I What Is the Frauchiger-Renner Theorem?

[DarMM]

I don't see how Eq. (31) comes from that assumption alone. I would like to see an explicit derivation. I will check the Fine's paper and comment it later.

If there are objective facts about the outcomes, then there is a distribution of those outcomes $p(a,b,c,d)$, the $E(i,j)$ then are marginals of this distribution and thus obey the inequality.

As above yes this would require them to commute, either this or the obeying of the CHSH inequality contradicts their QM behaviour, i.e. noncommuting and breaking the CHSH inequality.

 

[DarMM]

Leifer presenting Masanes's version:

@39:44

@41:55 the slide with $p(a,b,c,d)$ distribution is shown. It simply exists because the outcomes exist as Leifer says.

 

[Demystifier]

Just to be clear, this would mean many people don't understand the content of Fine's theorem as Healey's exact presentation of the argument has been used by Matt Leifer and Matthew Pusey. That's not to say you are wrong, just telling you the "scope" of the error if you are right.

Well, I am not sure about Pusey, but Leifer is one of those who think that the FR theorem is essentially correct and relevant. On the other hand, Healey thinks that FR theorem is not correct because it uses an unjustified assumption of non-contextuality, which he calls "intervention insensitivity". All I'm saying is that Eq. (31) also contains the assumption of non-contextuality (aka intervention insensitivity) so the Third Argument is wrong for essentially the same reason as the FR theorem.

 

[Demystifier]

If there are objective facts about the outcomes, then there is a distribution of those outcomes $p(a,b,c,d)$, the $E(i,j)$ then are marginals of this distribution and thus obey the inequality.

Ah, now I think I better understand the error in the Third Argument. The error is in the assumption that Eq. (31) is the same as Eq. (32). But it is not. The quantity $corr(a,b)$ in (31) is not the same as $E(a,b)$ in (32). The quantity $corr(a,b)$ in (31) is indeed a marginal of $p(a,b,c,d)$, but $E(a,b)$ in (32) is something else. The probability distribution $p(a,b,c,d)$ is obtained under a very specific measurement procedure that includes undoing previous measurements, while $E(a,b)$ describes a correlation obtained under a different, more ordinary, measurement procedure that does not include undoing previous measurements. The assumption that $E(a,b)$ is the same as $corr(a,b)$ corresponds to the assumption that it is not important how exactly something is measured, which is the same as the assumption of intervention insensitivity. Hence Eq. (32) would follow from (31) if intervention insensitivity was true. But intervention insensitivity is not true, so (32) does not follow from (31). Eqs. (31) and (33) are both correct, but there is no contradiction because Eq. (32) is not correct.

 

[DarMM]

Well, I am not sure about Pusey, but Leifer is one of those who think that the FR theorem is essentially correct and relevant.

Not that I've seen, in his lectures and talks he uses Masanes's version, not the actual FR result, although he still calls it "Frauchiger-Renner (Masanes version)". See the video I posted or this set of slides:
http://mattleifer.info/wordpress/wp-content/uploads/2009/04/FQXi20160818.pdf
(Slide 12)

Pusey can be seen using Masanes version here:

@18:38 he starts talking about it, @19:17 he mentions how it is not the actual FR result, but Masanes's version.

 

[DarMM]

The quantity $corr(a,b)$ in (31) is not the same as $E(a,b)$ in (32). The quantity $corr(a,b)$ in (31) is indeed a marginal of $p(a,b,c,d)$, but $E(a,b)$ in (32) is something else. The probability distribution $p(a,b,c,d)$ is obtained under a very specific measurement procedure that includes undoing previous measurements, while $E(a,b)$ describes a correlation obtained under a different, more ordinary, measurement procedure that does not include undoing previous measurements.

I don't see how this could be true, although I might be missing something. $E(a,b)$ and $corr(a,b)$ would occur before the reversal and seem identical.

So $corr(a,b)$ is the correlation between Alice and Bob's result. Let's displace the reversals to be 1,000 years after their measurments (for both of them), $corr(a,b)$ is then just the correlation between Alice and Bob's measurements on an entangled pair. I don't see how it could differ from $E(a,b)$, it can't involve a different quantum state or measurement operators as far as I can see and the set up is the same, except that a reversal awaits them in the far future, but I don't know of anything in the quantum formalism that means you should use a different state or measurement operator in such a case. When the measurement occurs it seems to me everything is the same.

If I'm wrong, which changes, the state or the operators, both? If the state how is it that the entangled particle pair themselves are affected?

 

[Demystifier]

I don't see how this could be true, although I might be missing something. $E(a,b)$ and $corr(a,b)$ would occur before the reversal and seem identical.

So $corr(a,b)$ is the correlation between Alice and Bob's result. Let's displace the reversals to be 1,000 years after their measurments (for both of them), $corr(a,b)$ is then just the correlation between Alice and Bob's measurements on an entangled pair. I don't see how it could differ from $E(a,b)$, it can't involve a different quantum state or measurement operators as far as I can see and the set up is the same, except that a reversal awaits them in the far future, but I don't know of anything in the quantum formalism that means you should use a different state or measurement operator in such a case. When the measurement occurs it seems to me everything is the same.

If I'm wrong, which changes, the state or the operators, both? If the state how is it that the entangled particle pair themselves are affected?

I was slightly sloppy by writing $a,b$ when I meant $i,j$. Let me explain all this once again, more carefully.

