Carroll interviews Barandes on Indivisible Stochastic QM

[Morbert]

Is it possible to summarize how they fall out of the formalism?

Section 3.7 in https://arxiv.org/pdf/2302.10778 outlines how.

Is it a variant of decoherence?

It is very closely related: If a system has become correlated with its environment (equation 45), then marginalizing over the environment (equation 50), which is like tracing over the environment in the QM formalism, will reveal a division event (t' in equation 53).

Could one take a description of wave function evolution in the standard formalism (or some other standard description, like using a density matrix for the evolution of an open system), translate it into an equivalent in Barandes' formalism, and get a unique answer about when the division events occur? If so is there some equivalent method that would give the same answer without Barandes' formalism?

This is close to what is done in 3.7. The time evolution of the system and environment wavefunction terms (equation 48) is used to construct the time evolution of the corresponding standalone probabilities in the stochastic formalism (equation 50).

 

[JesseM]

Section 3.7 in https://arxiv.org/pdf/2302.10778 outlines how.It is very closely related: If a system has become correlated with its environment (equation 45), then marginalizing over the environment (equation 50), which is like tracing over the environment in the QM formalism, will reveal a division event (t' in equation 53).This is close to what is done in 3.7. The time evolution of the system and environment wavefunction terms (equation 48) is used to construct the time evolution of the corresponding standalone probabilities in the stochastic formalism (equation 50).

My understanding is that the main reasons physicists don't see decoherence as fully resolving the measurement problem are that (1) it depends on the choice of how to make the division between "system" and "environment", which still has to be put in by hand rather than determined in some dynamical way, and (2) while decoherence shows how off-diagonal elements in the reduced density matrix for the "system" can rapidly become tiny due to interactions with "environment", getting you arbitrarily close to a classical mixture, they wouldn't exactly go to zero except maybe in the limit of infinite time. Is the system/environment boundary similarly not set dynamically in Barandes' interpretation, and is there also an issue where the timing of division events is somewhat approximate? (or if the timing is exact, does it depend on some other detail that has to be put in by hand, like a threshold for when quantum coherence has been sufficiently reduced?)

 

[Fra]

I do not think Barandes solves the measurement problem, its a refomulration.

In Barandes framework, the construction of the relevant gamma and the decomposition of the total system into subsystems, is not explained from first principles. Instead of saying that the subsystem becomes entangled with a huge environment in an enlarged Hilbert space, Barandes describes the subsystem in terms of a configuration and a stochastic transition constraint.

What the enlarged decoherence picture explains in terms of environmental entanglement is in barandes language, encoded in the effective gamma. So we get a new handle on the problem, but not yet a complete solution but the stochastic-quantum correspondence allows a translation between the two pictures.

Conceptually instead of thinking of the subsystem as entangled with a huge environment, one can think of it as a stochastic process whose transitions are constrained by gamma. For a subsystem, gamma can perhaps be read as encoding the effective influence of the environment or wider context on that process.

This gives a nice new perspective on the nature of dynamical law as the law is not a trajectory-generating flow but a transition constraint. But the deeper question remains how the relevant gamma is selected/evolved/stabilized?

Is that a "better question" than asking why this enlarged hilbert space and why these hamiltoninans? That may be a matter or opinon. I personally think its a more preferred question.

/Fredrik

 

[Morbert]

My understanding is that the main reasons physicists don't see decoherence as fully resolving the measurement problem are that (1) it depends on the choice of how to make the division between "system" and "environment", which still has to be put in by hand rather than determined in some dynamical way

This amounts to a choice of subsystem to model.

and (2) while decoherence shows how off-diagonal elements in the reduced density matrix for the "system" can rapidly become tiny due to interactions with "environment", getting you arbitrarily close to a classical mixture, they wouldn't exactly go to zero except maybe in the limit of infinite time. Is the system/environment boundary similarly not set dynamically in Barandes' interpretation, and is there also an issue where the timing of division events is somewhat approximate? (or if the timing is exact, does it depend on some other detail that has to be put in by hand, like a threshold for when quantum coherence has been sufficiently reduced?)

