# I Danger for the Many-Worlds Interpretation?

So inner observer has collapse, outer observer has uncollapsed with device-system correlation. There isn't any real problem here that I can see. What's the contradiction?
I don't know why you keep bringing up classical mechanics; it's not somehow above reproach and I don't know anything about interpretation of classical theories or probability.

The contradiction is that after I measure something I should believe according to collapse that the world is actually in that state. While according to unitary evolution I know the world can not evolve from an unsure state to just one of them. You can't just say it's my knowledge, it doesn't matter that it's inconsistent. Or at least, you aren't going to convince me of anything with that argument. Because what I care about is having a consistent model of reality and the only model of reality I have is based on what I know.

I think my position should be clear by now.

#### DarMM

Gold Member
I don't know why you keep bringing up classical mechanics
It's mathematically less complex and displays the same features and is known to have no contradictions despite containing the features you claim are contradictory.

The contradiction is that after I measure something I should believe according to collapse that the world is actually in that state. While according to unitary evolution I know the world can not evolve from an unsure state to just one of them
This isn't a contradiction though that's the issue. There's no inconsistency here that's the point, which is why Wigner's friend isn't considered a contradiction or inconsistency in the literature.

Of course the unitary evolution of an outer observer who knows less will not give the same state as an inner observer who knows more. That's just a basic fact of probability theory. There's no contradiction here though. The inner observer knows that $\uparrow$ has been obtained, the outer observer does not. You'd expect these kinds of different states in a probabilistic theory and quantum mechanics is mathematically a generalized probability theory.

Obviously since these views allow different observers to have different quantum states it's not a closed deterministic account, but it's not contradictory. You might not find it satisfying, but nothing you've mentioned shows a formal contradiction.

Last edited:

#### A. Neumaier

since QM allows different observers to have different quantum states
Only some interpretations of QM do allow that!

#### DarMM

Gold Member
Only some interpretations of QM do allow that!
The discussion here is about those interpretations that do such as Copenhagen. I'll edit it.

Obviously since these views allow different observers to have different quantum states it's not a closed deterministic account, but it's not contradictory. You might not find it satisfying, but nothing you've mentioned shows a formal contradiction.
I didn't mention multiple observers; just one. Assuming I am Wigner's friend, I know that I measured up, but I also know that both up and down were obtained and Wigner might cause the two branches to interfere. How do I know what to expect in this situation? The only way to reason correctly about it is by assuming unitary evolution of the isolated system. It might be impossible for a human, but judging by current trends it should be possible to have a simple AI in a quantum computer with that experience.

With 2 observers the contradiction is more obvious. I don't know why you say in the literature it's considered consistent.

#### DarMM

Gold Member
Assuming I am Wigner's friend, I know that I measured up, but I also know that both up and down were obtained
No, you know the external observer will have up and down in their state assignment. That's not the same as up and down actually existing at the same time as macroscopic facts, that's an MWI reading of the state. Of course if you read the states in an MWI way then the two of them are contradictory. However that's not the way they are viewed in Copenhagen.

In Copenhagen the different state assignments are just due to the different epistemic conditions of the agents.

I don't know why you say in the literature it's considered consistent.
If it were a genuine contradiction for Copenhagen views they would have been finished in 1963. Nobody considers Wigner's Friend to show a genuine contradiction for Copenhagen, that's why people are trying to formulate the extended Wigner scenarios like Frauchiger-Renner.

In Copenhagen the different state assignments are just due to the different epistemic conditions of the agents.
Okay, so you assume when I measure up the state is actually up? Then you need to involve Wigner, who will undo that measurement and then redo it, giving a 50% change of interacting with a version of me that measured down. Since Wigner will sometimes interact with a down-version of me and I'll never interact with a down-version of Wigner we can't be living in one world. If Wigner is only able to create an up-version then unitary evolution isn't universally valid for isolated systems.

If it were a genuine contradiction for Copenhagen views they would have been finished in 1963. Nobody considers Wigner's Friend to show a genuine contradiction for Copenhagen,
This is not true. For example David Deutsch thinks there is a contradiction. [1] [2]

#### DarMM

Gold Member
Okay, so you assume when I measure up the state is actually up?
There's no "the state" in this view. I agree the inner observer would use $|\uparrow\rangle$ and that the macroscopic degrees of freedom of their device are those corresponding to up.

