Is QM Inherently Non-local in EPR and Bell Discussions?

[DrChinese]

Hi CJ, you must realize that there is usually much more to the things you learn than you teachers tell you. Let me first make some comments and then argue why the predictions of QM are very strange indeed.
(a) It is not SCIENTIFICALLY correct to state that Bell experiments refute local hidden variables/objective local theories.

Start your own thread if you want to push your personal agenda on Bell tests.

 

[Careful]

This is a correct statement. However, what is usually meant with the statement is that in those situations where quantum mechanics makes idealised predictions that DO violate the Bell inequalities, if we add to that the (quantum-mechanically sound, even though often not derived from first principles) *usual* experimental corrections of apparatus and detectors, then it would be highly surprising that quantum mechanics being wrong, it would have in it a kind of self-correcting mechanism where its ideal predictions are wrong, but its experimental corrections are just as wrong in the opposite sense such as to result in agreement between realistic QM predictions of the experiment (including corrections) and actual experimental results, and that is what is observed: agreement between realistic QM predictions and experimental results.

Hi Vanesh,

You are simply repeating Bell's arguments... In the same way, I could argue that the local realist theories which, when interpreted according to the experimentators data massages, violate the Bell inequalities are merely realistic adjustements to the perfect local realist setup which does not violate these inequalities as long as the setup allows for the separability assumption to be made (such as in long distance correlation experiments). So we are not missing the point as you claim. I do not dispute that QM is successful in calculating the spectrum of the H and He atom (you cannot predict above He) and explaining the Lamb shift. However, I am convinced that these successes have perfect (albeit more difficult and subtle) classical explanations. A good step in this direction is the theory of stochastic electrodynamics which reproduces a good bunch of these so called exclusive results from second quantization in a firstly quantized framework (such as the Casimir effect, and the H atom I believe). Barut and Dowling have done quite some work on this issue ...

My intention is not at all to dispose of standard QM, I am perfectly aware of the fact that it provides an effective way of calculating the statistics of experiments with microscopic objects when applied on the *correct* problems with considerable thought. However, I also gives the wrong answers in some cases and it does not provide any insight into the dynamics of a single particle. I want to obtain *insight* into the microworld which I believe obeys the same laws as the macroworld (that is GR and electromagnetism), therefore entanglement is a crucial issue and one should not take it lightly.

Concerning your remark about the extrapolation of succes, I can only say that people have been looking for over 50 years for a perpetuum mobile; I hope one is not going to look 100 years for entanglement. By the way, Newton theory was also correct for 300 years.

Concerning Hartree, you have to include a classical radiation field determined by the probability current of the particles. In that way, you obtain a QM where each particle has its own wave function and where interactions propagate via a classical maxwell field determined by the sum over all probability currents times the appropriate charges (I think Barut called this the self field approach to QED, it's non-linear of course).

But I appreciate your honesty.

Cheers,

Careful

 

[Careful]

Start your own thread if you want to push your personal agenda on Bell tests.

I have no agenda, I just want people to know the exact scientific status. I think we all agree that this is important

 

[Sherlock]

Hi sherlock, the point is that the interaction between A and B has measurable consequences out of the lightcone of A.

One way to understand how this can happen in a universe which obeys the principle of locality is that the motions of A and B subsequent to their interaction are related. Isn't it?

Concerning quantum computing: it is impossibe to tell whether it is CLASSICAL correlations one is using or not (see my latest post).

I don't know much about quantum computing. I thought it required strictly quantum correlations of the entangled sort.

 

[Careful]

Hi sherlock,

true: A and B are related also in a locally causal universe, but only in the future lightcone of A. In the example I gave you, the influence of A on B travels outside the future lightcone. Moreover, you should not see A and B as ``particles´´ but as observables of a second quantized field.

Cheers,

careful

 

[Sherlock]

Moreover, I am glad you state explicitely that you can use faster than light signalling in an operational sense.

Where did I state that? My current understanding is that evolutions of any sort in our universe are limited by the speed of light.

So standard QM is not local either in the Bell or operational sense.

My understanding is that QM is non-local in the Bell sense, but that this is an artifact of limitations (as are, I assume, essentially correctly specified in the principles of quantum theory) placed on any fundamental theory by our (I assume locality obeying) universe.

I'm not sure what you mean by non-local in an operational sense. But, afaik, the the formal transformations to principle axes were developed in line with the assumption that Nature obeys the principle of locality.

