I Is the fork in the Einstein Podolsky Rosen argument correct?

[sahashmi]

TL;DR

Is the EPR argument correct?

The Einstein podolsky rosen argument is detailed here: https://plato.stanford.edu/entries/qt-epr/

The argument creates a logical fork and says there are only two options. In the case of perfect correlations where you have two photons that either both pass or are both absorbed by the filter, Einstein and the rest argue that if the particles are NOT physically influencing each other (spooky action at a distance), there are local hidden variables

So, he argues that either

a) there are local hidden variables

or

b) the particles are physically influencing each other (spooky action)

Now, his argument for a) relies on this. In the case of perfect correlations, as soon as Alice observes that her photon passes through the filter, she can predict with **certainty** that Bob on the other end must also have had a photon pass.

If you can predict a measurement with a certainty of 1, and neither particle is influencing each other, they then argue that there must be an “element of reality” to the particle that results in that (i.e. a local hidden variable).

This is no different from saying that if I can predict every time that Bob’s coin toss landed on heads or will land on heads, it implies that something in the coin determined it to land on heads.

Here’s the interesting part of this fork. If this fork is correct, and if this argument is correct, then physicists have no option but to say that the particles are influencing each other since Bell’s theorem already ruled out the local hidden variable option. This would contradict a lot of modern physicist beliefs (unless you believe in many worlds). There is no third option.

So the physicists who say that one can explain this by merely “dropping realism” are ultimately arguing that if you can predict X with a probability of 1, it doesn’t imply that X was determined to be 1. But isn’t that the very definition of determinism? How else do we scientifically figure out if something is determined? Denial of this seems to be solipsistic in nature and seems to point to absurdities. With that being said, I’d love to be proven wrong if I’m missing something here.

So I must ask, is this argument correct? Why or why not?

Original paper: https://cds.cern.ch/record/405662/files/PhysRev.47.777.pdf

Note that my question is not about whether local hidden variables exist (we know they don’t), but is more about whether his logical fork of the two options (local hidden variables vs action/influences between particles) is correct.

EDIT: I’m not exactly sure whether Einstein concluded that there were local hidden variables or not (it seems from the below comments that he didn’t fully conclude that so I omitted a detail in my post that said Einstein’s conclusion that local hidden variables are underlying QM is incorrect)

 

[lodbrok]

That is not an accurate reading of the EPR paper. The paper deals with whether the QM description is complete, not whether local hidden variables exist.

 

[Nugatory]

So the physicists who say that one can explain this by merely “dropping realism” are ultimately arguing that if you can predict X with a probability of 1, it doesn’t imply that X was determined to be 1

It implies that the measurement result was determined to be 1. But that is not the same thing as “X was determined to be 1” - that stronger conclusion requires an additional assumption, counterfactual definiteness.

 

[lodbrok]

Regarding the existence of local hidden variables, EPR did not address this in their paper. Note the following quote in their conclusion:

While we have thus shown that the wave function does not provide a complete description of the physical reality, we left open the question of whether or not such a description exists.

 

[lodbrok]

TL;DR Summary: Is the EPR argument correct?

If you can predict a measurement with a certainty of 1, and neither particle is influencing each other, they then argue that there must be an “element of reality” to the particle that results in that (i.e. a local hidden variable).

This is no different from saying that if I can predict every time that Bob’s coin toss landed on heads or will land on heads, it implies that something in the coin determined it to land on heads.

Counter examples: Alice's detector is Off. I predict with certainty that she will get zero clicks. What element of reality in the particle could possibly have caused that?
Another, instead of tossing, Bob simply places the coin heads up every time. I can predict the outcome will be heads with certainty. What element of reality in the coin could have caused that result?

 

[sahashmi]

Counter examples: Alice's detector is Off. I predict with certainty that she will get zero clicks. What element of reality in the particle could possibly have caused that?
Another, instead of tossing, Bob simply places the coin heads up every time. I can predict the outcome will be heads with certainty. What element of reality in the coin could have caused that result?

The argument doesn’t need the element of reality to be in the particle the same way a local hidden variable in entanglement doesnt need to literally be inside the particle. The element of reality just has to be somewhere that causes the result to come out with a probability of 1. In your “counterexample”, the thing that causes the detector to have zero clicks is it being turned off.

Ditto for the other example. Bob’s placement determines the fact that the coin will be heads every time.

Determinism both ways. And if locality remains, then the particles must be locally predetermined to explain the perfect correlations. This was ruled out implying that locality cannot remain.

Where’s the fault in this logic?

 

[sahashmi]

Regarding the existence of local hidden variables, EPR did not address this in their paper. Note the following quote in their conclusion:

Well in that case, one can’t even say that Einstein was wrong at all! My question is more about whether his logical fork of the two options (local hidden variables vs action/influences between particles) is correct.