Let me first make a purely logical analysis, without much physical insight. Basically, there are 3 potentially reasonable but mutually exclusive possibilities:
a) $corr(i,j)=E(i,j)$ for all $i,j$, (31) and (32) are wrong, while (33) is true.
b) $corr(i,j)=E(i,j)$ for all $i,j$, (31) and (32) are true, while (33) is wrong.
c) $corr(i,j)\neq E(i,j)$ for some $i,j$, (31) and (33) are true, while (32) is wrong.
Which of those 3 possibilities do you think is correct?

Now physically, my chain of reasoning is the following:
- Eq. (33) is standard QM, which I think is consistent. After all, (33) can be thought of as a violation of the CHSH inequality, which has been confirmed experimentally. Hence I exclude b).
- Eq. (31) is standard probability theory stemming from the existence of joint probability $p(a,b,c,d)$, so it must be right. Hence I exclude a).
- What remains is c), which requires to explain why exactly $corr(i,j)\neq E(i,j)$ for some $i,j$. My answer is that they are different because they correspond to different measurement procedures. $corr(i,j)$ corresponds to a thought experiment in which the joint probability $p(a,b,c,d)$ exists. $E(i,j)$ corresponds to an actual experiment in which (33) is true, in which case Fine (and others) proved that the joint probability does not exist. If $corr(i,j)$ was the same as $E(i,j)$ for all $i,j$, then it would mean that the joint probability both exists and doesn't exist, which would be a logical contradiction. Hence it must be that $corr(i,j)\neq E(i,j)$ for some $i,j$.

Concerning your last question, almost any measurement changes the state of the system. That's called quantum contextuality. The exact mechanism of this change is a matter of interpretation. In practice it can usually be described by "collapse", but in the case of a thought experiment that includes undoing previous measurements, the collapse postulate must be replaced by something else. One possibility is the Bohmian interpretation, which clearly distinguishes the change of the wave function from change of the particles themselves. Both change in the Bohmian interpretation, but since it is formulated in the Schrodinger picture, the observable operators do not change. For more conceptual details about the Bohmian interpretation see my "Bohmian mechanics for instrumentalists" linked in my signature below.

One additional comment. Experiment that corresponds to Eq. (33) can be described by a collapse, while experiment that corresponds to Eq. (31) cannot be described by a collapse. That's another evidence that those are two very different experiments, which is why $corr(i,j)$ and $E(i,j)$ are physically different for some $i,j$.

 

[DarMM]

I quote you thrice, but the main point of this comment is the response to the last quote. The rest might make more sense in light of it.

Let me first make a purely logical analysis, without much physical insight. Basically, there are 3 potentially reasonable but mutually exclusive possibilities:
a) $corr(i,j)=E(i,j)$ for all $i,j$, (31) and (32) are wrong, while (33) is true.
b) $corr(i,j)=E(i,j)$ for all $i,j$, (31) and (32) are true, while (33) is wrong.
c) $corr(i,j)\neq E(i,j)$ for some $i,j$, (31) and (33) are true, while (32) is wrong.
Which of those 3 possibilities do you think is correct?

It's not so much that I think any of them are right or wrong. It's that in this set up you can derive all three under the given assumptions, which is impossible as they can't be all right and thus the contradiction. So you have to forgo an assumption. Which one depends on your interpretation.

In a given interpretation you might take various ones to be wrong, e.g. QBism has (31) and (32) being wrong, Retrocausal theories would have $corr(i,j) \neq E(i,j)$ in some cases, i.e. at least one marginal does not match quantum predictions. What I have heard is that Bohmians would also take $corr(i,j) \neq E(i,j)$ (mentioned in the talk by Pusey)

- What remains is c), which requires to explain why exactly $corr(i,j)\neq E(i,j)$ for some $i,j$. My answer is that they are different because they correspond to different measurement procedures. $corr(i,j)$ corresponds to a thought experiment in which the joint probability $p(a,b,c,d)$ exists. $E(i,j)$ corresponds to an actual experiment in which (33) is true, in which case Fine (and others) proved that the joint probability does not exist. If $corr(i,j)$ was the same as $E(i,j)$ for all $i,j$, then it would mean that the joint probability both exists and doesn't exist, which would be a logical contradiction. Hence it must be that $corr(i,j)\neq E(i,j)$ for some $i,j$.

Note the bolded part as I believe that's exactly it. Hidden variable theories like Bohmian Mechanics and Retrocausal theories have exactly $corr(i,j)\neq E(i,j)$. As you say otherwise you'd have $p(a,b,c,d)$ both existing and not existing. However that is the point of the theorem. The given assumptions require it to both exist and not exist, thus something is wrong with them. Hopefully this will make more sense after the following...

Concerning your last question, almost any measurement changes the state of the system. That's called quantum contextuality. The exact mechanism of this change is a matter of interpretation. In practice it can usually be described by "collapse", but in the case of a thought experiment that includes undoing previous measurements, the collapse postulate must be replaced by something else. One possibility is the Bohmian interpretation, which clearly distinguishes the change of the wave function from change of the particles themselves. Both change in the Bohmian interpretation, but since it is formulated in the Schrodinger picture, the observable operators do not change. For more conceptual details about the Bohmian interpretation see my "Bohmian mechanics for instrumentalists" linked in my signature below.