Anyway, as to the question that this thread poses, I think that QM is inherently non-local only in an artificial sense, because it's inherently incomplete in a physical sense -- and I think this inherent incompleteness applies to any theory of fundamental processes (ie., in a universe constrained by the principles of relativity theory and the principles of quantum theory, then no hidden variable theory is possible).

Moreover, I do not know precisely which conservation laws you are referring to.

The classical conservation laws which were taken over directly into quantum theory. Conservation of energy, momentum, angular momentum, etc.

 

[Sherlock]

... true: A and B are related also in a locally causal universe, but only in the future lightcone of A.

Why not in the future lightcone of B also? The precise relationship between the motions of A and B subsequent to their interaction remains until one or the other, or both, are subjected to external influences (such as B interacting with C). The motions of B and C subsequent to their interaction will be, in part, due to B's prior interaction with A.

Moreover, you should not see A and B as ``particles´´ but as observables of a second quantized field.

It's just a convenient way to talk about it. Individual detections are particles.

 

[Sherlock]

This is an actual physical example of one thing being in two places at once.
... I think of the total breakdown of the classical concept that an object cannot be in two places at once.

This isn't a breakdown of the classical concept. It's actually an analogy to our more direct experience of the world. The different parts of the chair I'm sitting on are in many places at once. The different parts of an expanding water wave front are in many places at once.

Your idea of quantum correlations as involving separate parts of the same physical entity are one way to qualitatively conceptualize what's happening. But, it might not be what is actually happening in all cases of quantum correlations.

Suppose that the paired measurements are actually caused by opposite-moving, self-contained, separate wave structures of some sort. In this case, the results, A and B, really aren't caused by different parts of the same physical entity. But, the motions of the wave structures that caused A and B, and hence A and B, can still be related due to a common source or interaction.

 

[ttn]

But there is a mystery about collapse that it would be desirable to know more about. You touch on it above. You say that the WF collapses upon measurement, and sure, this is standard. So when there are 2 entangled particles, which one cause the collapse - measurement of Alice or of Bob? Sure, the results are apparently the same regardless of which one "causes" the collapse. But again, that's the mystery. We have no specific rule that defines this. And again, that is what I am referring to when I say "I am confused about whether WF collapse is physical" etc.

These are all good questions. In OQM, the collapse happens "instantaneously". So whoever measures first (Alice or Bob) collapses the wave function. That's an unambiguous answer as far as it goes.

The problem is, simultaneity is supposed to be relative. So "whoever measures first" isn't so clear after all. Or rather: in order to give a precise meaning to the *dynamics* of OQM, you have to add some extra spacetime structure such as a notion of preferred/absolute simultaneity. This is just another way of seeing the non-locality that is a real part of OQM.

This is of course precisely why MWI people want to get rid of the collapse rule entirely and get along with *just* the unitary dynamics.

But I am a bit confused about your marble box example. I think what you are saying is: this example does not violate Bell Locality because adding the information about Bob's outcome does not actually change the probability of the outcome at Alice. Am I close? (Or maybe the example isn't that important, not sure.)

No, it is important. If you don't understand that example, you won't understand why a violation of Bell Locality means a genuinely problematic "action at a distance" rather than something mundane and physically uninteresting like just learning something you didn't know about some distant thing.

Anyway, what you say is just right: the marble-in-box example involves no violation of Bell Locality. Here we have a change in the conditional probability of an event (Alice finding the marble) when we do/don't conditionalize on Bob's distant outcome. But if that's all Bell Locality required, it would be violated in situations (like this marble example) that *obviously* do not involve any spooky faster-than-light causality. So my point was just to clarify that Bell Locality is *not* in fact violated in this situation. It takes something more than this to violate it. The probability of an event has to change when we conditionalize on some space-like separated info *even though we've already conditionalized on the complete state of the world in the past light cone of the event in question*. In the marble case, *that* probability (namely: the probability for Alice to find the marble conditional on it either definitely already being in the box, or definitely already not being in the box) is either 0 or 100%, and it doesn't change if you *also* specify whether Bob found the marble. That information -- whether Bob found it -- is *redundant* because we've already specified the exact state of things near Alice, so the probabillities don't change.

