A Realization of a Basic Wigner's Friend Type Experiment

[PeterDonis]

It is observationally impossible to distinguish between "real" collapse and the appearance of collapse.

Yes, so which position one takes on this issue depends on whether one thinks of QM as the theory of everything and "Newtonian" classical behavior as an approximation that one is free to interpret however one likes (as your arguments in this thread do), or whether one thinks of QM as a theory that has been shown to work well for microscopic objects but which we have no evidence for when macroscopic objects are concerned (nobody has ever observed quantum interference, erasure, etc. with rocks), so one should be skeptical of claims like "we only need the appearance of collapse" that only make sense on the assumption that exact unitary QM applies to everything.

 

[DarMM]

I really don't understand what the confusion is here. I'm making two points:
1) Decoherence is a feature of the wavefunction dynamics of QM, and is therefore independent of interpretation.
2) After sufficient decoherence, any observer described by QM will observe what looks like collapse.

Decoherence shows (with certain caveats) that it is consistent for observer A with device A measuring observer B with device B to consider the latter as having reduced their wavefunction. In other words it shows the consistency of the kinematic constraints on the statistics on one level with a dynamical account on a higher one.

However we still lack collapse for observer A. Somebody will always lack collapse.

 

[kimbyd]

But this is just rejecting the premise of the thought experiment, which is that Wigner has complete control over all degrees of freedom in the closed friend/lab system, so Wigner can reverse decoherence at his leisure. Saying this is not feasible in practice is besides the point, as we are trying to assess the logical consistency of the theory under maximally extreme circumstances. Your argument is similar to saying since nobody can live inside a black hole, it doesn't matter if GR breaks down at the singularity.

I don't think that's an accurate characterization. Here is the abstract:

The scientific method relies on facts, established through repeated measurements and agreed upon universally, independently of who observed them. In quantum mechanics, the objectivity of observations is not so clear, most dramatically exposed in Eugene Wigner's eponymous thought experiment where two observers can experience fundamentally different realities. While observer-independence has long remained inaccessible to empirical investigation, recent no-go-theorems construct an extended Wigner's friend scenario with four entangled observers that allows us to put it to the test. In a state-of-the-art 6-photon experiment, we here realize this extended Wigner's friend scenario, experimentally violating the associated Bell-type inequality by 5 standard deviations. This result lends considerable strength to interpretations of quantum theory already set in an observer-dependent framework and demands for revision of those which are not.

Emphasis mine. This isn't supposed to just be a paper about thought experiments, but something that could actually be tested in practice.

Yes, so which position one takes on this issue depends on whether one thinks of QM as the theory of everything and "Newtonian" classical behavior as an approximation that one is free to interpret however one likes (as your arguments in this thread do), or whether one thinks of QM as a theory that has been shown to work well for microscopic objects but which we have no evidence for when macroscopic objects are concerned (nobody has ever observed quantum interference, erasure, etc. with rocks), so one should be skeptical of claims like "we only need the appearance of collapse" that only make sense on the assumption that exact unitary QM applies to everything.

Why do you think that QM should differ substantially at the macro scale, when it predicts macro-scale behavior correctly?

I'm not assuming that exact unitary QM applies to everything. All that I'm saying is that the specific features that people try to tack on to wavefunction dynamics are both unnecessary and observationally impossible to verify. It is certainly conceivable for the universe to not be entirely unitary, but stating that it's non-unitary in a specific, unmeasurable way is making unwarranted assumptions.

 

[PeterDonis]

Why do you think that QM should differ substantially at the macro scale, when it predicts macro-scale behavior correctly?

You're framing the question backwards. I'm not starting from the theory and asking what should happen to it at the macro scale. I'm starting from the observation that we see all sorts of quantum phenomena at the micro scale, but not at the macro scale. It's not that "QM should differ", it's that the observed phenomena do differ.

One possible way to deal with this is to try to modify QM so that its predictions change as you go from the micro to the macro scale. But it's not the only possible way. Another obvious way is to look for a different theory that has QM as one approximation, at the micro scale, and the appropriate macro-scale theory as another approximation at the macro scale. This is one way of viewing the current search for a theory of quantum gravity.

