Why the De Raedt Local Realistic Computer Simulations are wrong

[harrylin]

See:
http://drchinese.com/David/DeRaedtComputerSimulation.EPRBwithPhotons.C.xls

Thanks for the link!

Again it is a completely artificial mechanism, so what you call it is completely irrelevant. When talking about a suppression mechanism, I may call mention Detector Efficiency while they call it Coincidence Time Window.

If I hear you correctly, what you call "detector efficiency" (which refers to a physical characteristic of the detector) is in fact the data picking by means of the time window - that is, a human choice.

But nothing changes. There is no more one effect than the other. As you look at more of the universe, you get farther and farther away from the QM predictions and that never really happens in actual experiments.

To the contrary, their simulation matches Weihs' experiment rather well on exactly that issue. That topic is discussed here: https://www.physicsforums.com/showthread.php?t=597171

So the Suppression Mechanism must grow if you DO want it to match experiment! And THAT is the Unfair Sampling Assumption.

Now THAT is a less well defined term. Perhaps most people mean with the Unfair Sampling Assumption a detector characteristic, but I agree with you that their model is based on an unfair data picking explanation. That could equally well be called an Unfair Sampling, or more precisely, Sub-sampling Assumption.

 

[DrChinese]

If I hear you correctly, what you call "detector efficiency" (which refers to a physical characteristic of the detector) is in fact the data picking by means of the time window - that is, a human choice.

It was an incorrect use of language on my part. Mentally, I group all models in which there is some bias which causes the accepted sample to differ sufficiently from the universe as a whole. But there are definite legitimate differences between the models.

So my apologies. I will use the term "coincidence time window" instead of detector efficiency, with the understanding that in a computer simulation, some of this is arbitrary. If it were to be considered a candidate model, you would want to challenge whether such an effect really existed. Specifically, how does the photon get delayed without losing its entangled characteristic (i.e. perfect correlations)? Because if it lost that, it should NOT be considered at all.

If you vary the k= setting (in the spreadsheet, tab B. Entangled) from 1 to 30 to 100 you will see how things change in a very unphysical manner.

 

[harrylin]

[..] Mentally, I group all models in which there is some bias which causes the accepted sample to differ sufficiently from the universe as a whole. But there are definite legitimate differences between the models. [..]

Good to see that we now agree on this, and apology appreciated.

I will use the term "coincidence time window" instead of detector efficiency, with the understanding that in a computer simulation, some of this is arbitrary. If it were to be considered a candidate model, you would want to challenge whether such an effect really existed. Specifically, how does the photon get delayed without losing its entangled characteristic (i.e. perfect correlations)? Because if it lost that, it should NOT be considered at all.

That is exactly the kind of things that I try to discuss in the thread on "ad hoc" explanations. However, if I'm not mistaken it was you who pointed out that certain interactions do not or hardly affect entanglement.

If you vary the k= setting (in the spreadsheet, tab B. Entangled) from 1 to 30 to 100 you will see how things change in a very unphysical manner.

I'll try that.

 

[DrChinese]

That is exactly the kind of things that I try to discuss in the thread on "ad hoc" explanations. However, if I'm not mistaken it was you who pointed out that certain interactions do not or hardly affect entanglement.

That is true, generally I would not expect that the transport mechanism would be much of a factor. However, I guess it is *possible* that one photon could have an interaction that would reveal its spin (of course not to us) AND delay it both. If that case occurred, for example, it correctly should not be considered as the pair is no longer entangled on the polarization basis.

 

[DrChinese]

By the way, this thread is dredged up from some time ago. I would like to say that the de Raedt team was kind enough to work with me to refine my spreadsheet model. After they supplied me with some modifications to their original Fortran code, my primary objection* to their model disappeared. I have not come to understand why it was able to accomplish that feat - simply because I have not devoted the time to the matter.

So while I disagree with Hans and Kristel on the conclusions that should be drawn from the model, I agree with its operation.

