as I know) [...] From the Everettian perspective, the decoherence functional is a purely technical tool: its significance comes from the branching-decoherence theorem, which tells us that the vanishing of the decoherence function between any two distinct histories is a necessary and sufficient condition for a history space to have a branching structure. An alternative perspective, however—developed by Robert Griffiths (1984; 1996; 2002), Roland Omnés (1988; 1992; 1994), and (from a rather different viewpoint) by Murray Gell-Mann and James Hartle (1990; 1993; 2007)—was historically important and remains frequently discussed in the literature

 

[javisot]

It offers a reformulation that might prove to be useful beyond foundational/conceptual questions, and there would be no problem in someone adopting the formalism without subscribing to any speculative metaphysical hypotheses. This isn't unusual. E.g. David Wallace makes heavy use of the consistent histories formalism of QM to flesh out the many-worlds interpretation. In this sense, it is not, in and of itself, an interpretation.

But the reformulation does offer up a natural concomitant interpretation of quantum systems as non-Markovian stochastic systems.

True, what I mean is that we can use Copenhagen or many-worlds to calculate QM things, but we don't use Copenhagen or many-worlds to calculate things that aren't QM. In contrast, with stochastic processes we can calculate QM things and things that have nothing to do with QM.

I'm not sure we can call the Barandes stochastic interpretation an interpretation in the same sense as Copenhagen or many-worlds.

 

[Morbert]

True, what I mean is that we can use Copenhagen or many-worlds to calculate QM things, but we don't use Copenhagen or many-worlds to calculate things that aren't QM. In contrast, with stochastic processes we can calculate QM things and things that have nothing to do with QM.

I'm not sure we can call the Barandes stochastic interpretation an interpretation in the same sense as Copenhagen or many-worlds.

The interpretation would be the conceptualization of a quantum system as a system with a definite classical configuration, evolving stochastically in accordance with non-Markovian dynamics.

 

[Fra]

Is Barandés' stochastic "interpretation" really an interpretation?

Something in QM explained using the Copenhagen interpretation can be translated into stochastic terms. Obviously, the translation tells the same story, but in a different way.

It does explain the same experimental data differently, but is it truly an interpretation?

It also offers a unique alternative interpretation to the ontological status of dynamical law and thus causality.

Instead of seeing "dynamical law" as global given constraint, Barandes sees it as a local non-markovian stochastic process, which is encoded at the core by a transition probability.

What makes it non-trivial and distinguishes it from simplistic "classical" brownian type motion, is the non-markovian property.

So without providing the full answer, Barandes implitictly but naturally reformulates the problem of understanding the origin of a particular law or constraints, as understanding the origin of the transition probability! Which is helpful for the unification quest, beyond the pure "qm interpretation" issues.

So it offers new insights? Which is I think the hallmark of an interpretation, even if it does resolve all problems.

/Fredrik