Like in textbook QM, you can decide what subsystem to model, but it isn't an issue* in Barandes's formalism. If you choose to model a subsystem $\mathcal{S}$ as distinct from its environment $\mathcal{E}$, you will end up with potentially highly divisible $\Gamma^\mathcal{S}$. If you choose to model everything as a single system, you will end up with a highly indivisible $\Gamma^\mathcal{SE}$.

The formally approximate nature of decoherence corresponds to a formally approximate nature of division events: $\Gamma(t\leftarrow 0) \approx \tilde{\Gamma}(t\leftarrow t') \Gamma(t'\leftarrow 0)$. But like decoherence, this divisibility is exact enough that no mortal can carry out an experiment that renders classical records unreliable.

*I also don't think it's an issue for textbook QM but that's for another thread.

 

[JesseM]

This amounts to a choice of subsystem to model.Like in textbook QM, you can decide what subsystem to model, but it isn't an issue* in Barandes's formalism. If you choose to model a subsystem $\mathcal{S}$ as distinct from its environment $\mathcal{E}$, you will end up with potentially highly divisible $\Gamma^\mathcal{S}$. If you choose to model everything as a single system, you will end up with a highly indivisible $\Gamma^\mathcal{SE}$.

Is there any notion of an objective truth about what a maximally exhaustive list of division events in a given region would look like, such that modeling choices that leave some out are akin to macrostates in statistical mechanics that leave out some of the full microstate information? Or are the facts about division events fundamentally dependent on these modeling choices?

For example, say we are considering a Wigner's friend type experiment where the friend is inside the box making measurements on various quantum systems and recording results in memories of some kind (whether in their brain or in some external records). Until the box is opened, everything inside is isolated from interaction with the external world and so there is no decoherence with the outside. If we model a bunch of separate subsystems within the box we could get the conclusion that there are division events and an ordinary history of what happened inside which match up to the memories we observe when we open the box, correct? But if we modeled everything inside the box as a single subsystem, then there need not be any division events inside until the box is finally opened? If so would there be any sense that the former description is more true/accurate than the latter?

Also in the latter description, is it possible that the non-Markovian realizers could involve non-classical events at the macro level, like Wigner's friend's memories of the past shifting from one moment to another?

 

[jeffn1]

Here is a summary of Art Hobson's view in Wigner's friend (hope okay to post--feel fee to ignore):
Physicists like Art Hobson view the Wigner’s friend paradox and the idea of "superposed memories" as a misunderstanding of how quantum mechanics actually works.
Hobson, a staunch proponent of standard Quantum Field Theory (QFT) and quantum realism, argues that the paradox disappears when you correctly analyze the nature of quantum entanglement and decoherence.
From Hobson's perspective, Wigner's friend's memories are never in a paradoxical superposition. His critique relies on three core arguments:

1. Entanglement Does Not Create Macroscopic Superpositions
A classic mistake in interpreting Wigner's friend is assuming that because the friend and the particle are entangled, the friend's macroscopic brain or data register is "smeared" between two realities.

• The Hobson View: Entanglement transforms a superposition of separate individual states into a superposition of correlations.
• In his published analysis on the measurement problem, Hobson notes that when the friend measures the particle, the mathematical state doesn't mean the friend has two memories simultaneously. It means if the particle is up, the memory is up; if the particle is down, the memory is down. The macroscopic memory register itself is never fundamentally undecided.

2. A Single, Definite Outcome Occurs Instantly
Many interpretations (like Copenhagen) struggle to explain exactly when or where a wavefunction collapses.

• The Hobson View: Collapse occurs automatically and nonlocally the moment the interaction happens.
• Because of the nonlocal nature of quantum mechanics, when the friend makes a measurement, exactly one definite physical outcome occurs. The alternative possibilities instantly vanish nonlocally. Therefore, the friend has exactly one true, irreversible memory from the very beginning.

3. Wigner is Simply Ignorant, Not a "Superobserver"
In the traditional paradox, Wigner's mathematical description treats the lab as a giant quantum superposition.