Here is the link to the Excel spreadsheet model:

http://drchinese.com/David/DeRaedtComputerSimulation.EPRBwithPhotons.C.xls

* Which had to do with a specific case of PDC simulation, not the general case.

 

[billschnieder]

I have not come to understand why it was able to accomplish that feat - simply because I have not devoted the time to the matter.
...
So while I disagree with Hans and Kristel on the conclusions that should be drawn from the model, I agree with its operation.

So you agree with the way it works although you do not understand why it works but you disagree with their conclusion nonetheless?

If you could be kind as to explain why you disagree with their conclusion, despite agreeing that their model is local and realistic, then we can discuss that.

 

[DrChinese]

If you could be kind as to explain why you disagree with their conclusion, despite agreeing that their model is local and realistic, then we can discuss that.

There are a lot of problems with the model when you talk about it as more than a computer simulation. I.e. if you want to consider it as somehow corresponding to something physical. Hard to know where to begin really, so here are my opinions for what they are worth:

The good:
- It is 100% local and realistic, so no issue there.
- It does violate a Bell Inequality, so it succeeds there.
- It did model product state statistics correctly when it needed to, which was my original objection to the simulation itself.

The bad:
- It posits physical effects that are new, and subject to experimental rejection or confirmation (don't hold your breath on that one).
- It only matches experiment when the window size is very small, otherwise it deviates quite quickly towards the Bell boundary.
- It beats the Bell Inequality when the window size is made to be medium, but only barely.
- And most telling, it does not match QM for the full universe. Now, you don't seem to think this is a problem but it really is quite serious for a model of this type. Because there would be tests that could be constructed to exploit this difference. This is part of the reason that the team has attempted to construct further simulations to take things a few steps farther.
- It does not match the dynamics of actual data when the time window is varied. I.e. it is obviously ad hoc.

-DrC

 

[billschnieder]

- It only matches experiment when the window size is very small, otherwise it deviates quite quickly towards the Bell boundary.

Just like the experiments it is modelling.

- It beats the Bell Inequality when the window size is made to be medium, but only barely.

Just like the experiments it is modelling

- And most telling, it does not match QM for the full universe. Now, you don't seem to think this is a problem but it really is quite serious for a model of this type. Because there would be tests that could be constructed to exploit this difference.

There is no such thing as QM prediction for "full universe". QM only makes predictions about actual measurement outcomes.

- It does not match the dynamics of actual data when the time window is varied.

Just like the experiments it is modelling. QM doesn't match the experiments either when the time window is varied.

 

[DrChinese]

1. There is no such thing as QM prediction for "full universe". QM only makes predictions about actual measurement outcomes.

2. Just like the experiments it is modelling. QM doesn't match the experiments either when the time window is varied.

1. Of course it does. The expectation is cos^2(theta) always. But that is not the case with the De Raedt et al model.

2.Not so! Otherwise it wouldn't be an issue.

 

[billschnieder]

1. Of course it does. The expectation is cos^2(theta) always. But that is not the case with the De Raedt et al model.

2.Not so! Otherwise it wouldn't be an issue.

1) cos^2(theta) is the expectation value for OUTCOMES. QM does not predict anything other than what is observed! You change the time window you get a DIFFERENT observation! Looking at stuff that is not observed and calling ing "full universe" is simply wrong-headed.
2) This is false. Look at figure 2 in their article in which they analyze the actual experimental data, varying the window: http://arxiv.org/pdf/1112.2629v1.pdf. QM is violated by 5 standard deviations!

 

[harrylin]

1) cos^2(theta) is the expectation value for OUTCOMES. QM does not predict anything other than what is observed! You change the time window you get a DIFFERENT observation! Looking at stuff that is not observed and calling ing "full universe" is simply wrong-headed. [..]

While De Raedt et al's simulation did succeed in its intended purpose, it does appear that in principle (and likely also in practice) their model makes slightly different predictions from QM. That may allow for a comparison of both with existing data, as is intended in the thread on "Weih's data" (should have been Weihs' data).