Now, suppose we aren't talking about finding a marble, but doing some other experiment and getting some particular outcome. And suppose the probability for that particular outcome is different, depending on whether we do or don't conditionalize on some other information pertaining to a spacelike separated region -- and this *even though we've already specified the complete state of things near the experiment in question*. Well then, wouldn't we say that that distant event (information about which we do or don't conditionalize on) is having some kind of effect on the outcome -- an effect which *cannot* be accounted for by *local causes* in the past light cone of the event? This is what a violation of Bell Locality means.

 

[Careful]

Hi sherlock,

You seem to say that (a) QM gives the right predictions and (b) spacetime is real and all processes (which are real by assumption) have speed smaller or equal than light. (b) logically implies that the wave function of (a) must be real which implies that processes exist which go faster than with the speed of light. This is a problem of the Klein Gordon equation for a complex scalar field in first quantization, is remedied by hand for two measurements in quantum field theory, but pops up again in the situation I mentioned to you. Therefore my statement.

Concerning the conservation laws you mention: it is very hard to obtain an anomaly free interacting QFT and these equations do usually not make much sense anymore...

 

[Careful]

Why not in the future lightcone of B also? The precise relationship between the motions of A and B subsequent to their interaction remains until one or the other, or both, are subjected to external influences (such as B interacting with C). The motions of B and C subsequent to their interaction will be, in part, due to B's prior interaction with A.
It's just a convenient way to talk about it. Individual detections are particles.

This should be only so for that part of B which interacted causally with A (which is de facto in the future lightcone of A), not the part that did not interact with A at all. The real problem is that reduction of the state (in B) is an instantaneous non - local (causal) process and that is why the influence of A through B will travel to C.

 

[vanesch]

I do not dispute that QM is successful in calculating the spectrum of the H and He atom (you cannot predict above He) and explaining the Lamb shift.

?? I think that QM has rather more successes on its name than just H and He ! I am even convinced that the "classical field" approaches (the coupled Dirac-Maxwell fields + eventually some noise) have serious problems with higher than He configurations. At best these theories give the same predictions as the Hartree-Fock method with the self-consistent potential, but it is well-known in quantum chemistry that this gives a good approximation, but sometimes not good enough and one needs to add things like "configuration interaction" to get closer to experimental values.

However, I am convinced that these successes have perfect (albeit more difficult and subtle) classical explanations. A good step in this direction is the theory of stochastic electrodynamics which reproduces a good bunch of these so called exclusive results from second quantization in a firstly quantized framework (such as the Casimir effect, and the H atom I believe). Barut and Dowling have done quite some work on this issue ...

Yes, that's fascinating work, I agree. The problem is, however, with most of these approaches, that they tackle ONE SPECIFIC aspect of quantum predictions, and that we can then vaguely hope that they will, one day, be as successfull as standard quantum machinery in all the rest.
It is just not reasonable to accept the quantum machinery for about all the predictions it makes, *except* for those very few predictions that kill off your original belief of how nature ought to be.
The reason why this work is 1) fascinating and 2) probably misguided can be found by "reductio ad absurdum". Indeed, if these classical theories were correct, their computations would be gazillion times simpler than quantum computations. Non-linear partial differential equations in 3 dimensions are, computationally, peanuts as compared to, say the Feynman path integral in QFT, and can be attacked much much easier with finite-element methods than QFT. It would reduce quantum chemistry, and even nuclear physics, to something computationally just as easy as weatherforcasting. Not that weatherforecasting is so simple, but it is doable, while QFT calculations only start to be tractable with lattice techiques. So if it were possible to do so, it would have been done already since a long time.

My intention is not at all to dispose of standard QM, I am perfectly aware of the fact that it provides an effective way of calculating the statistics of experiments with microscopic objects when applied on the *correct* problems with considerable thought. However, I also gives the wrong answers in some cases and it does not provide any insight into the dynamics of a single particle.

I would like to know in what specific cases quantum theory comes up with the wrong experimental predictions which have been falsified by experiment.

I want to obtain *insight* into the microworld which I believe obeys the same laws as the macroworld (that is GR and electromagnetism), therefore entanglement is a crucial issue and one should not take it lightly.

In a way, I *also* adhere to a belief: it is that there are a few fundamental principles on which the entire formalism of physical theory has to be constructed.

Concerning your remark about the extrapolation of succes, I can only say that people have been looking for over 50 years for a perpetuum mobile; I hope one is not going to look 100 years for entanglement. By the way, Newton theory was also correct for 300 years.