Then you need to involve Wigner, who will undo that measurement and then redo it, giving a 50% change of interacting with a version of me that measured down
There's nothing strange here though for two reasons.

First if a theory is stochastic and there is some dichotomic events which occurs one way say $A$, then somebody rewinds time of course it can happen the other way $B$. There's nothing that requires multiple worlds here. In one world a random event occurred, then time was rewound for the physical system partaking in the event and the random event occurred again. If the radioactive decay of an atom occurred, then I rewound time on the second pass it might not occur. Of course because it is a random event.

Secondly these reversals are completely unphysical for a variety of reasons. The Hamiltonians required to achieve them are unbounded below and thus are undefined, the machines required to perform them cannot exist for relativistic reasons (both kinematical and dynamical), the state conjugations required to achieve them are impossible outside of very simple systems like a few qubits, Weak processes can't reversed from violation of T symmetry. Even if they could be performed, I don't really see the issue as per the point above. A random event run twice can have two different outcomes.

This is not true. For example David Deutsch thinks there is a contradiction.
Well really I don't mean literally nobody, I mean it isn't a view held by the majority in quantum foundations even those opposed to Copenhagen style views. Deutsch claims there is a contradiction there, but doesn't specify it.

The second link treats thermalized macroscopic systems like qubits.

However since both a treating Deutsch's paper the resolution is just the well known one. Why would the friend think the qubit is either in the state $|+\rangle$ or $|-\rangle$ states since he knows the qubit was subjected to a Hadamard gate after reversal and Wigner had access to his environment?

#### Minnesota Joe

You are evolving by the Schrodinger equation when you "do the mathematical operation of collapse".
You most certainly are not. This is one of the most basic possible mathematical points about QM. The Schrodinger equation most certainly does not tell you to take a linear combination of terms and discard all but one of them, and then re-normalize the one term that's left.
I'm not saying that. I was saying you evolving into a state where you represent the spin state as collapsed is compatible with the Schrodinger equation. So is you evolving into a state where you don't represent it correctly and getting your predictions wrong, but your ancestors in parent worlds learned better. I thought you were complaining about how we come to know it, but I missed your point, my apologies.

They are not describing things like you describe. They keep the amplitudes in front of the world states and simply decompose states, but I don't have enough examples to be sure how it all works, so I posted a comment on Sean Carroll's blog asking him if he could address on Sabine's argument. Maybe he will.

#### PeterDonis

Mentor
I was saying you evolving into a state where you represent the spin state as collapsed is compatible with the Schrodinger equation.
No, it isn't, because this claim of yours assumes that "you" has a particular definite meaning that it doesn't and can't have if the MWI is true.

According to the MWI, if you observe a spin measurement that has two possible outcomes, you become entangled with the measured system and measuring apparatus, and the whole process ends up with two decohered branches. The ordinary way of speaking about the MWI says that in one branch you observe spin up, and in one branch you observe spin down; but this way of speaking obscures a key issue with the term "you" and what it means.

If "you" means a particular subsystem of the quantum system as a whole, then "you" do not have a definite state after observing the measurement, because "you" are entangled, and entangled subsystems don't have definite states. And the Schrodinger equation does not say that this "you" represents the spin state as collapsed; this "you" does not represent the spin state as anything, because this "you" is not in a definite state of its own. So on this meaning of "you", your claim is false: "you" do not evolve into a state where "you" represent the spin state as collapsed; "you" evolve into being entangled, with no definite state at all.

If "you" means a particular component of a particular decohered branch after the observation, then "you" no longer has a single referent: it has two, because there are two branches. Each individual referent of "you" represents the spin state as being definite--up or down--but there are now two of them, not one, and neither one can be picked out as "the" one that corresponds to the "you" before the measurement, so there is no unique Schrodinger evolution involving "you". So in this meaning of "you", your statement is also false: "you" do not evolve into a state where you represent the spin state as entangled; "you" have two time evolutions, not one, and the two evolutions correspond to different spin states of the measured system.