Edited my original post to mark this though

 

[sahashmi]

It implies that the measurement result was determined to be 1. But that is not the same thing as “X was determined to be 1” - that stronger conclusion requires an additional assumption, counterfactual definiteness.

If a measurement result is determined to be 1, talk of what exists before measurement is irrelevant to the argument. The point is that either the measurement result is locally predetermined or that there are faster than light influences. The first is ruled out by Bell’s theorem.

 

[Nugatory]

talk of what exists before measurement is irrelevant to the argument.

It is not irrelevant to the EPR argument, which is that something existed before the measurement and the result is telling us what it is. Indeed, we could paraphrase the EPR argument as “QM is incomplete because it doesn’t tell us what existed before measurement”.

 

[sahashmi]

It is not irrelevant to the EPR argument, which is that something existed before the measurement and the result is telling us what it is. Indeed, we could paraphrase the EPR argument as “QM is incomplete because it doesn’t tell us what existed before measurement”.

The question is whether or not giving up realism makes a difference to the argument that either

a) there are local hidden variables
b) the particles influence each other

Whether or not this is the exact same as the EPR argument or not, I don’t think your contention makes a difference to this fork. This is because regardless of what you assume exists before measurement, the fact remains that you can predict Bob’s measurement once Alice knows hers in the case of perfect correlations.

So unless Bob’s measurement outcome determines Alice’s measurement outcome (option b), option a must be true (i.e. Alice’s measurement outcome must be independently predetermined).

Experiments showed option a) is impossible. Option b) is remaining. What is the third option?

 

[Nugatory]

Edited my original post to mark this though

[Participant hat off, moderator hat on]
DON’T DO THAT TO A POST THAT HAS BEEN REPLIED TO!
(Or if you must, show explicitly what you changed). This thread is going be to be read by hundreds or thousands of people over the next few years, they are the real audience for this discussion. Rewriting the original post destroys the context for the replies.

 

[sahashmi]

[Participant hat off, moderator hat on]
DON’T DO THAT TO A POST THAT HAS BEEN REPLIED TO!
(Or if you must, show explicitly what you changed. This thread is going be to be read by hundreds or thousands of people over the next few years, they are the real audience for this discussion. Rewriting the original post destroys the context for the replies).

Got it! I’ll be explicit. I didn’t rewrite much though. I just removed what I said was the conclusion of the paper but I’ll make that explicit

 

[Fra]

I think this is a good question, and I thin Einstes was right to be deeply disturbed about this.

The question is whether or not giving up realism makes a difference to the argument that either

a) there are local hidden variables
b) the particles influence each other

It definitely does for me, beacuse "realism" someone imples a constraint on the nature and effect of the hidden variables.

Realism local hidden variables are precisely the type of simple "ignorance" type, this is what Bells theorem disproves.

So for me personally there there is only one palatable third option: It's a different type of non-realistic hidden variable, that does not need spooky action at a distance. For me this involved implies subjective bayesian probabilities or interacting parts; not objective bayesian. But this explanation is still not on the table we are still looking for this, but the question was is ther a third option and IMO it is.

Nothing is more pathological that action at distance. Non-realism is not pathological at all, it just means that our former understanding of nature was likely way off.

/Fredrik

 

[sahashmi]

I think this is a good question, and I thin Einstes was right to be deeply disturbed about this.

It definitely does for me, beacuse "realism" someone imples a constraint on the nature and effect of the hidden variables.

Realism local hidden variables are precisely the type of simple "ignorance" type, this is what Bells theorem disproves.

So for me personally there there is only one palatable third option: It's a different type of non-realistic hidden variable, that does not need spooky action at a distance. For me this involved implies subjective bayesian probabilities or interacting parts; not objective bayesian. But this explanation is still not on the table we are still looking for this, but the question was is ther a third option and IMO it is.

Nothing is more pathological that action at distance. Non-realism is not pathological at all, it just means that our former understanding of nature was likely way off.

/Fredrik

Why do you think non realism escapes the conclusion of action between the particles?

It remains to be true that in the case of perfect correlations, Alice knows Bob’s measurement as soon as Alice makes hers. If Bob’s measurement is not locally predetermined, and neither does Alice’s measurement determine Bob’s (which is what non realism here implies), why are their results the same?

Note (and the following is just my personal thoughts) that the results of all experiments don’t need to be explained by action at a distance necessarily. It could be superluminal action but of a finite speed that propagates through space. It would just have to be way faster than light although this flatly contradicts relativity

 

[PeterDonis]

Why do you think non realism escapes the conclusion of action between the particles?

Because "action between the particles" implies that something "real" is happening, and the whole point of non-realism is that that's not the case.

 

[PeterDonis]

just my personal thoughts

Personal speculation is off limits here. Particularly when it's the same sort of idea that has already been discussed in a previous thread of yours that got closed. Please be advised.