If you want my guess, I don't think quantum mechanics in its present form will still be around (except as a useful approximation) 300 years from now - or it will, but then because of lack of progress (for instance, lack of experimental input on quantum gravity phenomena). But as of now, it is still the best thing we have - and it has to be admitted that it is vastly more successful in vastly different fields than anything that tries to rival with it. At best you get *identical* predictions in certain areas. Entanglement is a very standard part of the quantum formalism, and *is* confirmed by many experiments in the sense that these calculations DO correspond to predictions that are verified: see further.

Concerning Hartree, you have to include a classical radiation field determined by the probability current of the particles. In that way, you obtain a QM where each particle has its own wave function and where interactions propagate via a classical maxwell field determined by the sum over all probability currents times the appropriate charges (I think Barut called this the self field approach to QED, it's non-linear of course).

I expected that this is what you meant but wasn't sure. It is indeed the self-consistent field method used in quantum chemistry. A good approximation, but with known deviations from experiment, which is improved upon by configuration interaction techniques which are nothing else but "entanglements" of the different electrons. Even in the H2 molecule, this is experimentally visible (although small). More successes can be found with the H20 molecule, especially the angle between the two bonds, which for the H-F selfconsistent field method gives us 106.1 degrees, while the CI technique gives 104.9 degrees (experiment being 104.5 degrees). Took this from "Modern quantum chemistry" by Szabo and Ostlund.
There are many many examples like this. The problem with the CI technique is of course the huge system of equations that it generates - hence my proof by contradiction of a classical theory doing the same thing: if it worked, it would be done since long.

 

[Careful]

Hi Vanesh,

It is late for me, so I shall treat some part of your comments and shall be back tomorrow for more...
I think your most substantial argument is that the Hartree approximation although it is good is known to deviate slightly from experimental outcome. This is a fact. However, I did not say that Hartree is the full theory, neither did I claim that Fock corrections is what one should be looking for:
(a) your computability argument is incorrect, good computer experiments concerning, say the classical three body problem, have only very recently been performed and obtaining a thorough understanding of it is still on the way. The same comment applies to GR where the post Newtonian approximation is often known not to be adequate and obtaining the full solution (to the nonlinear equations) is a notoriously difficult problem (and picking out the right finite element method can take a considerable amount of time, even for the trained mathematician).
(b) I did not say I accept all predictions of QM except those which kill my belief : I accept all predictions which are confirmed by experiment but this does not imply that QM is the only way to get to these results.
(c) I am fully aware that these aternative approaches are in some sense behind QFT. The reason for this is is easy to think of: just compare the amount of money which is put into both research branches.
(d) I know it is the attitude of most researchers to conclude from Hartree is not equal to, but very close to, experiment implies that entanglement vindicates again. However, here I disagree : could we simply not have forgotten something? Why are these predictions so close while we totally ignore all non-local correlations? One could think now of adding other interaction terms between different wave packages, as you mention, in order fit accurately to the results but this is patchwork. I will come back to this tomorrow.

 

[Sherlock]

You seem to say that (a) QM gives the right predictions ...

It seems to be the most accurate across a wide range of experimental applications. Saying that QM's predictions are "right" is sort of an iffy statement, isn't it? After all, there are limits (due to error and due to fundamental constraints specified by QM itself) to what can be experimentally determined. QM predicts values that experimental runs will approach, and, afaik (I'm just learning), it's calculations agree with experiment.

... and (b) spacetime is real

Space and time are conventions.

... and all processes (which are real by assumption) have speed smaller or equal than light.

Yes, I assume that Nature obeys the principle of locality.

(b) logically implies that the wave function of (a) must be real which implies that processes exist which go faster than with the speed of light.

Space time is a convention. So is the wave function. The wave function is a complete description of what is known about the quantum system it refers to (at least it's one way of describing what is known). However, the wave function is necessarily an incomplete description of the physical reality of the quantum system it refers to. Hence, no superluminality is implied. The assumption that Nature is local is based on strong theoretical arguments which have thus far not been falsified by experiment, so it remains.

Concerning the conservation laws you mention: it is very hard to obtain an anomaly free interacting QFT and these equations do usually not make much sense anymore...

I haven't learned QFT yet, so if your main points depend on this theory, then I must excuse myself from the discussion.

 

[Sherlock]

This should be only so for that part of B which interacted causally with A (which is de facto in the future lightcone of A), not the part that did not interact with A at all.
The real problem is that reduction of the state (in B) is an instantaneous non - local (causal) process and that is why the influence of A through B will travel to C.

