I A new realistic stochastic interpretation of Quantum Mechanics

[martinbn]

They are describing their data in a particular way, in terms of "subensembles". The only reason for doing that is to perform statistical analysis. If we are discussing interpretations, such an analysis is only relevant for statistical interpretations.

I think they only say to look at the sets of results of Alice and Bob, and the subsets of for which e specific outcome happened at Victor's measurment on photons 2&3. No interpretation is needed here.

 

[iste]

Is that demanded by the math? Because I understand that this is the conclusion given on the paper, but when I look at the math used the only thing I gather is that in between division 1 and division 2 the particle is unknowable and trying to predict its properties gives us all of QM weirdness.

Yes, it is implied by the fact that you are talking about a stochastic system. If you sample the system, it has definite configurations at every point in time. Furthermore, you can always assign probabilities to any point in time, the system having started at an initial time. What you cannot do is take the system moving from initial time to another point in time and then construct conditional probabilities regarding the intermediate times. Particles are in definite configurations between division events.

I don't see why a regular particle with definite positions wouldn't be able to be describe by regular statistics.

There is a distinction between the statistics that describe particle behavior and the fact that it exists in definite configurations. Similar to how there is a distinction between the probabilities you assign to dice rolls and the fact that you can roll dice and get a specific outcome. Superpositions reflect the statistical description and so there is nothing inconsistent about the particle being in definite configurations during superposition.

 

[PeterDonis]

If interpretations are independent of the predictions of quantum mechanics then they should be independent of the statisticsl analyses experimenters perform.

To the extent that the statistical analysis is to compare predictions with experiment, of course it is independent of any interpretation, yes.

But that very fact means that you cannot use anything about the above to argue for or against any particular interpretation. In the post of yours that I responded to in post #383, you were trying to use the statistical subensembles to argue against the interpretation @DrChinese is using. At least, that's the only reading of your post I can come up with that makes it relevant at all to this thread.

 

[pines-demon]

In the end, you could simply diagonalize the density matrix at t=0, use a basis compatible with it (or a suitable coarse graining) at t=0, and then "rotate as fast as you wish" to the basis which you really want to work in (i.e. the one giving the "true ontology").

What do you mean here?

 

[gentzen]

In the end, you could simply diagonalize the density matrix at t=0, ...

What do you mean here?

You mean on a high level? Basically the same as he in https://arxiv.org/abs/2309.03085 on page 2:

Without any important loss of generality, the set of times $\mathcal{T}$ will be assumed to include an element denoted by $0$ and called the initial time. It will be further assumed that at the initial time $0$, the time-dependent dynamical map $f_t$ trivializes to the identity map $\operatorname{id}_\mathcal{X}$ on $\mathcal{X}$:
$f_0 = \operatorname{id}_X ,\quad\text{ or }\quad f_0(i) = i\quad[\text{for all i }\in \mathcal{X}]. (6)$

I mean: you can finesse this point in various ways if you want. But it is more a question of aesthetics and simplicity than a real limitation of generality.

You want to have a time dependent quadratic stochastic matrix as your starting point. Using a division event at $t=0$ is a simple way to achieve this, therefore Barandes did this. You could also be "more general" and introduce a linear dependence on an abstract probability vector (of same dimension as the phase space), without a connection to any specific time $t_0$. But it might feel a bit more complicated and less well motivated. But you could motivate this abstract probability vector as the eigenvalues of the density matrix.

What I described was how to stay with the division event at $t=0$, and still use the eigenvalues of the density matrix as the abstract probability vector.

 

[iste]

To the extent that the statistical analysis is to compare predictions with experiment, of course it is independent of any interpretation, yes.

But that very fact means that you cannot use anything about the above to argue for or against any particular interpretation. In the post of yours that I responded to in post #383, you were trying to use the statistical subensembles to argue against the interpretation @DrChinese is using. At least, that's the only reading of your post I can come up with that makes it relevant at all to this thread.