One additional comment. Experiment that corresponds to Eq. (33) can be described by a collapse, while experiment that corresponds to Eq. (31) cannot be described by a collapse. That's another evidence that those are two very different experiments, which is why $corr(i,j)$ and $E(i,j)$ are physically different for some $i,j$

I'm aware of these things, but I think what you are missing is the assumptions. I completely agree that they are different experiments and thus it is not unreasonable to ascribe $corr(i,j)\neq E(i,j)$. However what you are doing is secretly filling in the explanatory logic of an interpretation like Bohmian mechanics. For interpretations obeying the assumptions of the theorem there is no reason for $corr(i,j)\neq E(i,j)$.

This isn't a contradiction of all interpretations, but ones obeying all four of Masanes assumptions. In interpretations obeying those assumptions there is no reason that $corr(i,j) \neq E(i,j)$, in fact for them it is necessary that $corr(i,j) = E(i,j)$ and thus as you said perfectly in those interpretations "the joint probability both exists and doesn't exist". Which means they're self contradictory.

Interpretations like Bohmian Mechanics can naturally have $corr(i,j) \neq E(i,j)$ and thus the theorem has no relevance to them. Of course Masanes's proof can't show something nonsensical like $p(a,b,c,d)$ exists and doesn't exist at the same time. You're reasoning in physically logical terms that it must be that $corr(i,j)\neq E(i,j)$, but this isn't a proof about what is actually physically true, it's a proof about how some interpretations are ultimately self-contradictory.

Let's look at a type of nterpretation that falls victim to the no-go result, they have:

1. Observations are facts, i.e. if you were labelling events in spacetime you could assign something like $\sigma_z = \frac{\hbar}{2}$ to a spin measurement in a spacetime region. You don't need a pair like $(\sigma_z = \frac{\hbar}{2}, Measuring Entity)$. So the particle "had +-spin" as a simple universal fact, not "it manifested +-spin to Agent X".
   
   This means $p(a,b,c,d)$ exists
2. Quantum Theory is universal with no restrictions. This means I can describe Observers with quantum states and superpositions and they can be unitarily reversed like anything else.
   
   This permits the reversals of the measurements.
3. No hidden variables that permit differences from QM.
   So I cannot fall back on the contextual mechanics of some underlying theory to permit $corr(i,j) \neq E(i,j)$. Of course it is this assumption that is rejected by Bohmian Mechanics. Healey and Masanes don't really say this, it's just built into "QM is always right".
4. $\psi$ is not real, just ones knowledge/beliefs/etc. So collapse has no physical content, which prevents the "out" you have in the final paragraph.

Your ways out require $\psi$-collapse being physically real or additional variables, these interpretations don't have either. (Note "No Collapse" in Healey's paper)

Basically you're finding fault with the proof because it results in a contradiction about $p(a,b,c,d)$ and saying some assumptions must be wrong. However as I said, yes indeed, that's the point of the proof, one of the premises must be wrong. However they are all premises of Classic Copenhagen.

Thus if you want to be a Copenhagenist, or other similar interpretations, you need to jettison $p(a,b,c,d)$'s existence. You need to be perspectival. As Leifer and Pusey say, Observations are not facts.

 

[David Byrden]

Are we still talking about the FR paper now? I'm losing track.

I don't know how many times their error needs to be explained before everybody "gets" it.
Here, let me try to put it in simple terms. This is essentially what FR wrote:

Agent F says "I just got a qubit from Agent /F and measured it as UP. I will assume she's in a classical state. It must be TAILS".

And this is a corrected version:

Agent F says "I just got a qubit from Agent /F and measured it as UP. Since she's isolated from me, I will allow her to be in a superposition. Therefore she could be either TAILS or OK"

There. That's the whole thing solved, explained, finished and done with.

 

[DarMM]

Are we still talking about the FR paper now? I'm losing track.

We're talking about Masanes version, which works a bit differently.

Are we still talking about the FR paper now? I'm losing track.

I don't know how many times their error needs to be explained before everybody "gets" it.
Here, let me try to put it in simple terms. This is essentially what FR wrote:

Agent F says "I just got a qubit from Agent /F and measured it as UP. I will assume she's in a classical state. It must be TAILS".

And this is a corrected version:

Agent F says "I just got a qubit from Agent /F and measured it as UP. Since she's isolated from me, I will allow her to be in a superposition. Therefore she could be either TAILS or OK"

There. That's the whole thing solved, explained, finished and done with.

Maybe it's not because we don't "get it", perhaps your approach is wrong. It's been explained to you why this is wrong. If $F$ models $\bar{F}$ in superposition then they use the state:
$$\sqrt{\frac{1}{3}}\left(|\downarrow,h\rangle + |\downarrow,t\rangle + |\uparrow,t\rangle\right)$$

Thus if they measure $|\uparrow\rangle$ they can conclude $\bar{F}$ saw $|t\rangle$. That's just standard QM. To refute this you've said something about "knowledge in the environment" that doesn't seem part of QM. If the above was a typical particle state there's no way you'd conclude the second particle was in a superposition, so I genuinely don't understand your objection.

Also Masanes version doesn't use any of this, so even if this objection were valid it says nothing about Masanes version covered in Healey's paper.

 

[Demystifier]

Basically you're finding fault with the proof because it results in a contradiction about p(a,b,c,d) and saying some assumptions must be wrong. However as I said, yes indeed, that's the point of the proof, one of the premises must be wrong. However they are all premises of Classic Copenhagen.

