Do worlds of MWI ontologically have to exist?

[entropy1]

Is MWI equivalent to a superposition of possible Collapse-worlds? That is, is it equal to a superposition of the possible scenario's given by Collapse Interpretation?

Is it imparative that the "worlds" given by MWI all ontologically exist?

 

[Demystifier]

Is MWI equivalent to a superposition of possible Collapse-worlds? That is, is it equal to a superposition of the possible scenario's given by Collapse Interpretation?

Yes and yes.

Is it imparative that the "worlds" given by MWI all ontologically exist?

No.

 

[PeterDonis]

Is it imparative that the "worlds" given by MWI all ontologically exist?

Do you mean according to the MWI? According to the MWI, yes, they all exist, because they're all part of the wave function, and the wave function exists.

If you mean, does the MWI have to be true, no, it doesn't.

 

[DarMM]

Is MWI equivalent to a superposition of possible Collapse-worlds? That is, is it equal to a superposition of the possible scenario's given by Collapse Interpretation?

In simple situations currently modeled yes. Though in general it's not exactly certain what constitutes a "world" and there is a possibility that observers in the same world could disagree on the "world" of distant objects.

I know Renato Renner's group in Zurich is currently investigating violations of the Born rule in MWI and these kind of inter-agent consistency issues.

 

[Gerinski]

I don't know how correct this is, but I read somewhere that some people interpret MWI in such a way that many (if not most) of the worlds described by the wavefunction actually cancel each other so they do not produce a physical rendering.
When taken to the extreme, all of the worlds cancel out except one which is the one we experience. This would be an attempt to reconcile the MWI with the fact that we only experience one world.

 

[PeterDonis]

I read somewhere

Please give a specific reference. We can't discuss vague statements about what an unknown source said.

 

[Gerinski]

I may try to find where did I read that, it's in some of the popular science books I have, but I don't know if I will find it. Kindly forgive me if I don't.

 

[DarMM]

I may try to find where did I read that, it's in some of the popular science books I have, but I don't know if I will find it. Kindly forgive me if I don't.

If it helps, I think you might be talking about Feynman's sum over paths where all the paths cancel except one (our classical) one. Plenty of pop science books explain each path as another world and use this to explain our classical looking world.

It's not very accurate to the physics itself, but I think that might be your source.

 

[PeterDonis]

I may try to find where did I read that, it's in some of the popular science books I have, but I don't know if I will find it. Kindly forgive me if I don't.

Any time you find yourself saying "I read somewhere", you should stop posting and go find where, and if you can't, don't post. Again, we can't discuss vague descriptions of something said by an unknown source.

 

[entropy1]

Is MWI in principle the assertion that all possible outcomes occur simultaneously? So how does that solve collapse if not only by asserting that all collapses occur simultaneously?

 

[DarMM]

Is MWI in principle the assertion that all possible outcomes occur simultaneously? So how does that solve collapse if not only by asserting that all collapses occur simultaneously?

Because if you model a measurement with quantum mechanics, including the main macroscopic features of the experimental device itself in your model and you ignore the environment (environment here can be either the actual external environment, e.g. the air, or the atomic structure of the device itself) you get one term for each combination of "state of the quantum system and state of the device"

For example if you are measuring a particle that is in a superposition of spin up and spin down you get a superposition of "particle is spin up and device reads up" and "particle is spin down and device reads down". That's what quantum mechanics naturally gives.

Now in reality you only see one of these alternatives, so you have to use a "collapse" rule to get rid of the other.

Many Worlds doesn't need to because it simply says both alternatives occur, hence it doesn't need collapse as it isn't pruning away one of the options.

 

[A. Neumaier]

Many Worlds doesn't need to because it simply says both alternatives occur

But it does not explain why each observer sees only one alternative, and why different observers usually agree on the alternative seen.

 

[DarMM]

But it does not explain why each observer sees only one alternative, and why different observers usually agree on the alternative seen.

Even beyond this there is no non-circular reason for why each alternative is seen with the frequencies we observe, hence it cannot be said to be an interpretation of any of the formalisms of Quantum Mechanics (QM has three formalisms, no collapse, objective collapse and subjective collapse which give different predictions in certain obscure scenarios such as the Frauchiger Renner experiment. MWI cannot yet be said to be a functioning no collapse interpretation)

In addition one will often see the claim that MWI is local, but this is still an open issue.

 

[stevendaryl]

Even beyond this there is no non-circular reason for why each alternative is seen with the frequencies we observe, hence it cannot be said to be an interpretation of any of the formalisms of Quantum Mechanics (QM has three formalisms, no collapse, objective collapse and subjective collapse which give different predictions in certain obscure scenarios such as the Frauchiger Renner experiment. MWI cannot yet be said to be a functioning no collapse interpretation)

In addition one will often see the claim that MWI is local, but this is still an open issue.

It's true that it's difficult to understand probabilities in MWI, but I sort of feel like that's because there is no really good way to understand probabilities, when you get right down to it.

Let me sketch a "many-worlds" version of classical probabilities.

Suppose that we have a world that is deterministic except for one little aspect: There is a special coin such that flipping it produces heads/tails with equal probability, and there's no way, even in principle, to predict which.

Now, as far as anybody is concerned, the coin flip is completely nondeterministic. But secretly, this world is not real, but is a simulation in a gigantic computer. Every time in the simulation when someone (one of the simulated beings) flips the coin, the computer halts the simulation, makes a duplicate of the state, and then continues simulating both branches, giving one branch the result "heads" and the other branch the result "tails". As time goes on, there are more and more simulated universes, with different histories of coin flips.