I don't understand what you're saying here.

 

[vanesch]

I will come back to this tomorrow.

Maybe we should then start another thread, not hijacking this one ? This thread is about the non-locality or not of quantum theory, under the working hypothesis that QM predictions are correct, and it is NOT about whether or not this working hypothesis is acceptable.

 

[Careful]

Maybe we should then start another thread, not hijacking this one ? This thread is about the non-locality or not of quantum theory, under the working hypothesis that QM predictions are correct, and it is NOT about whether or not this working hypothesis is acceptable.

Ok, you start a new one although - as I mentioned before - the answer to the literal message of this thread is rather obvious (it might be pleasant to chat about it however and I am aware that recently some controversy about this has been on the Arxiv, entirely unnecessary and even misguided from time to time). People talk too much about the interpretation of QM and are afraid to really modify it, that's one good reason why progress (especially in quantum gravity) is slow...(not substantial) So, I will check this site later on.

 

[vanesch]

Ok, you start a new one although - as I mentioned before - the answer to the literal message of this thread is rather obvious

I don't think that the answer is obvious. QM presents us with a riddle: Bell locality is violated, but signal locality isn't.
If neither were violated, I think nobody would even think of saying that locality is violated by QM. If both were violated, again, it would be obvious that QM violates locality. But we're in between.

And then it depends on how you look at the internal workings of the theory to decide whether the mathematical operations you execute (and which you "believe" to be associated to an ontology or not) are respecting locality or not. So it depends on what exactly you understand by locality, and what exactly you assign a reality to in the mathematical framework of QM. This involves of course the interpretation you attach to it.

Concerning the predictions:
So if you say: a theory is local if it respects Bell locality, then, no, QM is not local (that's ttn's point of view).
If you say: a theory is local if it respects signal locality, then yes, QM is local (can't build a FTL phone that way).
Concerning the mathematical formalism and its relation to an ontology:
If you 1) assign a reality to the wavefunction and 2) consider the projection postulate as describing something that physically happens, then the inner workings of QM are bluntly non-local.
Denying 1) is the epistemological viewpoint of QM (just a technique for calculating outcomes of experiments) and you're back to the "predictions" side.
Denying 2) (like does MWI) allows you to consider the mathematical machinery of QM as respecting locality.
See, plenty of stuff to argue endlessly over, and spend time on PF :-)

 

[Careful]

I don't understand what you're saying here.

Look, it is very simple: locality in the operational sense means that a measurement at A cannot have measurable influence outside the lightcone of A. The example I gave you violates this. However, it is of crucial importance here that the measurement at B is non-local: such as the projection operator on a localized state or a Wilson loop, but not the integral of a local operator valued density. Such non-local operators are used all the time in QFT, so one cannot claim they are not physical. Therefore, if one assumes the validity of perfect von Neumann measurements and the existence of non-local observables, then one has to conclude that QFT is not local operationally. One could argue that such measurements are impossible, but then one has to develop an accurate measurement theory which respects locality. Such task has not been accomplished yet: therefore my statement is fair.

 

[Careful]

I don't think that the answer is obvious. QM presents us with a riddle: Bell locality is violated, but signal locality isn't.
If neither were violated, I think nobody would even think of saying that locality is violated by QM. If both were violated, again, it would be obvious that QM violates locality. But we're in between.
And then it depends on how you look at the internal workings of the theory to decide whether the mathematical operations you execute (and which you "believe" to be associated to an ontology or not) are respecting locality or not. So it depends on what exactly you understand by locality, and what exactly you assign a reality to in the mathematical framework of QM. This involves of course the interpretation you attach to it.
Concerning the predictions:
So if you say: a theory is local if it respects Bell locality, then, no, QM is not local (that's ttn's point of view).
If you say: a theory is local if it respects signal locality, then yes, QM is local (can't build a FTL phone that way).
Concerning the mathematical formalism and its relation to an ontology:
If you 1) assign a reality to the wavefunction and 2) consider the projection postulate as describing something that physically happens, then the inner workings of QM are bluntly non-local.
Denying 1) is the epistemological viewpoint of QM (just a technique for calculating outcomes of experiments) and you're back to the "predictions" side.
Denying 2) (like does MWI) allows you to consider the mathematical machinery of QM as respecting locality.
See, plenty of stuff to argue endlessly over, and spend time on PF :-)

I agree with you here except that you cannot send signals faster than light in QM. This is a much more subtle issue than just postulating commutation relations (see my previous post) at spacelike separated events. So, you should not talk but develop a theory of non-perfect Von Neumann measurements which respects locality in the operational sense.