 

[DrChinese]

1) cos^2(theta) is the expectation value for OUTCOMES. QM does not predict anything other than what is observed! You change the time window you get a DIFFERENT observation! Looking at stuff that is not observed and calling ing "full universe" is simply wrong-headed.

"Full universe" is what they usually call the portion that is not included in a sample (along with the sample itself of course). What do you call that? Because QM makes the same prediction for everything, while the de Raedt et al model does not. In that model, there is always a difference between the sample and the full universe.

 

[billschnieder]

"Full universe" is what they usually call the portion that is not included in a sample (along with the sample itself of course).

I know what "full universe" means, my point is that it does not apply to QM. What does QM predict for the "stuff" that is not sampled (ie, is not measured)? Don't you realize that QM says nothing about what is "there" beyond the measurement results?

The measurement is the sampling, QM predicts what the sample will show, not what exists apart from the sample which you call "full universe".

What do you call that? Because QM makes the same prediction for everything

No. The everything you refer to is "everything that is measured" as far as QM is concerned, not everything that exists apart from the measurement. What you call full-universe in their simulation would be "hidden" if it were a real experiment so you can't compare that with QM.

 

[DrChinese]

I know what "full universe" means, my point is that it does not apply to QM. What does QM predict for the "stuff" that is not sampled (ie, is not measured)? Don't you realize that QM says nothing about what is "there" beyond the measurement results?

Groan.

The simulation allows us to include as much as 1 pair from every trial, and in all cases shows us the other 2 pairs from every trial. Which of the 3 is selected for viewing is random and independent of the angle settings. That means it fulfills the local realism requirement. The 2 pairs not selected are then disposed of. That is not part of the full universe I am discussing.

The full universe is the portion we are sampling from. QM says the full universe is cos^2(theta). It is an experimental fact that the sample we actually measure respects that. Again, for QM the full universe does not include counterfactual angles, it only includes the angles we actually measure at.

By way of analogy: GR describes the relative mutual attraction of any 2 objects. The prediction for the full universe is the same as the prediction for any sample. This is normal in science, Bill. We have a theory which describes the full universe, and experiment which measures a sample. So too in this simulation. And the full universe does NOT match QM.

 

[billschnieder]

Groan.
The simulation allows us to include as much as 1 pair from every trial, and in all cases shows us the other 2 pairs from every trial. Which of the 3 is selected for viewing is random and independent of the angle settings. That means it fulfills the local realism requirement. The 2 pairs not selected are then disposed of. That is not part of the full universe I am discussing.

And that is not what I understand you to be discussing either!

The full universe is the portion we are sampling from.

Yes, I understand that this is what you mean.

QM says the full universe is cos^2(theta).

No! QM says no such thing. QM predicts, and agrees ONLY with the result of the sampling !

It is an experimental fact that the sample we actually measure respects that.

Yes the sample actually measured respects QM, it agrees with what QM predicted for the sample. The sample from the simulation also agrees with what QM predicted for the sample and it also agrees with the sample from the experiment.
Your suggestion that full universe of the simulation does not match QM's prediction for the sample is the wrong-headedness I'm pointing out to you.

Again, for QM the full universe does not include counterfactual angles, it only includes the angles we actually measure at.

AND again , I understand exactly what you mean by "full-universe". Now you try to understand what I mean when I say QM does not predict any "full-universe".
Your error is to ascribe the prediction of QM to a full-universe not realizing that QM's prediction is for a measurement outcome which is necessarily a sample.

 

[harrylin]

[..] AND again , I understand exactly what you mean by "full-universe". Now you try to understand what I mean when I say QM does not predict any "full-universe".

I kind of foresaw that issue and didn't get into that (although my reply did imply a middle ground answer) as I think that such arguments like "full universe" automatically disappear when discussing a real experimental data "universe".

 

[DrChinese]

Your error is to ascribe the prediction of QM to a full-universe not realizing that QM's prediction is for a measurement outcome which is necessarily a sample.

This is a useless statement. QM DOES make the same prediction, whether the measurement is performed or not, as long as it is possible in principle to make the measurement. The only thing about QM that is different than other scientific theories is that it makes no prediction for measurements that could not be performed, in principle, such as one counterfactual to an actual measurement.