• The Hobson View: Wigner’s external wavefunction does not dictate objective reality; it merely represents Wigner's lack of information.
• Because Wigner is outside the closed lab, he is not entangled with the system yet. His equations describe his own subjective uncertainty, not the actual, physical state of the friend's brain. Wigner cannot perform a magic "super-measurement" to rewrite or erase the friend's memory because macroscopic decoherence has already locked that definite outcome into the physical universe irreversibly.

Summary of Hobson's Verdict
To Art Hobson, the idea that Wigner's friend has "undecided or superposed memories" is a pseudoscientific over-interpretation. The universe is made of fundamental fields, not localized particles or magical observers. Once an interaction is recorded macroscopically, the measurement is over, the memory is real, and Wigner's outside math is just a tool tracking his own ignorance until he opens the door.

 

[JesseM]

Summary of Hobson's Verdict
To Art Hobson, the idea that Wigner's friend has "undecided or superposed memories" is a pseudoscientific over-interpretation. The universe is made of fundamental fields, not localized particles or magical observers. Once an interaction is recorded macroscopically, the measurement is over, the memory is real, and Wigner's outside math is just a tool tracking his own ignorance until he opens the door.

I think the difficulty with assuming definite truths about recorded outcomes that happen in the box has to do with scenarios where the records of those outcomes are thoroughly erased prior to the box being opened by external observers.

One could even imagine a variant of the delayed choice quantum eraser seen in the simplified schematic diagram here, where if the second "idler" photon is detected at D3 or D4 then you know which slit the "signal" photon exited from, and you won't see a double-slit interference pattern in the subset of signal photons for which this is true. But if the idler is detected at D1 or D2 then the which-path info has been erased, and if you look at the subset of signal photos whose idlers were detected at one of those, say D1, you do see a double-slit interference pattern (though the interference pattern for D1 is out of phase with the interference pattern for D2, so if you look at all the signal photons detected at either D1 or D2, the two interference patterns cancel each other out--Sean Carroll discusses this cancellation in his post on the experiment here).

Now imagine a version of the Wigner's friend experiment where electrons can exit from one of a pair of slits in the box, and Wigner's friend has detectors inside that can measure which slit the electron passed through and record the results. But prior to opening the box, this information can be erased, say by a bomb exploding inside which vaporizes everything so when the box is opened, external observers like Wigner just find a gas at maximum entropy. And say the experiment was performed a vast number of times with identical starting state (perhaps easier to imagine if the whole thing is a simulation on a big quantum computer) so that we had multiple trials where the particles of the gas were in some specific position eigenstate P_n at the moment the box was opened and measured, a state that just looks like a gas at maximum entropy where all the records of past measurements inside the box have been thoroughly erased.

Then if we looked at the position of the electrons escaping the slits on that subset of trials with final box state P_n, would this be analogous to looking at the position of the signal photons on the subset of trials where the idler went to D1 in the delayed choice quantum eraser, so that the exiting electrons in this subset would show a two-slit interference pattern? If so, then for any interpretation that denies there is a single definite truth about which slit the signal photon went through in the trials where the idler was detected at D1, I'd think that interpretation should also deny a definite truth about which slit the electron went through in the Wigner's friend variant that was found in position eigenstate P_n when the box was opened, despite the fact that Wigner's friend inside the box was measuring every electron as it exited.

 

[Morbert]

@JesseM Division events are nomolgical. They are features of dynamical laws. Supernatural entities like the Wigner superobserver don't change what actually occurred when Wigner's friend made their measurement. But they do change the dynamics of Wigner's friend.

The dynamical laws of Wigner's friend, the transition matrix, will have a division event at Wigner's friend's measurement only until Wigner makes his measurement. Extending the transition matrix beyond Wigner's measurement means the matrix will not have a division event at Wigner's measurement.

 

[jeffn1]

Does it matter that the notion of retrocausality initially attributed to the "delayed choice" experiment has largely been debunked:

It looks like the explanation of this fits well within Art Hobson's model.

 

[JesseM]

Does it matter that the notion of retrocausality initially attributed to the "delayed choice" experiment has largely been debunked:

It looks like the explanation of this fits well within Art Hobson's model.