I don't think a statistical conditioning interpetation of an experiment has to necessarily be tied to a specific interpretation: e.g. arguing against retrocausality in delayed choice.

https://scholar.google.co.uk/scholar?cluster=3918555048962557101&hl=en&as_sdt=0,5&as_vis=1

I was initially responding to a point where Dr. Chinese invoked delayed choice.

 

[PeterDonis]

I don't think a statistical conditioning interpetation of an experiment has to necessarily be tied to a specific interpretation

If "interpretation of an experiment" just means "comparing QM predictions to the results", then of course not.

But again, that means you can't use "statistical conditioning" to argue against any QM interpretation.

 

[JC_Silver]

Yes, it is implied by the fact that you are talking about a stochastic system. If you sample the system, it has definite configurations at every point in time. Furthermore, you can always assign probabilities to any point in time, the system having started at an initial time. What you cannot do is take the system moving from initial time to another point in time and then construct conditional probabilities regarding the intermediate times. Particles are in definite configurations between division events.



There is a distinction between the statistics that describe particle behavior and the fact that it exists in definite configurations. Similar to how there is a distinction between the probabilities you assign to dice rolls and the fact that you can roll dice and get a specific outcome. Superpositions reflect the statistical description and so there is nothing inconsistent about the particle being in definite configurations during superposition.

I'm not gonna lie, this is what gets me about the non-Markovian process, because the dice rolls isn't a non-Markovian process, it's a regular stochastic process. Since I'm not well versed in stochastic processes of any kind and I can't find good sources on non-Markovian processes online that are not by Barandes (as we used to say in the long past, my Google-fu isn't strong enough), I'm left not understanding exactly why the particle has definite positions between division events, because as far as I understand, the dice roll also doesn't exist between one roll and the next.

Again, sorry for bothering >.<

 

[DrChinese]

It occurs when you compare the experimental results to the predictions of QM. Those predictions are statistical. In order to do the statistics right, you have to separate the runs into buckets based on the BSM results, and test the observed statistics against QM predictions for each bucket separately.

No, there isn't, because there is no way to predict which of the four possible Bell states will be produced by the BSM interaction for each individual run. You can only post-select the runs into buckets after the fact, when you know the BSM results. That's not prediction, that's post-selection.

Yes, technically there is separation of the results into the 2 discernible buckets (Bell states) of 4 fold coincidences. Calling it "post-selection" may be correct at some level, but it is also misleading to readers. We combine the results of ALL 4 near-simultaneous detections per specification, which is normal for all experiments. Nothing is being "thrown away" because it doesn't fit our ideas of what is being studied. As I described, it is also possible to use polarizers instead of polarizing beam splitters; in that case, there is only a single bucket: would you call that "post-selection" too?

And calling the results in each bucket as being statistical is also misleading to some readers. The prediction is certain in each and every case. (And for this statement I am referring to the Ma experiment specifically, as it does not feature a CHSH-like inequality.)

 

[pines-demon]

I can't find good sources on non-Markovian processes online that are not by Barandes (as we used to say in the long past,

It is normal, Barandes himself says that there is almost no literature on non-Markovian processes, and most of the literature out there is about non-Markovian divisible processes, so Barandes stuff is very niche.

 

[DrChinese]

1. Its not if one thinks that a BSM is formally equivalent to statistical conditioning. This is what the Barandes paper would imply and I have read other papers that would imply similarly. Its not factually incorrect. And I'm am pretty sure a Bell state implies a correlation so I would say it is correlating. You are correlating by picking out ensembles where you always have the same two outcomes.


2. But they would have a correlation in the same basis right? They are entangled... which isn't even something I am contesting.

1. It is NOT statistical conditioning (the Barandes view). You read the exact same values for the 2 bits used to identify the Bell state (essentially as correlated or anti-correlated) regardless of whether there is a swap or not. The swap ONLY occurs if there is 2&3overlap in the beamsplitter. The experiment explains this.