I agree with everything you said in the great and very clear post above, except that I am not convinced that the bold sentence quoted above is true. What exactly is "Classic Copenhagen"? Is there a standard QM textbook in which it is unambiguously clear they all really are the premises? I ask this because it still looks to me as if the theorem (either FR or the Masanes version of it) rules out an "interpretation" that nobody seriously believed in the first place.

 

[Demystifier]

Also Masanes version doesn't use any of this, so even if this objection were valid it says nothing about Masanes version covered in Healey's paper.

In addition, the Masanes version is technically much simpler, so it is much easier to see the forest for the trees.

 

[DarMM]

I agree with everything you said in the great and very clear post above, except that I am not convinced that the bold sentence quoted above is true. What exactly is "Classic Copenhagen"? Is there a standard QM textbook in which it is unambiguously clear they all really are the premises? I ask this because it still looks to me as if the theorem (either FR or the Masanes version of it) rules out an "interpretation" that nobody seriously believed in the first place.

By Classic Copenhagen I mean roughly what Bohr, Heisenberg, etc thought. They're not consistent with each other or even themselves at different times, so in truth there is no "Copenhagen interpretation", but I mean roughly an interpretation following the list below which most of them would have held. Wasn't sure what to call, especially since there is no single Copenhagen interpretation, should have been clearer.

Before I go digging into Bohr's papers and those of the old Copenhagen group, I would just like to narrow the search for what you want. Which do you think Bohr for example didn't hold:

1. Observations are objective events, not agent experiences
2. There are no hidden variables, i.e. QM is complete
3. QM may be universally applied to any system (note this is not the same as saying there is no Heisenberg cut)
4. The wavefunction is not an ontic object

From my reading of him he said all four. I just want to know which one you think he didn't say, or what form of Copenhagen do you have in mind that denies one of these four statements and which one? Haag for example says something I believe in Local Quantum Physics, though I have to check that when I'm at my books.

To clarify does it need to be a textbook that says it, most textbooks are very operational and say nothing about the reality of $\psi$ for example. Would Omnés book "The Interpretation of Quantum Theory" count even though its not a "textbook" designed to teach QM?

Again most of the Foundations community would be wrong on this, they all seem to say it invalidates Classic Copenhagen, see Leifer's lectures I linked to before.

 

[DarMM]

In addition, the Masanes version is technically much simpler, so it is much easier to see the forest for the trees.

I 100% agree on this!

 

[atyy]

By Classic Copenhagen I mean roughly what Bohr, Heisenberg, etc thought. They're not consistent with each other or even themselves at different times, so in truth there is no "Copenhagen interpretation", but I mean roughly an interpretation following the list below which most of them would have held. Wasn't sure what to call, especially since there is no single Copenhagen interpretation, should have been clearer.

Before I go digging into Bohr's papers and those of the old Copenhagen group, I would just like to narrow the search for what you want. Which do you think Bohr for example didn't hold:

1. Observations are objective events, not agent experiences
2. There are no hidden variables, i.e. QM is complete
3. QM may be universally applied to any system (note this is not the same as saying there is no Heisenberg cut)
4. The wavefunction is not an ontic object

From my reading of him he said all four. I just want to know which one you think he didn't say, or what form of Copenhagen do you have in mind that denies one of these four statements?

To clarify does it need to be a textbook that says it, most textbooks are very operational and say nothing about the reality of $\psi$ for example. Would Omnés book "The Interpretation of Quantum Theory" count even though its not a "textbook" designed to teach QM?

Again most of the Foundations community would be wrong on this, they all seem to say it invalidates Classic Copenhagen, see Leifer's lectures I linked to before.

So what is wrong with Scott Aronson's objection?

 

[DarMM]

So what is wrong with Scott Aronson's objection?

He doesn't deal with Masanes's version.

 

[atyy]

He doesn't deal with Masanes's version.

OK, so basically the first 2 versions of the theorem are wrong?

For te Masanes version, I would count myself a believer in classic Copenhagen (ie, QM is formally complete), but also believe in hidden variables - I dare say many older physicists take this position - Dirac, Messiah, Bell, possibly even Landau & Lifshitz - because it has generally been assumed that Copenhagen has a measurement problem, the most important problem in foundations - not these horrible fake problems that Frauchiger and Renner solve.

 

[DarMM]

OK, so basically the first 2 versions of the theorem are wrong?

Not wrong, but the set of interpretations eliminated by them is almost null and I believe nobody held them outside of those who don't think much about Foundations. I don't think they have much content for serious interpretations.

For te Masanes version, I would count myself a believer in classical Copenhagen (ie, QM is formally complete), but also believe in hidden variables - I dare say many older physicists take this position - Dirac, Messiah, Bell, possibly even Landau & Lifshitz - because it has generally been assumed that Copenhagen has a measurement problem, the most important problem in foundations - not these horrible fake problems that Frauchiger and Renner solve.

What "fake problems" are you talking about?

You start off talking about Masanes's version and then switch back to Frauchiger-Renner. Are the fake problems with Masanes or the original Frauchiger-Renner.

As I said, like Healey I believe that the original FR papers have little force, but like most of the foundations community I think Masanes's version is a serious result. If Masanes's version has "fake problems" I'd like to hear them.