I wasn't suggesting any retrocausality, if you look at Sean Carroll's discussion of the delayed choice quantum eraser which I linked, he says that he says there is no need for it in the many-world interpretation ('There’s no need to invoke retrocausality to explain the delayed-choice experiment. To an Everettian, the result makes perfect sense without anything traveling backwards in time'), and there'd also be no need in Bohmian mechanics. The point I made in the last paragraph is that if an interpretation denies a single unique truth about which slit the photons came from in the standard DCQE experiment where information is erased, then it should also probably deny it in a more complex Wigner's friend style experiment where the friend plans to measure which slit each electron goes through but all records are erased before the box is opened. Bohmian mechanics would say there is a definite truth about the path in the standard DCQE, while MWI I think would deny that any measurement of a quantum system gives a single unique outcome, even an ordinary experiment that doesn't involve Wigner's friend style isolation.

BTW, I see in Hobson's paper Entanglement and the measurement problem he discusses a related experiment involving two entangled photons, where interference can be seen in the coincidence count for two photons but not for the photons individually:

each entangled photon “decoheres” [22] the other photon so that neither photon can interfere with itself. But coherence has not vanished. Coincidence probabilities such as Equations (10) and (11) show that coherence has been transferred to the biphoton. Thus the biphoton now interferes with itself across an arbitrary distance, i.e., nonlocally

I don't know whether in his interpretation this would mean there is a definite truth about each photon's path and we just lack information about it, though.

 

[JesseM]

@JesseM Division events are nomolgical. They are features of dynamical laws. Supernatural entities like the Wigner superobserver don't change what actually occurred when Wigner's friend made their measurement. But they do change the dynamics of Wigner's friend.

The dynamical laws of Wigner's friend, the transition matrix, will have a division event at Wigner's friend's measurement only until Wigner makes his measurement. Extending the transition matrix beyond Wigner's measurement means the matrix will not have a division event at Wigner's measurement.

Do you mean the answer to whether a division event happens at a particular moment (when Wigner's friend makes a measurement) can change at later times (when the box is opened and Wigner measures the interior)? If so does this depend on the conceptual choice of how you divide the world into subsystems, or would there be a division event when Wigner's friend inside the box makes their measurement regardless of the choice? Also would something similar be true for the delayed choice quantum eraser experiment, where the entanglement between signal and idler at the slit could be thought of as the idler having made a measurement of the signal photon, but one that can later be erased depending on where the idler is measured?

 

[PeterDonis]

I think the difficulty with assuming definite truths about recorded outcomes that happen in the box has to do with scenarios where the records of those outcomes are thoroughly erased prior to the box being opened by external observers.

Yes, but "thoroughly erased" is one thing when it's just a qubit--it's quite another when it's, say, Schrodinger's Cat, containing perhaps $10^{30}$ qubits. Even more so when you consider that the cat is emitting photons that are absorbed by the box walls, and that it's impossible to completely isolate the box itself from the rest of the universe, since the box also emits photons that can escape into space never to return.

In other words, even though, taken literally, QM only says that "thoroughly erased" is impossible for macroscopic objects "for all practical purposes", that might be more of an argument for not taking QM literally in this regime, and expecting that at some point we'll find some more complete theory that shows us how, once you're dealing with macroscopic objects and photons escaping to infinity, definite truths about recorded outcomes really are definite and irreversible, even according to the exact theory with no "for all practical purposes" qualifier needed.

Similar remarks would apply to "Wigner's Friend" type scenarios: for them to even work as described when you're dealing with humans and their macroscopic labs instead of single qubits, you're assuming that Wigner has the ability to undo decoherence on such scales, so that he can run experiments that detect the interference terms that are necessary for the scenario. But even leaving aside the number of degrees of freedom involved, what if the friend and his lab emit photons that escape to infinity? Then it's impossible for Wigner to undo decoherence or spot the interference effects: you can't catcn up with a photon. And that means the scenario simply can't work as described.

 

[JesseM]

Yes, but "thoroughly erased" is one thing when it's just a qubit--it's quite another when it's, say, Schrodinger's Cat, containing perhaps $10^{30}$ qubits. Even more so when you consider that the cat is emitting photons that are absorbed by the box walls, and that it's impossible to completely isolate the box itself from the rest of the universe, since the box also emits photons that can escape into space never to return.