First if a theory is stochastic and there is some dichotomic events which occurs one way say $A$, then somebody rewinds time of course it can happen the other way $B$.
I think now the contradiction might even occur earlier. Wigner is able to reverse the evolution of the friend's system because his friend's system is evolving unitarially. Though the friend should know his system evolved non-unitarially. Or in Copenhagen terms, he created some information, that his system has the property $\uparrow$ and not the property $\downarrow$

How is it possible that the system has both evolved linearly according to Wigner and non-linearly according to his friend, without having a split of the friend? I mean what kind of state describes the system at this time?
Secondly these reversals are completely unphysical for a variety of reasons.
That's a strong argument against scalable universal quantum computers, but this is still an open question. I am assuming for this argument they are able to someday make at least ones of a few kilobits. If they are able to scale then they can of course run a program backwards. If they are able to get below a certain noise floor it's shown error correction can let them scalable scale up.

Well really I don't mean literally nobody, I mean it isn't a view held by the majority in quantum foundations even those opposed to Copenhagen style views.
Do you have any evidence of this? I always got the impression most many worlders consider Wigner's friend a contradiction. Another person that thinks Wigner friend show's an incompatibility is Lubos Motl, no fan of many worlds:

At any rate, in principle, the conclusion is surely correct. In the Wigner's friend setup, the descriptions of the phenomena by the two people must be fundamentally distinct and impossible to unify into one "collective" let alone "objective" description. Different observers have different lists of observations with their generally different results, and therefore generally different predictions for the future and different microscopic descriptions what happened.​

However since both a treating Deutsch's paper the resolution is just the well known one. Why would the friend think the qubit is either in the state $|+\rangle$ or $|-\rangle$ states since he knows the qubit was subjected to a Hadamard gate after reversal and Wigner had access to his environment?
Because he assumed non-unitary evolution (there is only one version of him), so Wigner can't reverse the measurement, at least not exactly.

#### DarMM

Gold Member
I think now the contradiction might even occur earlier. Wigner is able to reverse the evolution of the friend's system because his friend's system is evolving unitarially. Though the friend should know his system evolved non-unitarially. Or in Copenhagen terms, he created some information, that his system has the property $\uparrow$ and not the property $\downarrow$

How is it possible that the system has both evolved linearly according to Wigner and non-linearly according to his friend, without having a split of the friend? I mean what kind of state describes the system at this time?
The friend is not treating the same systems as Wigner. The friend observed $\uparrow$ and may associate this with the macroscopic degrees of freedom of his device and the microsystem. Wigner is tracking the entire lab down to the atomic level. Thus the friend's state is a state on the "device-system" Hilbert space $\mathcal{H}_{S}\otimes\mathcal{H}_{D}$, Wigner's state is a state on the "device-system-environment" $\mathcal{H}_{S}\otimes\mathcal{H}_{D}\otimes\mathcal{H}_{E}$.
So the question is is Wigner's state on $\mathcal{H}_{S}\otimes\mathcal{H}_{D}\otimes\mathcal{H}_{E}$ compatible with the friend's state on $\mathcal{H}_{S}\otimes\mathcal{H}_{D}$? Due to the superselection caused by decoherence it is, for Wigner's induced state on $\mathcal{H}_{S}\otimes\mathcal{H}_{D}$ is mixed across the two possible states for the macroscopic degrees of freedom, which is consistent with what the friend sees.

That's a strong argument against scalable universal quantum computers
I don't see how. Here we are discussing $\mathcal{O}\left(10^{27}\right)$ non-isolated systems with multiple degrees of freedom and subject to all the complexities of thermalization, weak processes, etc. Scaling Quantum Computers concerns attempting to isolate systems composed of around $\mathcal{O}\left(10^{5}\right)$ two level systems.

I don't see any reason to think that the superselections derived for macroscopic degrees of freedom imply you can't scale quantum computers, i.e. create controlled collections of isolated qubits.
At any rate, in principle, the conclusion is surely correct. In the Wigner's friend setup, the descriptions of the phenomena by the two people must be fundamentally distinct and impossible to unify into one "collective" let alone "objective" description. Different observers have different lists of observations with their generally different results, and therefore generally different predictions for the future and different microscopic descriptions what happened.​
This is just complimentarity and a basic aspect of probability theory. At least that's all Motl seems to be discussing to me. Looking at a few of his posts on the topic it just seems to be what Peres discusses in his monograph "Quantum Theory: Concepts and Methods", i.e. that complimentarity, different previous observations and bayesian updating can lead to two observers having different states. Wigner's friend is a fairly simple example, but there are others just from basic Quantum Information theory if you want to see them.