2. No, they are NOT entangled when measured on the H/V basis UNLESS the swap occurs - which is the a) side. When there is distinguishability - the b) side - they are not entangled but they do yield the same statistics (as they authors show).

Even a broken clock displays the correct time twice a day. This is the same example.

 

[DrChinese]

How do you define "FTL action"?

I think the normal definition for an FTL action is: An effect that is occurs over a spacetime distance (between 2 spacetime points) that exceeds an effect that is limited to c. I include in that situations in which temporal order is blurred - delayed choice situations for example. And the effect need not go from one spot to the other (i.e. it need not be clearly directional).

 

[martinbn]

I think the normal definition for an FTL action is: An effect that is occurs over a spacetime distance (between 2 spacetime points) that exceeds an effect that is limited to c. I include in that situations in which temporal order is blurred - delayed choice situations for example. And the effect need not go from one spot to the other (i.e. it need not be clearly directional).

But in the situation above, the effect is in the past light cone of the cause, isn't it? They are not just space like seperated. You could send a message from an event to the event that caused it!

 

[Fra]

I think the normal definition for an FTL action is: An effect that is occurs over a spacetime distance (between 2 spacetime points) that exceeds an effect that is limited to c. I include in that situations in which temporal order is blurred - delayed choice situations for example. And the effect need not go from one spot to the other (i.e. it need not be clearly directional).

As you require no causal order for term, do you agree with

FTL action = Spacelike correlation ?

And then "correlation" can either be understood as statiatical dependence or on single instance basis? Or would you use different terms for the two variants?

/Fredrik

 

[iste]

You read the exact same values for the 2 bits used to identify the Bell state (essentially as correlated or anti-correlated) regardless of whether there is a swap or not. The swap ONLY occurs if there is 2&3overlap in the beamsplitter. The experiment explains this.

Nothing incompatible here with anything I have said.

2. No, they are NOT entangled when measured on the H/V basis UNLESS the swap occurs - which is the a) side. When there is distinguishability - the b) side - they are not entangled but they do yield the same statistics (as they authors show).

Again, I don't know if there is anything that prima facie contradicts statistical conditioning here given that the correlations in the raw data of separable measurements can be reproduced by mixing the data sets of Bell states, removing the correlations from the other bases but preserves correlations in one of them. Transitivity then allows correlations to show up in 1 & 4 as far as what the non-separable and separable cases allow.

 

[PeterDonis]

I don't know if there is anything that prima facie contradicts statistical conditioning here

If you mean, does this experimental data rule out a statistical interpretation of QM, no, it doesn't. No interpretation of QM can be ruled out by experimental data, since all interpretations use the same (or equivalent) math, the math of standard QM, and so they all make the same predictions for all experiments. That is a fundamental premise for all discussions in this forum. And because of that, trying to use experimental data to either prove or disprove any particular interpretation is out of bounds here. The best we can do is to express opinions about how we think particular experiments make some interpretations more or less plausible or acceptable to us. @DrChinese's position appears to be that the experiments he describes make a statistical interpretation of QM extremely implausible (to the point where it is unacceptable to him). Yours appears to be that they don't. There is no way to resolve such a disagreement (and that is also a fundamental premise for all discussions in this forum).

There are hypotheses in the literature which are sometimes called "interpretations" of QM, but which are really different theories, because they make different predictions. An example is the GRW stochastic collapse model (which is not the same, as far as I can tell, as the stochastic interpretation that the OP of this thread refers to.) Those should really be discussed in the Beyond the Standard Models forum.

 

[iste]

If you mean, does this experimental data rule out a statistical interpretation of QM, no, it doesn't.

No, I have been talking about normal statistical conditioning, not interpretation. We are talking about whether a very specific aspect of the experimental results that Dr. Chinese pulled out of a particular experiment can be explained by statistical conditioning on pairs of outcomes from two different entanglements. Its a question how the experiment works, my argument citing a specific passage where the authors report their data which is prima facie compatible with what I said about this in previous posts, regarding how the experiment works.