 

[atyy]

Not wrong, but the set of interpretations eliminated by them is almost null and I believe nobody held them outside of those who don't think much about Foundations. I don't think they have much content for serious interpretations.

Well, those are equivalent FAPP (as stevendaryl said, one can be charitable). Anyway, I'm glad I haven't spent much time with the first 2 FR versions.

What "fake problems" are you talking about?

You start off talking about Masanes's version and then switch back to Frauchiger-Renner. Are the fake problems with Masanes or the original Frauchiger-Renner.

As I said, like Healey I believe that the original FR papers have little force, but like most of the foundations community I think Masanes's version is a serious result. If Masanes's version has "fake problems" I'd like to hear them.

Fake problems in the sense that they are, as you say, completely irrelevant. Given the bad track record of this theorem, I'm going to wait before investing time in the Masanes version. However, I would like to know how that accounts for the view that QM is only formally complete, but not de facto complete because of the measurement problem. I think many would count this a variety of classic Copenhagen - Dirac, Messiah, Bell, possibly L&L. Demystifier himself is, I think, a proponent of this version of classic Copenhagen: https://www.physicsforums.com/insights/stopped-worrying-learned-love-orthodox-quantum-mechanics/.

 

[DarMM]

Given the bad track record of this theorem, I'm going to wait before investing time in the Masanes version.

Just to be clear Masanes's version is almost as old as the FR result itself. The FR result had problems recognised almost immediately, Masanes has yet to have any pointed out.

I don't think it can be judged as poor based on the track recorded of a related theorem. They're not the same result.

Although I understand your weariness.

The rest of your post requires a longer answer which I'll only have time for over the weekend. To help though what do you mean by "formally complete". That it isn't self-contradictory, even though it leaves somethings unexplained like measurement?

Also do you take a perspectival view of Copenhagen, i.e. the measurement outcomes aren't fully objective but associated with the agent observing them, relational to some degree.

 

[atyy]

The rest of your post requires a longer answer which I'll only have time for over the weekend. To help though what do you mean by "formally complete". That it isn't self-contradictory, even though it leaves somethings unexplained like measurement?

Formally complete, in the sense that it is (or can be) consistent after one has defined one classical/quantum cut. However it is not complete from the larger point of that the observer has a special status in quantum mechanics. Here one views QM as a complete theory that is emergent from a more complete theory, just as Newtonian mechanics is a complete theory emergent from more complete theories like special relativity and quantum mechanics.

Also do you take a perspectival view of Copenhagen, i.e. the measurement outcomes aren't fully objective but associated with the agent observing them, relational to some degree.

Within QM, outcomes are fully objective to all agents on the classical side of the classical/quantum cut. I'm not sure what you mean by perspectival, but we have to make only one classical/quantum cut - and things on the other side of the cut cannot be granted the status of agent or observation outcome.

 

[DarMM]

To be clearer then, in a Wigner's friend type set up, do Wigner and the friend have the same cut or different ones?

 

[atyy]

To be clearer then, in a Wigner's friend type set up, do Wigner and the friend have the same cut or different ones?

Wigner is classical, and the friend is quantum.

 

[DarMM]

From Wigner's perspective I get that, but do you think the friend should model themselves as quantum as well?

 

[atyy]

From Wigner's perspective I get that, but do you think the friend should model themselves as quantum as well?

The friend is not an observer and cannot model himself.

Edit: The question of whether the cut can be consistently shifted is interesting. However, I don't think it is in formal QM, since making the cut itself already requires subjectivity. However, I do like Hay and Peres's https://arxiv.org/abs/quant-ph/9712044. I think the question more generally requires a more complete theory like Bohmian Mechanics to be correctly dealt with.

 

[DarMM]

Okay you would be saying the presence of Wigner's measuring device alters the context of the situation and means the friend could not be placed on the classical side of the cut, even by himself.

The highly unusual confined nature of the friend and the contextual effects of Wigner's device means the friend evolves nonclassically?

I think this is legitimate, as an issue I always have with these set ups is what has to actually be done to get a Wigner's friend scenario is very extreme.

I'll need to think a bit.

 

[atyy]

Okay you would be saying the presence of Wigner's measuring device alters the context of the situation and means the friend could not be placed on the classical side of the cut, even by himself.

The highly unusual confined nature of the friend and the contextual effects of Wigner's device means the friend evolves nonclassically?

I think this is legitimate, as an issue I always have with these set ups is what has to actually be done to get a Wigner's friend scenario is very extreme.

I'll need to think a bit.

Yes, the friend is just the same as a Schroedinger's cat or a qubit.

 

[Demystifier]

Before I go digging into Bohr's papers and those of the old Copenhagen group, I would just like to narrow the search for what you want. Which do you think Bohr for example didn't hold:

1. Observations are objective events, not agent experiences
2. There are no hidden variables, i.e. QM is complete
3. QM may be universally applied to any system (note this is not the same as saying there is no Heisenberg cut)
4. The wavefunction is not an ontic object

My problem with this list is that 2 and 4 contradict 1 in a rather trivial way. 2 implies that there is no ontology which is not $\psi$. 4 says that there is no ontology which is $\psi$. Hence 2 and 4 together say that there is no any ontology at all. Yet 1 says that there is some ontology. A contradiction!