 

[sahashmi]

Because "action between the particles" implies that something "real" is happening, and the whole point of non-realism is that that's not the case.

Well the action could be after the measurement of one of the particles (which is what current nonlocal theories postulate anyways), and the measurement outcome we all agree is real.

 

[Fra]

Why do you think non realism escapes the conclusion of action between the particles?

Because the ansatz in bells theorem where you partition or divide the probabilities into the HV - makes no sense if the HV are subjective to the entangled subsystems only. Ie where HV are relational just to explain their correlated responses even before the relation to the ennviromnet(ie external spacetime) is set.

The ansatz in this case would be much more complex. And its not what bell considers.

/Fredrik

 

[sahashmi]

Personal speculation is off limits here. Particularly when it's the same sort of idea that has already been discussed in a previous thread of yours that got closed. Please be advised.

Sure sure, I’ll stick to just thoughts about this fork. Because if this fork is correct, how are physicists speculating a local non realistic option? There must be something wrong in the argument otherwise

 

[PeterDonis]

his logical fork of the two options (local hidden variables vs action/influences between particles)

Please specify (as in, with actual quotes) where in the paper this "logical fork" is given.

 

[PeterDonis]

Well the action could be after the measurement of one of the particles (which is what current nonlocal theories postulate anyways), and the measurement outcome we all agree is real.

So? That's still realism.

 

[sahashmi]

So? That's still realism.

Yes but that would be non local realism which is one measurement influencing another. You are saying that local non realism is an option. Einstein/Bell did not think this was one.

I’ll attach Bell’s thoughts on his last talk before he died:

“And similarly here the reasonable assumption, for any reasonable person, is that we are concerned here with the case of genetics. That these photon unknown to us carry some small packets of information, the same information in the two cases, which dictates that when they meet the same circumstances they will behave in the same way. And that was the Einstein-Podolsky-Rosen (EPR) argument, from no action at a distance to determinism. If you didn't want to accept that hypothesis of determinism, I think you are
obliged to accept that in some way things can agree at a distance, without any explanation. And that for Einstein would have been an objection about action at a distance.”

Full link: https://iis-edu.org/wp-content/uploads/2022/10/Bell-indeterminism-and-nonlocality.pdf

 

[PeterDonis]

that would be non local realism

You're confusing yourself. The question you asked, and I answered, had nothing to do with "non local realism". It has to do with non-realism. You asked how non-realism (not "non-local realism") escapes the conclusion of action between the particles. And I answered that it does so because "action between the particles" requires realism. I never said it requires local realism. So pointing out that there are versions of "action between the particles" where the realism is non-local, while true, does not address what I said at all.

You are saying that local non realism is an option.

I said no such thing. See above.

That said, you have given no argument for why "local non realism" is not an option. You have shown how "non-local realism" is possible (and indeed that's no news to anyone who knows of the Bohmian interpretation). But that in no way shows that "local non realism" is not possible. Again, you're confusing yourself.

 

[sahashmi]

You're confusing yourself. The question you asked, and I answered, had nothing to do with "non local realism". It has to do with non-realism. You asked how non-realism (not "non-local realism") escapes the conclusion of action between the particles. And I answered that it does so because "action between the particles" requires realism. I never said it requires local realism. So pointing out that there are versions of "action between the particles" where the realism is non-local, while true, does not address what I said at all.


I said no such thing. See above.

That said, you have given no argument for why "local non realism" is not an option. You have shown how "non-local realism" is possible (and indeed that's no news to anyone who knows of the Bohmian interpretation). But that in no way shows that "local non realism" is not possible. Again, you're confusing yourself.

I’ll try to be explicit again.

1. In the case of perfect correlations, when Alice makes a measurement, she can predict with certainty Bob’s measurement

2. This implies that either Alice’s measurement causes Bob’s measurement outcome (non local action) or that Bob’s measurement outcome was pre determined locally.

But if there is local “non realism” and indeterminism where Alice’s measurement outcome does not affect Bob’s, then they cannot be perfectly correlated. This is because if I can predict X with certainty, X must be determined (this is the Einsteinian argument). For if X was not determined, it makes no sense to be able to predict it.

Even if you believe that we can’t make definitive conclusions about the particle’s reality before measurement, it still is true that the measurement outcome is real and undeniable, and that outcome for Bob can be 100% predicted once Alice’s outcome is known.

So that’s my argument for why local non realism isn’t an option

 

[Nugatory]

"I think you are obliged to accept that in some way things can agree at a distance, without any explanation.And that for Einstein would have been an objection about action at a distance.”

For Einstein, yes. But Einstein also accepted (in historical context back in the 1930's may not even have recognized it as an assumption) counterfactual definiteness.