In the simulation there is a sample and there is a full universe. YOU CAN SEE BOTH, so don't say they don't exist. We can talk about them meaningfully, and we can compare to the QM expectation value for the full universe, which has meaning according to the EPR definition. And the sample is not a faithful representation of the full universe in many cases, and in all cases the universe does not match a QM universe. In accordance with Bell.

You would stick an ice pick in your eye before you would admit how wrong you are, so I am not going to keep going in circles with you.

 

[billschnieder]

QM DOES make the same prediction, whether the measurement is performed or not, as long as it is possible in principle to make the measurement.

Of course, that is what "prediction" means. But realize what the prediction is for. It is a prediction "for an experimental outcome". Not a prediction for the full-universe that you are talking about.

In the simulation there is a sample and there is a full universe. YOU CAN SEE BOTH, so don't say they don't exist.

No question there. Nobody says they don't exist, in the simulation you can see everything, you can even see photons that get lost, you can even trace the photons one by one and see what happens to each one. The point you refuse to see is that the QM prediction is for the outcome, so you must compare the QM prediction with the outcome of the simulation, not what exists in the simulation beyond the outcome. The only relevant question is: Have they obtained the outcome in a manner that, in principle, is reasonably consistent with the way real outcomes are obtained in the real experiments? If the answer to this question is yes, then you compare the outcome with the outcome of the real experiments, and to the QM prediction for the real experiments. You already agree that their outcomes agree with QM. You already agree that they obtain the outcomes in a manner consistent with Local reallity. So what is the beef? This "full-universe argument does not make sense.

We can talk about them meaningfully

Of course you can.

and we can compare to the QM expectation value for the full universe, which has meaning according to the EPR definition.

I've said this too many times already. The QM expectation value is for the outcome. QM is answering the question "What is the expectation value E(a,b) for the outcome IF we measure along a and b"?

And the sample is not a faithful representation of the full universe in many cases, and in all cases the universe does not match a QM universe. In accordance with Bell.

I could not be any clearer, what you call "QM universe" is not comparable to what you call "full-universe" in the simulation. They are apples and oranges.

 

[lugita15]

billschnieder, this is what I told you in another thread:

The "full universe" issue you're talking about concerns the existence of counterfactual outcomes. But the "full universe" issue that DrChinese is discussing in regard to de Raedt's model is that it exploits the fair sampling loophole: the model only reproduces the predictions of QM if we take a small coincidental detection window, but if we had better experiments that would detect ALL entangled pairs emitted by the source, then de Raedt's model would be in stark disagreement with the predictions of QM.

 

[billschnieder]

billschnieder, this is what I told you in another thread:

Thanks Lugita,
I understand what DrC means by full-universe, the disagreement concerns the fact that he thinks QM predicts a full-universe apart from what is actually measured, and I think QM predicts ONLY what is measured and no more.

Another way of looking at it as per your quote is that DrC thinks in QM coincidence window is infinite, but I think in QM the coincidence window is 0.

 

[lugita15]

Thanks Lugita,
I understand what DrC means by full-universe, the disagreement concerns the fact that he thinks QM predicts a full-universe apart from what is actually measured, and I think QM predicts ONLY what is measured and no more.

I still think you're not understanding DrChinese. When he says the phrase "full universe" in the context of de Raedt's model, he is NOT talking about realism. He is talking about the fair sampling loophole, which only exists because of current experimental limitations. If our experimental equipment was good enough, the fair sampling loophole would be closed, and it may be possible to test the differences between standard quantum mechanics. Do you not agree that that's what he's talking about, or do you agree that that's what he's talking about but disagree with him on it?

Another way of looking at it as per your quote is that DrC thinks in QM coincidence window is infinite, but I think in QM the coincidence window is 0.