 

[DarMM]

My problem with this list is that 2 and 4 contradict 1 in a rather trivial way. 2 implies that there is no ontology which is not $\psi$. 4 says that there is no ontology which is $\psi$. Hence 2 and 4 together say that there is no any ontology at all. Yet 1 says that there is some ontology. A contradiction!

Well I would say (2.) means that your ontology doesn't involve a world that admits an objective description with mathematical terms. It might be non-mathematical, or it might be mathematical but completely reltional. Bohr thought the former, Schrödinger thought the latter.

Regardless it seems to me pretty clear that many believed 1-4. I understand why you think they are daft, I'm not disagreeing with you on this, but they're there in Bohr, Heisenberg and others writings. As silly as one might find them, there was no proof they were self-contradictory until Masanes.

 

[DarMM]

Basically @Demystifier and @atyy , the "problem" is that both of you already hold non-self-contradictory positions* that Masanes's proof doesn't affect and consider the positions it does refute as obviously wrong. However there was no clear proof they were wrong until now and they are the Copenhagen of Bohr and others of early QM.

*e.g. @atyy you have a similar view to Bub, QM must be used from the perspective of a final ultimate classical user and rejects the idea that the Wigner's friend set up has the friend capable of being considered on the classical side, even by himself. I think it's a very clear resolution and you seem to consider it obvious, but note that Bub had to publish a paper about it, it's not a widely known position.

 

[DarMM]

Yes, the friend is just the same as a Schroedinger's cat or a qubit.

Okay sorry for all the questions! I think this is the last one, what about the situation makes this true for you explicitly? Basically do you think that Wigner completely confining his friend to a totally sealed lab and the physical presence of Wigner's equipment causes quantum effects to propagate up to the macroscopic scale for the friend invalidating him considering himself or his equipment as classical?

I can imagine this, Wigner's device would be pretty "extreme" equipment whatever it is.

 

[martinbn]

My problem with this list is that 2 and 4 contradict 1 in a rather trivial way. 2 implies that there is no ontology which is not ψ\psi. 4 says that there is no ontology which is ψ\psi. Hence 2 and 4 together say that there is no any ontology at all. Yet 1 says that there is some ontology. A contradiction!

I don't understand this. How does 2. imply that there is no ontology other than $\psi$? In fact, in my opinion, it is an extreme abuse of language to say that $\psi$ is ontology. What does it even mean?

 

[DarMM]

What does it even mean?

It's true in Many-Worlds and Bohmian mechanics, the state space of ontic objects $\Lambda$ has the form $\Lambda = \mathcal{A} \times \mathcal{H}$ with $\mathcal{H}$ the quantum Hilbert space. So the ontology has $\psi$ as an element.

 

[martinbn]

It's be true in Many-Worlds and Bohmian mechanics, the state space of ontic objects $\Lambda$ has the form $\Lambda = \mathcal{A} \times \mathcal{H}$ with $\mathcal{H}$ the quantum Hilbert space. So the ontology has $\psi$ as an element.

May be I don't know what the word means. But for me "ontology" is related to existence/being. So something like an element in a Hilbert space cannot have ontology. It makes no sense to say that it exists. What exists is the particles/fields, not $\psi$.

 

[Demystifier]

I don't understand this. How does 2. imply that there is no ontology other than $\psi$?

Let me try to answer this step by step. (2) says that "There are no hidden variables, i.e. QM is complete". It really means that in the mathematical description of the world, we don't need anything else except $\psi$. I guess you agree so far.

Now consider ontology. It's a philosophical term, but its meaning is not essential here. All what we explicitly need to prove what I want to prove is the following assumption:
Assumption 1: Ontology can be described mathematically and we need it to describe the world.
Then, from (2) and Assumption 1, it follows that there is no ontology which is not $\psi$. Is it clear now?

So if you want to avoid my conclusion, you must deny Assumption 1, i.e. you must hold that either ontology cannot be described mathematically, or that we don't need ontology to describe the world. The first option looks like mysticism to me, which is logically legitimate but not scientific in spirit. The second option implies that, in order to describe the world, we don't need to believe that the Moon exists when nobody observes it, which is also logically legitimate, but contradicts common sense.

 

[Demystifier]

May be I don't know what the word means. But for me "ontology" is related to existence/being. So something like an element in a Hilbert space cannot have ontology. It makes no sense to say that it exists.

I would basically agree with that.

What exists is the particles/fields, not $\psi$.

A priori, it is not obvious that particles or fields exist. What is obvious is that things such as the Moon exist. But the Moon must be made of something more elementary, which could perhaps be particles, fields, strings, or something else. In my "Bohmian mechanics for instrumentalists" I explain why particles, as objects with well defined positions at all time, are the most natural possibility.

 

[martinbn]

Let me try to answer this step by step. (2) says that "There are no hidden variables, i.e. QM is complete". It really means that in the mathematical description of the world, we don't need anything else except ψ\psi. I guess you agree so far.

For the purpose of this conversation, yes, I agree. Strictly speaking there may be more things needed, say equations, a choice of Hilbert space representation, specific operators, boundary conditions etc.

Now consider ontology. It's a philosophical term, but its meaning is not essential here. All what we explicitly need to prove what I want to prove is the following assumption:
Assumption 1: Ontology can be described mathematically and we need it to describe the world.
Then, from (2) and Assumption 1, it follows that there is no ontology which is not ψ\psi. Is it clear now?