Do note that the non-local influence required to explain quantum correlations is something different from and weaker than "action at a distance" as the term is generally understand. There's no causality by the non-communication theorem, the distinction between cause and effect is frame-dependent so can have no physical significance, all observable results are the same no matter which measurement comes first. You should interpret that as a very strong hint that whatever is going on is not "action".

 

[sahashmi]

For Einstein, yes. But Einstein also accepted (in historical context back in the 1930's may not even have recognized it as an assumption) counterfactual definiteness.

Do note that the non-local influence required to explain quantum correlations is something different from and weaker than "action at a distance" as the term is generally understand. There's no causality by the non-communication theorem, the distinction between cause and effect is frame-dependent so can have no physical significance, all observable results are the same no matter which measurement comes first. You should interpret that as a very strong hint that whatever is going on is not "action".

The no communication theorem says that we can’t communicate using entanglement so far which is an anthropomorphic notion. It doesn’t say there’s no causality. Otherwise, theorems such as Bohmian mechanics which do posit “action” would be ruled out.

 

[PeterDonis]

1. In the case of perfect correlations, when Alice makes a measurement, she can predict with certainty Bob’s measurement

True. Note, though, that in QM interpretations like the MWI, unpacking how it's true is not straightforward. You'll see in a moment why I mention this.

2. This implies that either Alice’s measurement causes Bob’s measurement outcome (non local action) or that Bob’s measurement outcome was pre determined

Or that measurements don't have single outcomes--as in the MWI. In the MWI, there is no "non local action", but also no single measurement outcome is predetermined. All outcomes happen, and the correlations between them (whether they're perfect or not) are enforced by the wave function. The wave function is arguably nonlocal, but calling what it does "non local action" does not mean at all the same thing as what "non local action" means in, say, the Bohmian interpretation.

In short, the possibilities here are not as simple as you seem to think.

locally.

Invalid assumption--"predetermined" does not have to mean "predetermined locally". It's true that it does in the kind of model EPR implicitly had in mind when they wrote their paper. But that's not the only possible model.

if there is local “non realism” and indeterminism where Alice’s measurement outcome does not affect Bob’s, then they cannot be perfectly correlated. This is because if I can predict X with certainty, X must be determined (this is the Einsteinian argument). For if X was not determined, it makes no sense to be able to predict it.

Wrong. The MWI itself is a counterexample. See above.

 

[PeterDonis]

The no communication theorem says that we can’t communicate using entanglement so far which is an anthropomorphic notion.

Not in the sense in which "communication" is used in the theorem. Or the term "signaling"--since the "no signaling theorem" is a common alternate name for it. "Communication" and "signaling" in the theorem are perfectly objective notions that do not even require humans at all to happen.

It doesn’t say there’s no causality.

If two events are spacelike separated, their time ordering is frame dependent. On the usual understanding of "causality", where the "cause" must precede the "effect", this means neither of the two events can possibly be the cause, or the effect, of the other, since their time ordering is not invariant.

Otherwise, theorems such as Bohmian mechanics which do posit “action” would be ruled out.

Whether Bohmian mechanics is even compatible with relativity is an open question. There are other PF threads on this topic.

 

[Nugatory]

The no communication theorem says that we can’t communicate using entanglement so far which is an anthropomorphic notion. It doesn’t say there’s no causality. Otherwise, theorems such as Bohmian mechanics which do posit “action” would be ruled out.

The "communication" in the no-communication theorem isn't anthropomorphic, it is any form of information transfer. The theorem is significant precisely because it shows that correlated measurements do not and cannot violate causality.

Most presentations of Bohmian mechanics consider only the non-relativistic form of the theory so the question of compatibility with relativity never comes up (or maybe I should say that incompatibility with relativity has already been conceded). I won't speak to the relativistic formulations of Bohmian mechanics because we have people here who are far more qualified than I to do so, and because the topic is not altogether free of controversy.

 

[sahashmi]

True. Note, though, that in QM interpretations like the MWI, unpacking how it's true is not straightforward. You'll see in a moment why I mention this.


Or that measurements don't have single outcomes--as in the MWI. In the MWI, there is no "non local action", but also no single measurement outcome is predetermined. All outcomes happen, and the correlations between them (whether they're perfect or not) are enforced by the wave function. The wave function is arguably nonlocal, but calling what it does "non local action" does not mean at all the same thing as what "non local action" means in, say, the Bohmian interpretation.

In short, the possibilities here are not as simple as you seem to think.


Invalid assumption--"predetermined" does not have to mean "predetermined locally". It's true that it does in the kind of model EPR implicitly had in mind when they wrote their paper. But that's not the only possible model.


Wrong. The MWI itself is a counterexample. See above.

In regards to MWI, that’s why I wrote “unless you believe in many worlds” in my post

You said that predetermined does not mean locally predetermined. But I said it’s locally predetermined if there are no non local influences. If it’s predetermined, and there’s locality, it must be locally predetermined