I'm not assuming that exact unitary QM applies to everything.

Then whatever interpretation of QM you are using, it isn't the MWI, which requires exactly that assumption.

stating that it's non-unitary in a specific, unmeasurable way is making unwarranted assumptions.

Only if you are taking "exactly unitary at all scales" as the null hypothesis, the one we should accept in the absence of evidence to the contrary. But how did that particular hypothesis, which comes from a particular theory that only has experimental validation at the micro scale, somehow get itself to be the null hypothesis?

 

[kimbyd]

Decoherence shows (with certain caveats) that it is consistent for observer A with device A measuring observer B with device B to consider the latter as having reduced their wavefunction. In other words it shows the consistency of the kinematic constraints on the statistics on one level with a dynamical account on a higher one.

However we still lack collapse for observer A. Somebody will always lack collapse.

There's no actual collapse anywhere in MWI. There's only the appearance of collapse.

Decoherence causes interactions between different components of the wavefunction to be suppressed. Once those interactions are suppressed enough, an observer who is described by one branch of that wavefunction will not have any access to information in the other branches. That observer's experimental results look like only one outcome occurred.

That fact is the case whether you're observer A or observer B. The point of the Wigner's Friend thought experiment is to show that observer A will (ideally) still be able to measure some quantum effects for observer B, even if observer B doesn't see them.

The entire point I'm trying to get across is that decoherence places limits on the complexity of observer B in order for observer A to be able to measure anything. What I would expect to see in an experiment where the amount of decoherence in observer B is tunable is that there will be a regime where observer A sees the quantum effects, but as the decoherence is turned up (e.g. by increasing the intensity of a light source), those quantum effects should disappear.

Ideally, if this observation is to be useful at all, there will be a regime where observer B reports no visible quantum effects, but observer A still sees them.

 

[charters]

This isn't supposed to just be a paper about thought experiments, but something that could actually be tested in practice

But outside of the authors and their press releases, most do not really agree this recent optics experiment was a valid investigation of the Wigner's friend problem. The full fledged humanity of the internal observer is essential to avoiding loopholes that distract from the point of the thought experiment.

 

[kimbyd]

But outside of the authors and their press releases, most do not really agree this recent optics experiment was a valid investigation of the Wigner's friend problem. The full fledged humanity of the internal observer is essential to avoiding loopholes that distract from the point of the thought experiment.

But why should that be the case? All that you'd need to show is that the quantum effects disappear as decoherence is dialed up. If the observer isn't human, and the quantum effects disappear once the decoherence is turned up sufficiently, then that proves that you'll never get anything but a null result if the observer is human (assuming that we can't effectively isolate the human from the outside observer, which is probably always going to be the case).

 

[DarMM]

There's no actual collapse anywhere in MWI. There's only the appearance of collapse.

Of course hence @PeterDonis 's statement that this is fine in MWI, but it doesn't resolve it in other views like Copenhagen. Thus it's not interpretation neutral.

That observer's experimental results look like only one outcome occurred.

It shows that the higher observer's statistics for the lower observer, prior to measuring the lower observer, are consistent with the assumption that an outcome has occurred. This is not actual collapse though. The higher observer retains all terms, they have not reduced to one term. When they observe the lower observer they then reduce using an axiom separate to the unitary dynamics.

 

[charters]

But how did that particular hypothesis, which comes from a particular theory that only has experimental validation at the micro scale, somehow get itself to be the null hypothesis?

For me, Occam's razor and the non existence of an objective collapse theory that can reproduce the Standard Model and exactly explain the collapse triggers.

 

[charters]

But why should that be the case? All that you'd need to show is that the quantum effects disappear as decoherence is dialed up. If the observer isn't human, and the quantum effects disappear once the decoherence is turned up sufficiently, then that proves that you'll never get anything but a null result if the observer is human (assuming that we can't effectively isolate the human from the outside observer, which is probably always going to be the case).