I am using the phrase "coincidence window" in a very precise experimental sense, you are using it with a meaning that I don't recognize. The coincidence window is not something theory-specific, so it makes no sense to ask what the coincidence window is "in QM". The coincidence window is how long you let the photon detectors run, waiting for each entangled particle to hit the respective detector. If we set the window too short, we may miss some of the particles that are still on their way. If we set the window too long, we may get confused as to which photons belonged to which particle pair. This is just a practical experimental problem, and if and when it is resolved the predictions of QM and de Raedt will presumably no longer be experimentally indistinguishable.

 

[billschnieder]

I still think you're not understanding DrChinese. When he says the phrase "full universe" in the context of de Raedt's model, he is NOT talking about realism.

You think you know what I understand but you don't. I understand exactly what DrC is talking about.

He is talking about the fair sampling loophole, which only exists because of current experimental limitations. If our experimental equipment was good enough, the fair sampling loophole would be closed, and it may be possible to test the differences between standard quantum mechanics.

And this is precisely the "blinders" both of you have on which is preventing you from understanding what I'm saying (and the simulation). You assume that lack of fair sampling is a loophole only due to problems with equipment and we can "fix" it by improving the experiment.

I am using the phrase "coincidence window" in a very precise experimental sense, you are using it with a meaning that I don't recognize. The coincidence window is not something theory-specific, so it makes no sense to ask what the coincidence window is "in QM".

Let us think about your ideal experiment in which the photons paths are exactly the same length and the clocks are exactly synchronized, and no stray photons are present. I presume you mean that in such a case, this "loophole" will not exist because the two photons will have exactly the same time of arrival. Yes? So setting the coincidence window to zero should more accurately represent the QM case right? And increasing it much larger than zero should deviate from QM right? So hopefully now you understand why W=0 is equivalent to the QM prediction for an ideal setup.
Here is what De Raedt says:

In this case, both the simulation and a rigorous mathematical treatment of the model lead to the conclusion that for d = 3 and W → τ → 0, the model reproduces the results (see Table I) of quantum theory for a system of two S = 1/2 particles.

So what has "full-universe" got to do with it. What is the full universe in your view that is supposedly violates QM. Now maybe this phrase from their paper addresses exactly what I'm talking about:

Another deceptive point may be that in our model, one can compute the correlation of the particles right after they left the source. This correlation is exactly minus one. However, this correlation has no relevance to the experiment: To measure the correlation of the particles, it is necessary to put in the Stern-Gerlach magnets, detectors, timing logic and so on. We emphasize that the simulation procedure counts all events that, according to the same criterion as the one employed in experiment, correspond to the detection of two-particle systems.
Our simulation results also suggest that we may have to reconsider the commonly accepted point of view that the more certain we are about a measurement, the more ”classical” the system is. Indeed, according to experiments and in concert with the prediction of our model, this point of view is in conflict with the observation that the more we reduce this uncertainty by letting W → 0, the better the agreement with quantum theory becomes.
Both in experiments and in our model, the uncertainty is in the time-tag data and it is this uncertainty that affects the coincidences and yields the quantum correlations of the singlet state (if W → 0). Isn’t it then very remarkable that the agreement between experiment and
quantum theory improves by reducing (not increasing!) the uncertainty by making W as small as technically feasible?

So think about a coincidence window of zero. I ask again, what is this full-universe which DrC claims violates QM?

This is just a practical experimental problem, and if and when it is resolved the predictions of QM and de Raedt will presumably no longer be experimentally indistinguishable.

So let me get this straight, you are saying that the simulation agrees with QM and Experiment because the experiments are faulty, but when the experiment becomes ideal, they will continue to agree with QM but not with the simulation? Does that make sense to you?

 

[harrylin]

[..] The coincidence window is not something theory-specific, so it makes no sense to ask what the coincidence window is "in QM".

I agree with that. However:

The coincidence window is how long you let the photon detectors run, waiting for each entangled particle to hit the respective detector. [..]

Here I think that you misunderstand the usual experimental set-up, or at least those that De Raedt et al have in mind. The photo detectors run nearly all the time, and anyone who has the data can freely chose the coincidence window for data analysis.