No, not at all clear. As I said we might be using the words differently. Let me give you an example. Classical mechanics of several particles. The ontology of the theory is that there are several particles. That's what exists in the physical world. Their behavior may be described by a function (plus possibly other things as above), but it is meaningless and abuse of language to say that the ontology is the function. If your argument was correct, it would imply that in this example the only ontology is the function. Which is absurd.

It seems to me that by ontology you mean the minimum of mathematical apparatus that is needed to describe the world within a given theory.

 

[Demystifier]

Their behavior may be described by a function (plus possible other things as above), but it is meaningless and abuse of language to say that the ontology is the function. If your argument was correct, it would imply that in this example the only ontology is the function. Which is absurd.

Ah, I see what is your problem. Suppose that the ontology is the particle with a well defined position. I think you are fine with that. That position can be described by 3 numbers (x,y,z), but it doesn't mean that those 3 numbers are ontology. The ontology is the position itself, not our mathematical coordinatization of that position. Is it what you are saying?

Well, strictly speaking I think that you right, but that problem can be cured relatively easy. Consider a physical ontological object $\tilde{O}$ (in the case above it is a particle with a well defined position in physical space). Let its all mathematically describable properties be described by some mathematical object $O$ (in the case above it is the numbers (x,y,z)). When we say that $O$ is ontology, it is just an imprecise manner of speak, which really means that $\tilde{O}$ is ontology.

So when someone says that the wave function $\psi$ is ontology, it really means that there is an ontological object $\tilde{O}$ such that its all mathematically describable properties can be described by $O=\psi$.

Does it make more sense now?

EDIT: This abuse of language is similar to calling the numbers $(x,y,z)$ a vector, which really means that the object $x{\bf e}_x+y{\bf e}_y+z{\bf e}_z$ is a vector. Physicists usually do not have problems with calling $(x,y,z)$ a vector, which often annoys mathematicians.

 

[DarMM]

May be I don't know what the word means. But for me "ontology" is related to existence/being. So something like an element in a Hilbert space cannot have ontology. It makes no sense to say that it exists. What exists is the particles/fields, not $\psi$.

@Demystifier has already stated this, but basically in a $\psi$-ontic view $\psi$ directly describes the physical properties of the objects of the theory. In a $\psi$-epistemic view, $\psi$ instead describes an agent's knowledge of those properties.

 

[atyy]

Okay sorry for all the questions! I think this is the last one, what about the situation makes this true for you explicitly? Basically do you think that Wigner completely confining his friend to a totally sealed lab and the physical presence of Wigner's equipment causes quantum effects to propagate up to the macroscopic scale for the friend invalidating him considering himself or his equipment as classical?

I can imagine this, Wigner's device would be pretty "extreme" equipment whatever it is.

Well, the friend can model himself as classical, but then Wigner will also model the friend as classical. Basically, we only allow one observer - Wigner. Wigner can subjectively choose where to put the classical/quantum boundary (ie. whether it includes the friend or not). However, what Wigner regards as an objective measurement outcome differs depending on whether the friend is classical or quantum. If Wigner regards the friend as classical, then the quantum state used by Wigner will collapse when the friend makes the measurement (ie. the friend's measuring apparatus is also Wigner's measuring apparatus).

 

[atyy]

I don't think Healey's third argument is correct. If Alice knows that Dan has recorded a definite outcome, then Dan has collapsed the state. So Eq 23 is not correct. It should be a proper mixture, not a pure state.

 

[stevendaryl]

I don't understand this. How does 2. imply that there is no ontology other than $\psi$? In fact, in my opinion, it is an extreme abuse of language to say that $\psi$ is ontology. What does it even mean?

I think that the issue is whether the wave function is subjective--that is, its value reflects the knowledge of the observer--or objective--it has an actual value, even if the observer may not know what that value is.

 

[martinbn]

I think that the issue is whether the wave function is subjective--that is, its value reflects the knowledge of the observer--or objective--it has an actual value, even if the observer may not know what that value is.

But I still don't understand Demistifier's argument. I don't see how 2. and 4. contradict 1.

 

[stevendaryl]

But I still don't understand Demistifier's argument. I don't see how 2. and 4. contradict 1.

I'm not sure I understand that specific argument. But EPR-type correlations seem to me to make it difficult to understand how the wave function can be epistemological (if that's the antonym of ontological).

I've stated this simple argument several times before, but I don't know of a good answer to it.

Let's take the case where Alice and Bob are measuring spins of anti-correlated spin-1/2 particles. To make it definite, let's assume that they are both planning to measure spin along the z-axis. Let's pick an inertial coordinate system in which Alice and Bob are both at rest, and assume that Bob is farther from the source of twin particles than Alice, so he measures his particle's spin slightly later than Alice does (although the two measurements have a spacelike separation).

1. Immediately before Alice performs her measurement, she would rate the probabilities for Bob's results to be 50/50 spin-up or spin-down.
2. Immediately after Alice performs her measurement and gets the result "spin-up", she knows with 100% certainty that Bob will get the result "spin-down".
3. So the statement "Bob will get spin-down" goes from being uncertain with 50/50 probability to true.
4. So it seems that Bob's situation (as understood by Alice) makes a nearly-instantaneous change.
5. There are two different possible interpretations of this sudden change. They are (1) epistemological, or (2) physical. (Maybe there are more than two possibilities, but I don't know of others.) Under the epistemological interpretation, the fact that Bob will get spin-down was true BEFORE Alice performed her measurement, and Alice's measurement simply allowed her to know this. Under the physical interpretation, the fact that Bob will get spin-down wasn't true before Alice performed her measurement, but became true as a side-effect of her measurement.
6. The epistemological interpretation seems to be contradicted by Bell's proof.
7. The physical interpretation seems to violate causality (no effects can travel faster than light).