 

[PeterDonis]

I have been talking about normal statistical conditioning, not interpretation.

This thread is in the interpretations subforum. The subject is a stochastic interpretation. If you are not talking about interpretation, then your posts are off topic for this thread, in this subforum.

We are talking about whether a very specific aspect of the experimental results that Dr. Chinese pulled out of a particular experiment can be explained by statistical conditioning on pairs of outcomes from two different entanglements.

Of course it can. That is an obvious consequence of the fact that the predictions of the standard math of QM, independent of any interpretation, match the experimental results.

It is also irrelevant to this thread. This thread, in this subforum, is about QM interpretations, not the comparison of QM predictions with results. As far as interpretations go, as I have already said, the comparison of QM predictions with results cannot possibly rule out any interpretation, since they all use the same math of standard QM and all make the same experimental predictions.

Basically, as I've already said, you are saying that you don't think these experiments make the preferred interpretation of @DrChinese any more plausible, or a statistical interpretation any less plausible. Okay, noted. But continuing to belabor the point serves no purpose.

 

[iste]

I'm not gonna lie, this is what gets me about the non-Markovian process, because the dice rolls isn't a non-Markovian process, it's a regular stochastic process. Since I'm not well versed in stochastic processes of any kind and I can't find good sources on non-Markovian processes online that are not by Barandes (as we used to say in the long past, my Google-fu isn't strong enough), I'm left not understanding exactly why the particle has definite positions between division events, because as far as I understand, the dice roll also doesn't exist between one roll and the next.

Again, sorry for bothering >.<

The particle is in a definite position at every point in time but the position is subject to random change over time. Because of the randomness, the behavior of definite particles can only be described by attaching probabilities on top. Indivisibility is in regard to the ability to assign certain probabilities describing particle behavior. Division events momentarily allow the application of that type of probability, but only temporarily.

 

[iste]

Basically, as I've already said, you are saying that you don't think these experiments make the preferred interpretation of @DrChinese any more plausible, or a statistical interpretation any less plausible. Okay, noted. But continuing to belabor the point serves no purpose.

Well presumably everytime Dr. Chinese brings up entanglement swapping to attack another interpretation he is breaking the rules then.

 

[DrChinese]

But in the situation above, the effect is in the past light cone of the cause, isn't it? They are not just space like seperated. You could send a message from an event to the event that caused it!

No signal possible faster than light, no signal to the past either

 

[DrChinese]

Nothing incompatible here with anything I have said.

Again, I don't know if there is anything that prima facie contradicts statistical conditioning here given that the correlations in the raw data of separable measurements can be reproduced by mixing the data sets of Bell states, removing the correlations from the other bases but preserves correlations in one of them. Transitivity then allows correlations to show up in 1 & 4 as far as what the non-separable and separable cases allow.

Except that there is one result that is different when there is physical overlap that does not occur otherwise. If you don’t want to call that action, fine.

There is no such thing as transitivity in this context.

 

[PeterDonis]

Well presumably everytime Dr. Chinese brings up entanglement swapping to attack another interpretation he is breaking the rules then.

"Attack" is a vague term. His position, as I have already said, is that he thinks entanglement swapping makes certain interpretations much less plausible. That is his opinion. You are free to have your own, different opinion.

 

[PeterDonis]

calling the results in each bucket as being statistical is also misleading to some readers. The prediction is certain in each and every case. (And for this statement I am referring to the Ma experiment specifically, as it does not feature a CHSH-like inequality.)

"Certain" is an idealization. The prediction is only certain if it is guaranteed that all of the polarization measurements are made in exactly the same direction (or in exactly orthogonal directions). In any real experiment, that cannot be guaranteed. There is always some finite error involved. That means there is always some need for statistical analysis just to deal with that source of error.