Again, a proper a Wigner's friend scenario assumes the decoherence can always be reversed in principle, ignoring the practically insurmountable realistic challenges. The most philosophically interesting problem (for assessing the consistency of the theory) is when we imagine a decohered human being gets unitarily reversed.

 

[DarMM]

Again, a proper a Wigner's friend scenario assumes the decoherence can always be reversed in principle, ignoring the practically insurmountable realistic challenges. The most philosophically interesting problem (for assessing the consistency of the theory) is when we imagine a decohered human being gets unitarily reversed.

There are two levels of Wigner's friend constructions in the literature. One where the superobserver is able to perform arbitray measurements on the observer and a second where they also have full unitary control of the observer. Frauchiger-Renner and Brukner's theorem are in the first category, Masanes version of Frauchiger-Renner is in the second.

 

[kimbyd]

You're framing the question backwards. I'm not starting from the theory and asking what should happen to it at the macro scale. I'm starting from the observation that we see all sorts of quantum phenomena at the micro scale, but not at the macro scale. It's not that "QM should differ", it's that the observed phenomena do differ.

They differ only in terms of their ontological interpretation of what's occurring. The experimental predictions are absolutely identical once you're outside the quantum regime. So no, I don't reject at all that the macro scale and quantum scale are presenting fundamentally different behavior, because quantum dynamics correctly predicts what happens at all scales, without modification.

It is certainly possible that something we can't observe is occurring in the regime where quantum effects aren't measurable. But it would be stepping far outside the available evidence to try to stake a claim as to what that is.

One possible way to deal with this is to try to modify QM so that its predictions change as you go from the micro to the macro scale. But it's not the only possible way. Another obvious way is to look for a different theory that has QM as one approximation, at the micro scale, and the appropriate macro-scale theory as another approximation at the macro scale. This is one way of viewing the current search for a theory of quantum gravity.

I mean, you can modify QM. That's been the prevailing strategy in QM for a long time, and has been a primary component of most interpretations of QM.

But then it was pointed out that these modifications are unnecessary.

Then whatever interpretation of QM you are using, it isn't the MWI, which requires exactly that assumption.

Yes, MWI generally does assume unitarity. I'm not assuming MWI. I'm rejecting the assumption that additional dynamics need to be added based upon current observational evidence, though it remains plausible that the dynamics we know of today are incorrect in some manner we don't yet understand.

Rejecting the collapse assumption doesn't limit me to MWI, as there are still multiple mechanisms to go from the wavefunction dynamics to observational effects. Those other interpretations generally offer the same general picture of what's going on as MWI, but do so in different ways.

Only if you are taking "exactly unitary at all scales" as the null hypothesis, the one we should accept in the absence of evidence to the contrary. But how did that particular hypothesis, which comes from a particular theory that only has experimental validation at the micro scale, somehow get itself to be the null hypothesis?

Because it requires fewer assumptions, and predicts the same large-scale behavior.

It's very true that the realization that the assumption of wavefunction collapse was not a necessary component of the theory did not occur until decades after quantum theory first appeared, and that the interpretation of probability if you don't have collapse only had a relatively firm grounding quite recently. But the fact remains that a theory which assumes only evolution via $i\hbar{d \over dt}|\Psi(t)\rangle = H|\Psi(t_0)\rangle$ where $H$ is the appropriate Hamiltonian correctly predicts all behavior at both small and large scales, with the exception of gravity.

 

[charters]

There are two levels of Wigner's friend constructions in the literature. One where the superobserver is able to perform arbitray measurements on the observer and a second where they also have full unitary control of the observer. Frauchiger-Renner and Brukner's theorem are in the first category, Masanes version of Frauchiger-Renner is in the second.

In either case, the decoherence of the lab along the friend's measurement basis is getting reversed

 

[DarMM]

In either case, the decoherence of the lab along the friend's measurement basis is getting reversed

Sorry that wasn't to disagree, just as a point of interest.

 

[kimbyd]

Of course hence @PeterDonis 's statement that this is fine in MWI, but it doesn't resolve it in other views like Copenhagen. Thus it's not interpretation neutral.