If Wigner seeing an uncollapsed state is consistent with what the friend see's then, he can't assume the state has collapsed. I think it's clear why people see this as an inconsistency and that many people do, so I am going to leave it at that.

observations and bayesian updating can lead to two observers having different states
It's not about having different states - it's about those states being incompatible. That's why Motl's position is a consistent version of Copenhagen; because he claims it's not allowed to reconcile the states of two different observers.

EPR

#### DarMM

Gold Member
If Wigner seeing an uncollapsed state is consistent with what the friend see's then, he can't assume the state has collapsed
Why though? Wigner's uncollapsed state for the entire "micro-device-lab" system is not in contradiction with the friend having obtained a result and using a collapsed state for the "micro-device" system because of dynamical and environmentally driven superselection in the macroscopic degrees of freedom. This is the reason the basic Wigner's friend set up is not a problem for Copenhagen. See Richard Healey's "The Quantum Revolution in Philosophy" Chapter 11 for the simplest exposition. I really don't see anything wrong with this classical solution since decoherence means that Wigner's state induces a mixed state on the "micro-device" system.

Of course all of this ignores that Wigner's friend is an impossible unphysical situation.

It's not about having different states - it's about those states being incompatible. That's why Motl's position is a consistent version of Copenhagen; because he claims it's not allowed to reconcile the states of two different observers.
In this case the states are different, but they are not incompatible. I don't see how they are given the superselection amongst the macroscopic degrees of freedom.
As I said what Motl mentions is something that occurs in quantum information generally. It isn't viewed as a contradiction there. Bayesian updating permits observers to have opposed states even without encapsulation.

Last edited:

#### DarMM

Gold Member
I think it's clear why people see this as an inconsistency and that many people do, so I am going to leave it at that.
I don't doubt that some see it as an inconsistency, but the consensus in quantum foundations is that it is not. See the notes about Copenhagen interpretations for Matt Leifer's (who is not a Copenhagener) Foundations course here:

He explicitly says about the basic Wigner's friend on p.19:
In any case differences of quantum state assignment do not prove that objective observations do not exist. For that we need Frauchiger-Renner.

#### EPR

Gold Member
There is a reason why the MWI is not a textbook quality interpretation but the CI is. No unwarranted assumptions needed for the CI instead of the usual "classical" pointer states, einselection by an introduced but unexplained macro environment etc. stuff under the rug.

Last edited:

#### Michael Price

I think the problem is this:

The wave-function collapse, I have to emphasize, is not optional. It is an observational requirement. We never observe a particle that is 50% here and 50% there. That’s just not a thing. If we observe it at all, it’s either here or it isn’t. Speaking of 50% probabilities really makes sense only as long as you are talking about a prediction.

If Hossenfelder's statement is true then I think the MWI fans have a problem, because they can't deny that outcomes are observed which then requires the wave-function collapse.

But is this statement true?
No, there are no required additional assumptions in MWI. The whole point of MWI is to ditch all these measurement related assumptions, since they are unnecessary, including the Born Rule and the classical-quantum boundary - as assumptions. SH is playing with words. There are a host of (putative)) ways of deriving the Born Rule, as Demystifier says - I think they're all mostly pretty good.

#### Michael Price

There is a reason why the MWI is not a textbook quality interpretation but the CI is. [...]
MWI does not feature in textbooks so as to not to "scare the horses". Just as QED is not taught at kindergarten. It would blow too many tiny minds.

#### gill1109

Gold Member
Suppose we have N Stern-Gerlach devices. Measure N spins each initially in an equal weight superposition of up and down. We now have 2^N worlds all with equal weights. Now let N tend to infinity and you will find that in almost all of those worlds, very close to half of the spins came up "up" and half came up "down". Some people see this as an argument for the Born rule within the "no-collapse" MWI (there is only unitary evolution of the wave function of the universe). I think that this is an FAPP argument, saying that one mathematical model gives similar predictions to another, it does not derive us the Born rule *within* MWI. I have to read Everett. So far I never bothered, I thought MWI was just many *words*, deliberately mixing upmetaphorically the branching of collapse with the parallel worlds of unitary. i.e., poetry, not science.