In other words, even though, taken literally, QM only says that "thoroughly erased" is impossible for macroscopic objects "for all practical purposes", that might be more of an argument for not taking QM literally in this regime, and expecting that at some point we'll find some more complete theory that shows us how, once you're dealing with macroscopic objects and photons escaping to infinity, definite truths about recorded outcomes really are definite and irreversible, even according to the exact theory with no "for all practical purposes" qualifier needed.

Similar remarks would apply to "Wigner's Friend" type scenarios.

As I suggested in my first comment to jeffn1, to make a Wigner's friend thought experiment a bit more realistic we could imagine a large quantum computer running a simulation of the friend in the box, which can incorporate whatever sorts of technologies (eg temperatures near absolute zero and superconducting materials) are needed to make coherence of a large system more plausible. And we could also imagine the simulation involves an algorithmically much simpler model of an experimenter making measurements than a full physical simulation of a biological organism. In addressing the question of whether this simplified observer's measurements actually occurred in a single definite way if records were erased prior to our own measurement (Wigner opening the box), I'd think that whatever answer we give here, Occam's razor would tend to favor giving the same answer in the hypothetical case of a really huge quantum computer that could simulate an organism in microscopic detail.

 

[PeterDonis]

to make a Wigner's friend thought experiment a bit more realistic we could imagine a large quantum computer running a simulation of the friend in the box, which can incorporate whatever sorts of technologies (eg temperatures near absolute zero and superconducting materials) are needed to make coherence of a large system more plausible.

But that just underscores the point that no such thing happens with us humans and the measurements we make under ordinary conditions. Basically you're making an ordinary quantum eraser experiment involving many more qubits by artificially creating conditions that allow it. But what happens under conditions that don't allow it? That's the real question at issue.

 

[JesseM]

But that just underscores the point that no such thing happens with us humans and the measurements we make under ordinary conditions. Basically you're making an ordinary quantum eraser experiment involving many more qubits by artificially creating conditions that allow it. But what happens under conditions that don't allow it? That's the real question at issue.

Sure, but my intent was not to generalize to ordinary measurements. I'm just curious about what an interpretation like Barandes' would say about this type of exotic scenario because it seems to me that the answer would probably then generalize to much simpler experiments like the 2-particle delayed choice quantum eraser. If Barandes would commit to saying the measurements by Wigner's friend have definite outcomes even in cases where the information is erased before the box is opened by Wigner (and that those outcomes are part of a single consistent history before the erasure, without weird jumps between inconsistent memory states analogous to the jumps people have suggested might be found in non-Markovian realizer models of single particle behavior), shouldn't Barandes say the same about definite truths about which slit-like location the signal and idler came from even in cases where idler is measured at a location such that there is no which-path information available? And if not, what is the relevant distinction?

 

[PeterDonis]

my intent was not to generalize to ordinary measurements. I'm just curious about what an interpretation like Barandes' would say about this type of exotic scenario

Ok, fair enough.

 

[Morbert]

Do you mean the answer to whether a division event happens at a particular moment (when Wigner's friend makes a measurement) can change at later times (when the box is opened and Wigner measures the interior)? If so does this depend on the conceptual choice of how you divide the world into subsystems, or would there be a division event when Wigner's friend inside the box makes their measurement regardless of the choice? Also would something similar be true for the delayed choice quantum eraser experiment, where the entanglement between signal and idler at the slit could be thought of as the idler having made a measurement of the signal photon, but one that can later be erased depending on where the idler is measured?

Assuming the Wigner superobserver performs some recoherent operation analogous to a quantum eraser experiment: If by "Does a division event happen?" you mean "Do the subsystem dynamics $\Gamma(t)$ factorize at some time $t'$?", then the answer is "If the dynamics concern the stochastic process before recoherence, the dynamics factorize at $t'$. If the dynamics concern the stochastic process extended to after recoherence, they don't". If we are ever invaded by these superobservers capable of macroscopic quantum erasures, we will have to come to terms with our dynamics becoming very strange.