 

[vanesch]

I agree with you here except that you cannot send signals faster than light in QM. This is a much more subtle issue than just postulating commutation relations (see my previous post) at spacelike separated events.

I have seen what you allude to, but I can't make much sense of it. I would think that what is sufficient is that the Green's functions (the propagators) vanish outside of the lightcone ? (and this is related to the commutation relations vanishing at spacelike separated events) How are you going to modify the field in a spacelike way if the Green's function is 0 ?

 

[Careful]

I have seen what you allude to, but I can't make much sense of it. I would think that what is sufficient is that the Green's functions (the propagators) vanish outside of the lightcone ? (and this is related to the commutation relations vanishing at spacelike separated events) How are you going to modify the field in a spacelike way if the Green's function is 0 ?

It is very simple: quantum field theory has NO measurement theory. There is NO principle of reduction of the state functional (that is why the vanishing of the Green function outside the lightcone is sufficient for your purposes) such as in standard QM (you should read Sorkin's paper). This is clearly unsatisfactory and the only thing QFT is good for is to compute S matrices. Summary: in QFT you are restricting yourself to a unitary evolution. Bringing in any discrete/non-unitary reduction of the state principle allows for the possibility of measurable correlations outside the lightcone at least when you do it in the naive Von Neumann sense. So it seems to me you have two possibilties: either (a) you admit that the idea behind QFT needs improvement in order to incoorporate for a suitable measurement theory or (b) you refuse fundamental investigations in the principles of QFT and accept either (i) superluminal signalling or (ii) the fact that QFT allows for a limited number of questions to be asked.

 

[vanesch]

Summary: in QFT you are restricting yourself to a unitary evolution.

But that is maybe a very very good idea

 

[Careful]

But that is maybe a very very good idea

No, it is not
Let me make the full reasoning: let X be the orginal density matrix
(a) you want f(x) and f(y) to commute when x and y are spatially separated since you want MEASUREMENT of f(x) and f(y) to be independent (otherwise there is no sense in doing this)
(b) Let A be in the past of B and C in the future of B but not in the future of A (A,B and C are domains in spacetime). Suppose a and c correspond to local operators on A and C that is : a = integral(f(x), x in A) and c = integral(f(y), y in C). Then clearly a and c commute and if I do a after c or c after a, it does not matter. However the causal relations impose a temporality on the order in which a,b and c have to be performed : that is c after b after a. Now if b is not a local operator (for example the integral of quasi local observables) then
sum_{i,j,k} P_i Q_j R_k X R_k Q_j P_i is not equal to sum_{i,j} P_i Q_j X Q_j P_i even when both density matrices are restricted to the complement of the lightcone of A (P_i Q_j R_k are the orthogonal projection operators associated to c,b and a respectively). Non local operators are for example integrated hamiltonian densities with respect to some observers.

This is clearly a problem. So I expect a better answer from you. On one hand you claim that f(x) and f(y) have to commute since measurements have to be independent and on the other you claim that you do not want to do state reduction when it becomes troublesome. Even funnier, if you would claim that no measurement can be made in QFT, then it is impossible to even tell something about this issue at all :-) It is clear that any quantum theory MUST have a consistent measurement theory for it to be taken seriously. So either you propose one, or otherwise I see no reason why superluminal signalling is banned.

 

[Tez]

No, I think you have it right.
For me personally, the confusion begins when you talk about an object outside the lightcone. Alice makes a measurement, which causes collapse of the shared wave function. Now you know something about something somewhere else, true, and that is outside the lightcone.
But what has happened that is really so weird? We project the knowledge we have back to the point at which the entangled particle pair was created. This is the same thing that happens when only one particle is involved, nothing strange about that. The particle acts as if it had that orientation from the last point something happened.
Say Alice sees a V orientation with a polarizer at 0 degrees. Naturally, all subsequent measurements will be consistent in EVERY WAY with this knowledge AS IF it was always that way from the creation of the particle. So in that sense there is absolutely nothing happening outside any light cone.
In other words, all quantum measurements find a particle in an eigenstate and its eigenvalue is consistent with the quantum measurement rules. Entangled particles are no different in this respect. So the real question to me is: why does a measurement at time T2 cause the particle to assume a specific value as if it had that value at time T1 (where T1 is before T2) ? Does that make oQM non-local? Or is that a case of backwards causality? I am not sure that anything physical occurs along with the collapse, and I think that is a relevant question too.
Naturally, some of these issues show up in our definition of locality. You can see that there is no information transfer which is FTL, and there is no clear causal effect which is FTL. Yet the Bell Locality condition is violated with a strict application of its definition. So what does that condition actually tell us? Of course, it fits with the Bell Inequality too so that is very important.
Inquiring minds want to know...