This goes back to what I was saying earlier: it's technically possible for wavefunction collapse to occur before decoherence. And if said collapse occurs before decoherence it could, in principle, be measurable.

But if the collapse happens after decoherence, then it can't be measured and it should be disregarded. The most simplistic Copenhangen interpretation, which places the boundary of collapse far after the boundary where decoherence occurs, cannot ever possibly be tested as a result.

The Copenhagen interpretation is useful because it's simple. But that doesn't mean it's correct. And it should be expected that its usefulness will degrade whenever the precise details of wavefunction collapse (whether effective or real) are important for the behavior of a given system, as in quantum computing.

It shows that the higher observer's statistics for the lower observer, prior to measuring the lower observer, are consistent with the assumption that an outcome has occurred. This is not actual collapse though. The higher observer retains all terms, they have not reduced to one term. When they observe the lower observer they then reduce using an axiom separate to the unitary dynamics.

So what distinguishes observer A from observer B that permits observer B to avoid collapse where observer A collapses? What physical property is being used to separate them?

In simple Copenhagen, the answer is simple: both A and B are fully-collapsed, so observer A can never measure quantum effects from observer B's measurement.

In MWI (and similar), the answer is that neither observer A nor observer B fully collapse, but some quantum effects are still apparent to observer A. The statement that observer A has collapsed is not intended to be a real statement of what has occurred, but instead a tool to make evaluation of the experimental result easier to interpret. And it should be a reasonable approximation as long as observer A is sufficiently complex to never directly see any quantum effects (as is the case for humans not using very specialized experimental equipment).

I'm not aware of any interpretation that would allow observer B to avoid collapse while observer A collapses, except for the outside possibility that observer B's nature allows them to experience less decoherence.

 

[charters]

Sorry that wasn't to disagree, just as a point of interest.

No worries. By the way, did you ever happen to check out the new Bub paper on Wigner' friend I replied with here a few days/maybe a week ago?

 

[DarMM]

The Copenhagen interpretation is useful because it's simple. But that doesn't mean it's correct. And it should be expected that its usefulness will degrade whenever the precise details of wavefunction collapse (whether effective or real) are important for the behavior of a given system, as in quantum computing.

Copenhagen has collapse as being epistemic, thus there are no details in this sense either effective or real.

So what distinguishes observer A from observer B that permits observer B to avoid collapse where observer A collapses? What physical property is being used to separate them?

Observer B does not avoid collapse. When they apply the theory they must use collapse. When observer A models observer B with QM the statistics they have prior to the measurement of macroscopic observables on observer B are consistent with B having an outcome. However they must use collapse when they finally observe B.

 

[DarMM]

No worries. By the way, did you ever happen to check out the new Bub paper on Wigner' friend I replied with here a few days/maybe a week ago?

Yes I've read it. Any particular aspect you are interested in. To me there are three points of discussion in it.

The consistency of NeoCopenhagen views with Frauchiger-Renner
That MWI violates both the S and C conditions of Frauchiger-Renner
That MWI is empirically wrong

That's also the order of how contentious others may find them coincidentally.

 

[charters]

Yes I've read it. Any particular aspect you are interested in.

Yes, on page 10 he says:

"Since the two Wigners measure commuting observables on separate systems, W[bar] can communicate the outcome‘ok’of her measurement to W, and her prediction that she is certain, given the outcome‘ok,’ that W will obtain ‘fail,’ without ‘collapsing’ the global entangled state. Then in a round in which W obtains the outcome ‘ok’ for his measurement and so is certain that the outcome is ‘ok,’ he is also certain that the outcome of his measurement is not ‘ok.’"

I think this is just wrong. He is letting W-bar know both her own |ok> and F-bar's |tails> at the same time, which is just a complementarity violation. Same error when he just changes the subscript from F-bar to W-bar in the subscripts in (16) to (17). If you keep W-bar and F-bar distinct in the correct way, the whole argument against MWI/representational QT falls apart (and actually instead shows the flaw in his own informational view)

 

[DarMM]

(and actually instead shows the flaw in his own informational view)

I'll have to think about the rest of your post, but regarding this part I think at best it shows Bub's defense of informational views is not correct. Richard Healey has already shown that Informational views can escape Frauchiger-Renner by rejecting intervention insensitivity.