I don't see a satisfactory way out if you accept Bell's proof and you also accept FTL limitations for effects. There are more exotic ways out, and they are Many-Worlds (Bob's result doesn't have a definite value) and superdeterminism, but those are unsatisfactory for other reasons.

 

[1977ub]

... The physical interpretation seems to violate causality (no effects can travel faster than light).
I don't see a satisfactory way out if you accept Bell's proof and you also accept FTL limitations for effects. There are more exotic ways out, and they are Many-Worlds (Bob's result doesn't have a definite value) and superdeterminism, but those are unsatisfactory for other reasons.

Can you articulate or link your favorite reasons why SD is unsatisfactory? Thank you.

 

[RUTA]

I'm not sure I understand that specific argument. But EPR-type correlations seem to me to make it difficult to understand how the wave function can be epistemological (if that's the antonym of ontological).

I've stated this simple argument several times before, but I don't know of a good answer to it.

Let's take the case where Alice and Bob are measuring spins of anti-correlated spin-1/2 particles. To make it definite, let's assume that they are both planning to measure spin along the z-axis. Let's pick an inertial coordinate system in which Alice and Bob are both at rest, and assume that Bob is farther from the source of twin particles than Alice, so he measures his particle's spin slightly later than Alice does (although the two measurements have a spacelike separation).

1. Immediately before Alice performs her measurement, she would rate the probabilities for Bob's results to be 50/50 spin-up or spin-down.
2. Immediately after Alice performs her measurement and gets the result "spin-up", she knows with 100% certainty that Bob will get the result "spin-down".
3. So the statement "Bob will get spin-down" goes from being uncertain with 50/50 probability to true.
4. So it seems that Bob's situation (as understood by Alice) makes a nearly-instantaneous change.
5. There are two different possible interpretations of this sudden change. They are (1) epistemological, or (2) physical. (Maybe there are more than two possibilities, but I don't know of others.) Under the epistemological interpretation, the fact that Bob will get spin-down was true BEFORE Alice performed her measurement, and Alice's measurement simply allowed her to know this. Under the physical interpretation, the fact that Bob will get spin-down wasn't true before Alice performed her measurement, but became true as a side-effect of her measurement.
6. The epistemological interpretation seems to be contradicted by Bell's proof.
7. The physical interpretation seems to violate causality (no effects can travel faster than light).

I don't see a satisfactory way out if you accept Bell's proof and you also accept FTL limitations for effects. There are more exotic ways out, and they are Many-Worlds (Bob's result doesn't have a definite value) and superdeterminism, but those are unsatisfactory for other reasons.

I'll give my response again, for the readers who might not have seen it. QM is in the business of supplying the distributions of quantum exchanges in 4D (block universe). The change in Alice's knowledge is purely epistemic, since the pattern (QM distribution) is "already there" in the block universe. Unitary evolution simply represents our (necessary) ignorance about what lies in the future.

 

[stevendaryl]

Can you articulate or link your favorite reasons why SD is unsatisfactory? Thank you.

I'm sorry, what is SD?

 

[1977ub]

I'm sorry, what is SD?

I meant 'superdeterminism'

 

[DarMM]

The epistemological interpretation seems to be contradicted by Bell's proof.

I wouldn't say Bell's theorem says anything too strong about the epistemological view. In fact I would say it has the same implications for ontic and epistemic views. It tells you the underlying ontology can't be a non-superdeterministic mathematical* local causal** single world, regardless of whether $\psi$ is part of that ontology or not.

It's the PBR theorem that has stronger implication for epistemic theories, meaning they basically have to go the retro/acausal route or the antirealist (non-mathematical) route.

* I prefer mathematical to realist, as what's actual rejected is that the underlying reality has no mathematical description not that it's not "real"

** Causal meaning not retrocausal or acausal

 

[DarMM]

I don't think Healey's third argument is correct. If Alice knows that Dan has recorded a definite outcome, then Dan has collapsed the state. So Eq 23 is not correct. It should be a proper mixture, not a pure state.

You're assuming an interpretation with some sort of reasonably objective collapse, not that that's wrong, but it doesn't affect Masanes proof.

Under the kind of Copenhagen advocated by Bohr and others since QM is just a calculus of expectations (in modern terminology a Bayesian framework) from the perspective of Alice you would have:
$$|\uparrow\rangle|d-ready\rangle \rightarrow |\uparrow\rangle|d-up\rangle$$
and
$$|\downarrow\rangle|d-ready\rangle \rightarrow |\downarrow\rangle|d-down\rangle$$
you'd have to have:
$$\sqrt{\frac{1}{2}}\left(|\downarrow\rangle + |\uparrow\rangle\right) |d-ready\rangle \rightarrow \sqrt{\frac{1}{2}}\left(|\uparrow\rangle|d-up\rangle + |\downarrow\rangle|d-down\rangle\right)$$

You're saying it should be a mixed state instead, but that implies you should know for some systems`you have yet to observe you should decide whether to apply unitary evolution or collapse. Would you measure this via decoherence?