#### EPR

Gold Member
The MWI is a poor quality interpretation. The human perception is not a reliable copy of the external world as we've come to (re)discover in quantum mechanics. We can't presuppose the existence of that we are trying to prove(classical states and classical environment). It's a fable for weak minds who can't deal with the(nonexistence of a fixed classical) reality.

#### vanhees71

Gold Member
No, there are no required additional assumptions in MWI. The whole point of MWI is to ditch all these measurement related assumptions, since they are unnecessary, including the Born Rule and the classical-quantum boundary - as assumptions. SH is playing with words. There are a host of (putative)) ways of deriving the Born Rule, as Demystifier says - I think they're all mostly pretty good.
If you say the Born rule is unnecessary, how do you make sense of what's implied by a specific state in terms of what's observable? I've never experienced that I split myself in parallel universes, where the one or the other outcome of an observation occurs. I always get one result when observing something, though even given complete knowledge (i.e., the preparation of the system measured in a pure state) and measuring some observable not determined by the state, I get a random result with the probabilities predicted by Born's rule but always one of the possible results and not some ambivalent splitting of the system, the measurement device and myself.

Perhaps you can have such an experience by taking some drugs, but then I'd not consider this as an objective observation of nature anymore ;-)).

#### Michael Price

If you say the Born rule is unnecessary, how do you make sense of what's implied by a specific state in terms of what's observable? I've never experienced that I split myself in parallel universes, where the one or the other outcome of an observation occurs. I always get one result when observing something, though even given complete knowledge (i.e., the preparation of the system measured in a pure state) and measuring some observable not determined by the state, I get a random result with the probabilities predicted by Born's rule but always one of the possible results and not some ambivalent splitting of the system, the measurement device and myself.
QM is linear, so the result(s) you get on some subset of worlds is (as Everett pointed out) indifferent to the presence of absence of other worlds. This ensures the splitting is not an observable process. If the cat is dead, you see the cat dead. If the cat is alive, you see the cat alive. If the cat is a superposition of alive and dead, then you become a superposition of seeing the cat alive alive and seeing the cat dead - NOT one observer aware of zombie-cats.

#### vanhees71

Gold Member
If the cat is in a superposition of "dead" and "alive" (for me a non-sensical idea to begin with, because "dead" and "alive" are so coarse grained information that for sure you cannot describe them as a pure state, but nevertheless we can argue, considering a q-bit), the state of the observable which can take the values "dead" and "alive" is indetermined. When I investigate the cat to see whether it's "dead" or "alive", given the state I only know probabilities for finding the cat's "dead-or-alive observable" taking the one or the other possible value. When I measure it, I get a unique result. Nothing splits, and it doesn't have any advantage of thinking that something splits since it has no observable consequences at all to make that assumption. So I can just leave out this idea of splittings and live with minimally interpreted QT.

#### Minnesota Joe

No, there are no required additional assumptions in MWI. The whole point of MWI is to ditch all these measurement related assumptions, since they are unnecessary, including the Born Rule and the classical-quantum boundary - as assumptions. SH is playing with words. There are a host of (putative)) ways of deriving the Born Rule, as Demystifier says - I think they're all mostly pretty good.
Where is she "playing with words" specifically? Also, where does she complain about deriving the Born Rule? She seems to address that in the comments where she elaborates:
It's not the probability measure problem, as I am not worried about how to weigh different branches.

It's not the preferred basis problem as that is basically solved by decoherence.

I am asking why is the forward evolution of what is a detector at t_0 no longer a detector at t_1>t_0. The answer to this is that, by assumption, the forward-evolved detector is not what many worlds fans want to call a detector. So you need an additional assumption and this assumption is virtually equivalent to the measurement postulate in Copenhagen. I use virtually to mean "up to interpretation".
http://backreaction.blogspot.com/2019/09/the-trouble-with-many-worlds.html?showComment=1569568819019#c6700348820680074317

#### Michael Price

If the cat is in a superposition of "dead" and "alive" (for me a non-sensical idea to begin with, because "dead" and "alive" are so coarse grained information that for sure you cannot describe them as a pure state,[....]
Yes you can so describe the two coarse grained states in a single pure state.