[edit] - User slightly clearer language

 

[JesseM]

Assuming the Wigner superobserver performs some recoherent operation analogous to a quantum eraser experiment: If by "Does a division event happen?" you mean "Do the subsystem dynamics $\Gamma(t)$ factorize at some time $t'$?", then the answer is "If the dynamics concern the stochastic process before recoherence, the dynamics factorize at $t'$. If the dynamics concern the stochastic process extended to after recoherence, they don't".

And would you say the statement about dynamics factorizing prior to recoherence doesn't depend on any choice of how we divide the world into subsystems, i.e. we either aren't allowed to treat everything inside the box as a single "subsystem", or even if we are allowed to do that, the statement that its dynamics factorize is still true?

Either way, how would this apply to a case like the delayed choice quantum eraser, at a short time after the interaction between signal and idler before there's been time for the idler to reach the "erasing" paths, so that if we chose to measure the idler at that time its location could tell us at which of the two locations the original signal/idler interaction happened? Would you similarly say that the dynamics of this two-particle system factorize at that moment, and if so would this depend on whether we treat each of the two particles as a separate subsystem?

If we are ever invaded by these superobservers capable of macroscopic quantum erasures, we will have to come to terms with our dynamics becoming very strange.

As I suggested in comments to others above, we could imagine the Wigner's friend experiment as a simplified measuring apparatus running in a quantum computer simulation, and then maybe erasure could mean something like a simulated bomb going off and scrambling all the structure that could qualify as records of the past, so that when the final output state of the simulation is measured, we'd just see something like a simulated gas at maximum entropy. This would require much larger quantum computers than anything we'll likely be able to create in the near future, but it doesn't seem as deeply fantastic as perfectly isolating a warm biological organism for long periods.

 

[Morbert]

And would you say the statement about dynamics factorizing prior to recoherence doesn't depend on any choice of how we divide the world into subsystems, i.e. we either aren't allowed to treat everything inside the box as a single "subsystem", or even if we are allowed to do that, the statement that its dynamics factorize is still true?

Different subsystems will have different dynamics, so the choice of division into subsystems will yield different subsystems and hence different dynamics models. For example if we include all degrees of freedom, then we are concerning ourselves with the unistochastic dynamics of the entire system, which will not have a division event at t'.

Either way, how would this apply to a case like the delayed choice quantum eraser, at a short time after the interaction between signal and idler before there's been time for the idler to reach the "erasing" paths, so that if we chose to measure the idler at that time its location could tell us at which of the two locations the original signal/idler interaction happened? Would you similarly say that the dynamics of this two-particle system factorize at that moment, and if so would this depend on whether we treat each of the two particles as a separate subsystem?

The actual quantum eraser experiment is a little trickier as it doesn't reduce to a simple reversion operation. The signal can be modeled as a stochastic process with a $\Gamma^\mathcal{s}(t)$ that always has a division event at $t'$ (the time when the signal and idler become correlated) regardless of what happens to the idler, as we marginalize over the idler. But if we instead consider the subsystem of the signal and the idler detectors, then $\Gamma^\mathcal{sD_i}(t)$ extended across the the entire experiment will not have a division event at $t'$.

 

[Sambuco]

and (2) while decoherence shows how off-diagonal elements in the reduced density matrix for the "system" can rapidly become tiny due to interactions with "environment", getting you arbitrarily close to a classical mixture, they wouldn't exactly go to zero except maybe in the limit of infinite time.

As an additional comment, in an ideal case, if the system is sufficiently isolated, the limit for very long times doesn't have to be with zero non-diagonal elements, since there could be a quantum Poincaré recurrence.

Lucas.

 

[Sambuco]

I do not think Barandes solves the measurement problem, its a refomulration.

If we consider his proposal that, at least for non-relativistic quantum mechanics, the position of particles taking a defined value at each time constitutes the ontology of the theory, then it does solve the measurement problem. It doesn't differ much from how Bohmian mechanics does it.

In Barandes framework, the construction of the relevant gamma and the decomposition of the total system into subsystems, is not explained from first principles. Instead of saying that the subsystem becomes entangled with a huge environment in an enlarged Hilbert space, Barandes describes the subsystem in terms of a configuration and a stochastic transition constraint.