 

[PeterDonis]

quantum dynamics correctly predicts what happens at all scales, without modification.

Only if you include collapse--call it "effective collapse" or whatever you want to allow room for interpretations like the MWI--whenever a measurement occurs. But QM does not tell you when a measurement occurs. QM does not tell you when to include collapse. It just says "do it when it makes sense to do it". That's not a dynamical prediction. It's an admission that there is no dynamical prediction, so the physicist just has to make the best of it.

Decoherence does not help here because decoherence is a continuous process, not an instantaneous change. Decoherence does not tell you "when decoherence has reached point X, include a collapse". It still says "include collapse when it makes sense to include it".

 

[PeterDonis]

a theory which assumes only evolution via $i\hbar{d \over dt}|\Psi(t)\rangle = H|\Psi(t_0)\rangle$ where $H$ is the appropriate Hamiltonian correctly predicts all behavior at both small and large scales

I strongly disagree, for the reasons I gave in my last post.

 

[charters]

I'll have to think about the rest of your post, but regarding this part I think at best it shows Bub's defense of informational views is not correct. Richard Healey has already shown that Informational views can escape Frauchiger-Renner by rejecting intervention insensitivity.

Which Healey paper is this in?

 

[DarMM]

Which Healey paper is this in?

https://arxiv.org/abs/1807.00421
Section 3

 

[DarMM]

Decoherence does not help here because decoherence is a continuous process, not an instantaneous change. Decoherence does not tell you "when decoherence has reached point X, include a collapse". It still says "include collapse when it makes sense to include it".

Precisely, in my language above it causes no inconsistencies if observer A assumes observer B included collapse at any point after B's device has decohered in A's application of quantum theory.

After decoherence A's dynamical model of B is not in contradiction with any presumed application by B of collapse, but nothing says when to apply collapse.

 

[charters]

https://arxiv.org/abs/1807.00421
Section 3

Thanks. I think Healey is correct that F&R don't show a contradiction among the 2 superobservers. But really this was never the issue for informational interpretations, and the whole move of going to 4 players is an unhelpful detour. The issue for informational interpretations is present with just Wigner and Friend, and it is that the superobserver and internal observer disagree about whether the internal (human) observer is a quantum mechanical subsystem or an observer external to quantum theory.

A new paper that explains this very nicely on pg 6-7: http://philsci-archive.pitt.edu/16238/

 

[DarMM]

Thanks. I think Healey is correct that F&R don't show a contradiction among the 2 superobservers. But really this was never the issue for informational interpretations, and the whole move of going to 4 players is an unhelpful detour. The issue for informational interpretations is present with just Wigner and Friend, and it is that the superobserver and internal observer disagree about whether the internal (human) observer is a quantum mechanical subsystem or an observer external to quantum theory.

A new paper that explains this very nicely on pg 6-7: http://philsci-archive.pitt.edu/16238/

I've read her paper before and don't agree with the analysis. The response is covered in many papers and the anaylsis is very similar to one by Deutsch as improved by Brukner's theorem that is the originator of this thread.

Wigner is measuring very different classes of observables to the friend. Thus $|0\rangle_S$ does not allow an inference of $|0\rangle_L$ with $S$ the system the observer is looking at and $L$ their lab.

This inference doesn't even hold in Spekkens toy model or epistemically restricted classical theories, so performing it in QM seems pointless to me.

 

[charters]

This inference

What is your |0>L above? When Wigner simply opens the box to ask Friend his result of the S measurement?

 

[DarMM]

What is your |0>L above? When Wigner simply opens the box to ask Friend his result of the S measurement?

The state of the friend's lab.

 

[charters]

The state of the friend's lab.

I don't understand what the validity of this inference has to do with the issue of the parties disagreeing about whether Friend is a quantum system or an external observer.