What you say here is one of the most common misunderstandings of what Bell's theorem tells us. It is categorically not the interesting nonlocality of QM.

Ask yourself if that type of nonlocality would enable you to "win" this game:

The game is you and a friend are imprisoned, and told you're going to be separated. Once separated you will each randomly be asked either "what is X?" or "what is Y?" to which you must answer either 1 or -1. If you are both asked the X question then you must give opposite answers, but in all other cases (one of you asked X the other Y, or both asked Y) you must give the same answer. You win the game, you get released.

A minutes thought will let you know that unless you can tell what question the other person is asked there's no way to guarantee you winning the game. Your best bet is simply to agree to always answer the same thing and rely on the 3/4 chance that this'll win you the game.

But wait: if you carried entangled particles you can delay the decision of what to answer to your captors - once asked the question you make a measurement on the particle and output 1 or -1 according to the outcome. This way your probability of being released goes up to 85%. How did the entangled particles do it unless they knew something about what the other particle had been "asked".

In a 3 prisoner version the probability of release can go up to 100%, despite every "logical" strategy allowing for a maximum of 75%.

 

[DrChinese]

But wait: if you carried entangled particles you can delay the decision of what to answer to your captors - once asked the question you make a measurement on the particle and output 1 or -1 according to the outcome. This way your probability of being released goes up to 85%. How did the entangled particles do it unless they knew something about what the other particle had been "asked".

But they don't really! (Obviously, otherwise we could use that for FTL signalling.)

All we know is that our captors communicated in our past light cone and we are using that locally transmitted knowledge to play a logic game. There is no superluminal anything over and above a normal interpretation of a Bell test. After all, our captors can't release us until they compare our answers.

 

[Careful]

But they don't really! (Obviously, otherwise we could use that for FTL signalling.)
All we know is that our captors communicated in our past light cone and we are using that locally transmitted knowledge to play a logic game. There is no superluminal anything over and above a normal interpretation of a Bell test. After all, our captors can't release us until they compare our answers.

Well, enlighten us with a measurement theory which bans superluminal signalling consistently. I did not meet anyone until now who can do this, perhaps a texan chinese can be the first one.

 

[Sherlock]

You can see that there is no information transfer which is FTL, and there is no clear causal effect which is FTL. Yet the Bell Locality condition is violated with a strict application of its definition. So what does that condition actually tell us?

The Bell Locality condition tells us that A and B aren't observationally or statistically independent. This has a local explanation via quantum theory which also tells us that A and B aren't independent. (Paired results are the macroscopic manifestation of quantum-level disturbances that came from the same emitter via the same emission process. They're thus related by the applicable conservation laws, and, when they're entangled, they're entangled due to the ambiguity of certain intermediate states described by the emission process model.)

My current understanding of, and answer to, your original question is that quantum theory is not inherently non-local.

EDIT: I think that maybe the Bell Locality condition is poorly named. Calling it the Bell Independence condition would be less confusing. The assumption that statistical independence of A and B is required in a local universe is, I think, incorrect.

 

[Sherlock]

How did the entangled particles do it unless they knew something about what the other particle had been "asked".

The entangled particles don't need to know anything about what the other particle had been asked. Their motions just need to be related in some way.

Entanglement implies a relationship between the motions of two particles. Exactly how they're related is unknown. But the assumption of some sort of relationship has a purely local basis, and this assumption is conceptually adequate to understand the predictable results.

The fact that a detailed description of the motions (the sub-microscopic evolutions) of the two particles is impossible according to the principles of quantum theory is why the theory can't be made to be explicitly local. But it certainly isn't explicitly non-local either.

 

[Sherlock]

Well, enlighten us with a measurement theory which bans superluminal signalling consistently.

Does Special Relativity qualify?