I completely agree, but this wouldn't be necessary if Barandes' proposal were merely an "interpretation" of quantum mechanics. What you're saying implies a reconstruction program, and I agree that this would be desirable within Barandes' stochastic formulation, as it also is in other $\psi$-epistemic interpretations, such as RQM or QBism.

This gives a nice new perspective on the nature of dynamical law as the law is not a trajectory-generating flow but a transition constraint. But the deeper question remains how the relevant gamma is selected/evolved/stabilized?

I completely agree! If the theory proposes that the entire formulation in terms of Hamiltonian and states in Hilbert spaces is simply a mathematical convenience, then there should be a way to arrive at the formulation without going through it.

Lucas.

 

[Sambuco]

The formally approximate nature of decoherence corresponds to a formally approximate nature of division events: Γ(t←0)≈Γ~(t←t′)Γ(t′←0). But like decoherence, this divisibility is exact enough that no mortal can carry out an experiment that renders classical records unreliable.

An interesting aspect related to this is that Barandes mentions in his conversation with David Albert that, in order to formulate the theory, something like a "past hypothesis" is needed in which it is assumed that there was a division event that gave rise to the subsequent dynamics.

Lucas.

 

[Sambuco]

The Hobson View: Wigner’s external wavefunction does not dictate objective reality; it merely represents Wigner's lack of information.

In that case, Hobson's "interpretation" is not an interpretation, but a different theory with objective collapse, since it doesn't allow Wigner to model the laboratory where his friend is located as a system with unitary evolution.

An additional problem is that, unlike other objective collapse theories, such as the Ghirardi-Rimini-Weber theory, Hobson doesn't seem to suggest a criterion for determining when unitary evolution ceases to hold. He simply states that a collapse has occurred, and nothing more.

Wigner cannot perform a magic "super-measurement" to rewrite or erase the friend's memory because macroscopic decoherence has already locked that definite outcome into the physical universe irreversibly.

This is simply wrong. Decoherence can lead to quasi-classical states where irreversibility is true for all practical purposes, but not in principle.

Lucas.

 

[Sambuco]

Similar remarks would apply to "Wigner's Friend" type scenarios: for them to even work as described when you're dealing with humans and their macroscopic labs instead of single qubits, you're assuming that Wigner has the ability to undo decoherence on such scales, so that he can run experiments that detect the interference terms that are necessary for the scenario. But even leaving aside the number of degrees of freedom involved, what if the friend and his lab emit photons that escape to infinity? Then it's impossible for Wigner to undo decoherence or spot the interference effects: you can't catcn up with a photon. And that means the scenario simply can't work as described.

I think they are nothing more than overly idealized thought experiments that seek to understand the implications of assuming the universality of quantum mechanics. Of course, in practice, it's virtually impossible to achieve something like that, but the analysis remains valid. In other words, I see an analogy with the idea that the second law of thermodynamics, which can be derived à la Boltzmann from statistical mechanics, is not fundamental, but only true for all practical purposes. In fact, the Poincaré recurrence is assumed to exist.

In other words, even though, taken literally, QM only says that "thoroughly erased" is impossible for macroscopic objects "for all practical purposes", that might be more of an argument for not taking QM literally in this regime, and expecting that at some point we'll find some more complete theory that shows us how, once you're dealing with macroscopic objects and photons escaping to infinity, definite truths about recorded outcomes really are definite and irreversible, even according to the exact theory with no "for all practical purposes" qualifier needed.

Why not take quantum mechanics as valid in that regime? In fact, the prediction of quantum mechanics is precisely what you say: that the possibility of carrying out such an experiment is impossible, for all practical purposes.

Lucas.

 

[PeterDonis]

the analysis remains valid.

Not in the case where photons escape to infinity. In such a case, this...

the Poincaré recurrence is assumed to exist.

...or its quantum analogus, is simply wrong. Photon that escapes to infinity will never show interference with whatever other particles remain behind, and that means that decoherence can never be reversed--not even in a "for all practical purposes" sense, but in the "in principle" sense of the straight theoretical predictions.

Why not take quantum mechanics as valid in that regime?