I agree that experiments in which the idealized prediction is not an inequality but a definite single result are more definitive in terms of confirming the QM predictions independently of any particular interpretation, because the role statistics plays is narrower.

However, they do not pose any problem for a statistical interpretation of QM, because "statistics" includes the edge cases where the prediction is 100% probability for one result and 0% probability for any other result. And to robustly test such predictions, you still need a large number of runs. Just one run with the predicted result is a very weak test.

 

[iste]

Except that there is one result that is different when there is physical overlap that does not occur otherwise. If you don’t want to call that action, fine.

And the distinction between nonseparable and separable systems exists in the Barandes formulation so nothing is missing. The separable HHVV measurement data is statistically the same as when you just mix the Bell state HHVV+ and HHVV- data together. Mix the HHVV+- with HVHV+- and all correlations disappear. A hierarchy of correlations in the raw data.

There is no such thing as transitivity in this context.

There is always transitivity. A is only ever correlated with D because A is correlated with B is correlated with C is correlated with D, entanglement or separable.

 

[Fra]

As you require no causal order for term, do you agree with

FTL action = Spacelike correlation ?

And then "correlation" can either be understood as statiatical dependence or on single instance basis? Or would you use different terms for the two variants?

/Fredrik

Except that there is one result that is different when there is physical overlap that does not occur otherwise. If you don’t want to call that action, fine.

I think this response to the another post answers my question too with a yes?

I think the confusion was the word "action" as i tend to associate it with "causal influence" in that, someone has the CHOICE to take an action, and this influences something else, but its not quite what you meaa I think.

Yes the experimenter can choose to mess upp the experiment at will, to make sure you get NO "tags" to send, or it can choose to NOT send the BSM results, but destroy it, but even when doing it's BEST, it can not CHOOSE which of the possiblities that you get for an individual 2&3 pair. Thus the "choice" the experimenter have, is not a choice of that kind.

So with your definition then I agree there is "FTL action", but it is not conceptually problematic to my understanding.

I think any physical experiments requires extreme leves of control. So it is obvious that an experiementer has a "choice" to destroy the inferences from the experiments. But I think that physical laws represents are somehow the "optimal inferences" that are possible; ie given a perfect experimental control and given that exprimenters make no human errors.

/Fredrik

 

[Fra]

It is normal, Barandes himself says that there is almost no literature on non-Markovian processes, and most of the literature out there is about non-Markovian divisible processes, so Barandes stuff is very niche.

This makes sense, I think the potential extensios to Barandes perspective may required a radical change in the paradigm of which we view physical law. I relate to his, as the same is try for my own interpretation.

But I would propose that a nice context to view Barandes stochastics (though I am not sure what he thinkgs of it) is via an agent based model, where agents actions are "stochastic", but guided by conditional probabilities. This may give insight to the "nature" of the stochastics in a hidden variabl pace, in constrast to bell style HV.

This is paper on AI that associates the need for non-markovian decisions to the existence of "hidden states", hidden from the agents "immediate sensation". This is also the explicit meanig of the "hidden layers" in neural network models. So there is IMO many reasons to associate the non-markovian appearance with this type of "hidden variables"
https://doi.org/10.7551/mitpress/2026.003.0019

The insight you can more easily get from this, when considering "interacting" agents, is that interference patterns can happen, because their "hidden variables" are NOT of the type that Bell envisions; so they will not restore determinism, but they might provide other benefits in model building.

I think the synthesis of these ideas with foundational physics, is very immature subject and this is why there isn't alot of published. But I expect alot more from this in the future.

While that is beyond anything Barandes says, he writes himeself that he sees future possibilities in various areas, and it would be nice to see him elaborate on this. Had Barandes been a mathematician first, I could have imagine that he has no such idea. But as he seems to come from philosophical angles, it would appear strange to me, if he didn't have at least something to add here?