I used the word "literally", not "valid", to emphasize that things like photons escaping to infinity are simply ignored in these thought experiments, even though there is no way to keep them from happening, not even in principle. It's impossible to have your lab and its contents at exactly absolute zero temperature, and anything at a finite positive temperature emits photons, and you can't catch them all. In an idealized analysis where all you want to do is predict measurement results for qubits, sure, that's fine; but the idealized thought experiments we've been discussing go way beyond that.

 

[Sambuco]

Not in the case where photons escape to infinity. In such a case, this...


...or its quantum analogus, is simply wrong. Photon that escapes to infinity will never show interference with whatever other particles remain behind, and that means that decoherence can never be reversed--not even in a "for all practical purposes" sense, but in the "in principle" sense of the straight theoretical predictions.


I used the word "literally", not "valid", to emphasize that things like photons escaping to infinity are simply ignored in these thought experiments, even though there is no way to keep them from happening, not even in principle. It's impossible to have your lab and its contents at exactly absolute zero temperature, and anything at a finite positive temperature emits photons, and you can't catch them all. In an idealized analysis where all you want to do is predict measurement results for qubits, sure, that's fine; but the idealized thought experiments we've been discussing go way beyond that.

Okay, now I understand your point of view better. Thanks for the reply!

Lucas.

 

[Fra]

We seems to partly understand each other, but just one comment on this

If we consider his proposal that, at least for non-relativistic quantum mechanics, the position of particles taking a defined value at each time constitutes the ontology of the theory, then it does solve the measurement problem. It doesn't differ much from how Bohmian mechanics does it.

From my perspective there is a huge difference. Yes some common denominators is

• the idea of "hidden variables" or "hidden configurations" that are "physically real"(whatever that means)
• these at least influence the future (wether via initial conditions or basic transition probabilities).

But other than that, there are some IMO critical differences.

I have two issues

• Two isomorphic spaces can be seem as the same mathematical space, but it doesnt mean its the same physical space. This is I think one difference between a HV beables in bohmain mechanivs and the configurations of each subsystem in Barandes picture, which are explicitly not beables. The configurations spaces of the subsystems has not a prioi "relation". Note this is the same in hilber cpiture. Two quantum systems has their OWN hilbertspace - the hilber solution is just to enlarge the space and consider them PARTS of the same bigger space. but that really does not solve anything, it moves the problem outwards.
• In particular thinking of barandes picture in terms of particles and 3D positions is questionable, except in simple examples perhaps. The generael case, I understnd his configurations to be abstract. Not representing position in old fashioned 3D space. And for sure, the configurations of TWO difference particles does not live in the "same physical space", at leat no a priori I think. It is something whose relations remains to be explained. I personally thing of "configurations" more is something abstract lie the subsystems physical state and memory record (in some abstract sense). And in this picture, constraining the mental picture to "particle postiions" misses out a much more beautiful concept, that has the potential to even make sense before there is emergent spacetime relations. (All this would then be implicilt in expecting a first principle construction of Gamma)

What you're saying implies a reconstruction program, and I agree that this would be desirable within Barandes' stochastic formulation, as it also is in other $\psi$-epistemic interpretations, such as RQM or QBism.

Yes we agree here. For me there "reconstruction program" is unavoidable though. It makes not sense to me to come up with something like Barandes does and then just leave it at that. That is weird.

I see it as giving the rest of the community a conceptual passing here!

/Fredrik

 

[Sambuco]

In particular thinking of barandes picture in terms of particles and 3D positions is questionable, except in simple examples perhaps.

I was referring only to Barandes' current formulation in which, for the case of non-relativistic QM, the configurations are effectively given by the positions of the particles. However, I agree that all of that should only be an intermediate step if the formulation is intended to be anything more than a simple interpretation.

Yes we agree here. For me there "reconstruction program" is unavoidable though. It makes not sense to me to come up with something like Barandes does and then just leave it at that. That is weird.

I fully agree! It's like the usual attempts to deduce Born's rule within the Everettian interpretation. If the only thing that exists is the wave function evolving unitarily, then the probabilities we observe experimentally must arise naturally and not as an additional postulate.

Lucas.