/Fredrik

 

[DrChinese]

1. And the distinction between nonseparable and separable systems exists in the Barandes formulation so nothing is missing. The separable HHVV measurement data is statistically the same as when you just mix the Bell state HHVV+ and HHVV- data together. Mix the HHVV+- with HVHV+- and all correlations disappear. A hierarchy of correlations in the raw data.

2. There is always transitivity. A is only ever correlated with D because A is correlated with B is correlated with C is correlated with D, entanglement or separable.

Again, let’s get the facts straight.

1. There is no mixing! The same 2 bits are revealed at the BSM - identifying the possible Bell swapped state - regardless of whether there is distinguishable 2&3 or not. That’s my point!! So the correlation does disappear for the distinguishable case + for example, because there is no swap. But not because that case wasn’t identified. Ditto for the - case. These were separately identified and counted.

Please note that your mention of HHVV and HVHV type cases is actually not useful, because the correlation exists there even without a swap. It is the +HH+ and -VH+ type cases that demonstrate the effect.

2. Note again there is NO transitive effect as you imagine. Yes, there appears to be some - even when entanglement is not swapped- when you measure on the H/V basis. But all normal swapping tests use a different basis such as +/- or L/R. Those are not “transitive” in your sense. The actual experimental results show this point, so there is nothing to dispute on your part.

 

[DrChinese]

I think the confusion was the word "action" as i tend to associate it with "causal influence" in that, someone has the CHOICE to take an action, and this influences something else, but its not quite what you meaa I think.

Yes the experimenter can choose to mess upp the experiment at will, to make sure you get NO "tags" to send, or it can choose to NOT send the BSM results, but destroy it, but even when doing it's BEST, it can not CHOOSE which of the possiblities that you get for an individual 2&3 pair. Thus the "choice" the experimenter have, is not a choice of that kind.

So with your definition then I agree there is "FTL action", but it is not conceptually problematic to my understanding.

I think any physical experiments requires extreme leves of control. So it is obvious that an experiementer has a "choice" to destroy the inferences from the experiments. But I think that physical laws represents are somehow the "optimal inferences" that are possible; ie given a perfect experimental control and given that exprimenters make no human errors.

/Fredrik

Again, if the experimenter’s distant control of entanglement does not conflict with your (or your reading of Barandes’) interpretation: then all good. No one is claiming that there is FTL action that is useful for signaling. The specific Bell state that results is random, and not under the experimenter’s free choice.

 

[iste]

because the correlation exists there even without a swap. It is the +HH+ and -VH+ type cases that demonstrate the effect

The point is that it seems to be the case that if you take all of the possible data from 2 & 3, there is always ways of sorting the data where you have Bell state correlations, separable correlations and no correlations whatsoever. The fact that you can mix the Bell state data and make the correlations in other bases disappear is indicative of this because even though those correlations have disappeared, its just the same data mixed up. The new mixed up data is then statistically the same as the separable measurements data (even though it is made up of swap data).

This seems strongly analogous to other delayed-choice eraser scenarios where adding together the coherent interference patterns on eraser idler photon screens just results in the same clump patterns as on the which-way idler photon screens and also the signal photon screen. What is happening in these experiements is that, via interference, the beam-splitter is affecting how the data from the entanglements is physically being sorted in the idler photons. Obviously the beamsplitter and which-way cases are very different physically, but the data just adds up. You can then formulate the analogue of this in entanglement-swapping.

The same 2 bits are revealed at the BSM - identifying the possible Bell swapped state - regardless of whether there is distinguishable 2&3 or not

And this is also pretty clearly the case in the part where they unambiguously describe mixing Bell states and report the statistical correlations they do and do not find. The mixing I am referring to is very unambiguously decribed in the paper you linked. I think you need to re-read that part of the paper.

Those are not “transitive” in your sense

They are in precisely the sense I mean - in the sense that the particles possess correlations and if you should chose to make measurements you would find those correlations. The fact that you happen to choose to measure some bases or not is immaterial, just like the fact that the entangled photons will have correlations in all bases but you clearly cannot ever simultaneously measure them at the same time.