Now for any simulated being living in one of the simulations, the coin seems perfectly random, with equal chance of heads and tails. But the operation of the master computer is deterministic. So determinism in an ensemble is perfectly compatible with nondeterminism in each element of the ensemble.

But what about probability? As the number of flips gets larger and larger, the number of simulations in which half the coin flips turned up "heads" gets much larger than the number of simulations in which there is a large imbalance of "heads" and "tails". So we can sort of understand the probabilistic statement "the coin has 50/50 chance of landing heads or tails" in terms of "typical" and "atypical" branches. The overwhelming majority of simulations will have 50/50 results, so we can call those "typical" worlds. In the typical worlds, the relative frequency of heads approaches 1/2.

However, this is really unsatisfying, for the following reason: What if, instead of creating two copies every time a coin is flipped, the computer creates 3 copies, and allows two copies that are given result "heads" and 1 copy that is given the result "tails"? Now, the "typical" branch has 2/3 heads, rather than 1/2. So with this variation, the coin does not have 50/50 odds, but 33/67 odds. That makes sense, except for the fact that in each world, the simulated people empirically determine the odds of heads and tails by just flipping many times and counting. Obviously, the counting process is not affected by the existence of more alternate worlds. What's changed by the addition of alternate worlds is not the experience of anyone in any of the worlds, but simply the definition of which worlds are considered "typical".

I claim that there is no completely satisfying way to give meaning to the claim "the coin has 50/50 chance of being heads". You want it to be the case that probabilities correspond to relative frequencies, but that will only be true in some possible worlds, and not others. You can make it true by definition by calling the worlds where it is false "atypical", and restricting your probabilistic claims to the typical worlds. But that makes it all pretty tautological.

It seems to me that there is no good explanation for probability in a many-worlds setting, and furthermore, you can't empirically distinguish between such a many-worlds setting and a single-world with nondeterminism (as far as you're concerned, there is no difference, as long as there is no interaction between possible worlds---you can just define your world to be the "real" one, and consider the others just hypotheticals).

 

[bhobba]

Is MWI in principle the assertion that all possible outcomes occur simultaneously? So how does that solve collapse if not only by asserting that all collapses occur simultaneously?

It's the assertion all there is, is a universal wave-function. There are a few variations from that point on - some are like you suggest, others are more subtle. Best you study an actual paper on it:
https://arxiv.org/ftp/arxiv/papers/1801/1801.08587.pdf

And listen to Gell-Mann:


Thanks
Bill

 

[bhobba]

It seems to me that there is no good explanation for probability in a many-worlds setting, and furthermore, you can't empirically distinguish between such a many-worlds setting and a single-world with nondeterminism (as far as you're concerned, there is no difference, as long as there is no interaction between possible worlds---you can just define your world to be the "real" one, and consider the others just hypotheticals).

You and many others - its one of it disputed features as the paper I posted above explains:
'Finally, a still disputed claim of Everett’s and DeWitt’s was that the probabilistic predictions of standard quantum theory arose naturally from the formalism with no necessity of postulating a statistical interpretation (i.e., the Born rule).'

That's one reason I say its a bit subtle - there are points of disagreement about it.

Thanks
Bill

 

[bhobba]

But it does not explain why each observer sees only one alternative, and why different observers usually agree on the alternative seen.

Yes there are a number of hidden assumptions in the theory. I think some like Wallace believe that one is solved - but it has been thrashed out a number of times on this forum - no need to go over it again.

Personally I am with Gell-Mann who thinks decoherent histories is just MW without the many worlds and doesn't really understand why some want to invoke the actual existence of many worlds. That's his view - I can say mine - but it is just a view and will only lead to stuff that has been discussed a number of times before.

Thanks
Bill

 

[DarMM]

@stevendaryl I think you are coming upon the typical back and forth one sees between Adrian Kent and David Wallace.

The typical resolution is simply to say one should prove that the typical worlds have Born frequencies, which becomes a deterministic statement in Many Worlds as the whole ensemble is real.

However there is no non-circular proof of this either.

I should say recent work has shown that in Many Worlds there isn't a countable number of worlds, but a world volume. So then you are hoping to show that the volume of worlds that experience an outcome is given by the worlds' amplitude squared. It's been shown that the square of the amplitude is the only volume preserving way of subdividing world volumes.

In addition, one should look to more recent work, like the Pusey-Leifer theorem concerning operational time symmetry, which show Many Worlds would have to be fine tuned to avoid disagreeing with QM.

 

[A. Neumaier]

Every time in the simulation when someone (one of the simulated beings) flips the coin, the computer halts the simulation, makes a duplicate of the state, and then continues simulating both branches, giving one branch the result "heads" and the other branch the result "tails".

You may like Bertrand Russell's The metaphysician's nightmare.

 

[stevendaryl]

But it does not explain why each observer sees only one alternative, and why different observers usually agree on the alternative seen.

It's not clear to me that there is anything to explain here. You have to ask what it would mean to see more than one alternative.

Suppose you decide, in the case of Schrodinger's cat, that you will write down on a piece of paper, either the word "Dead", "Alive" or "Both", depending on what kind of cat you see (both meaning a superposition of dead and alive). Then if the cat is dead, you'll definitely write "Dead". If the cat is alive, you'll definitely write "Alive". But given those two facts, what you do in the case of a half-dead/half-alive cat is determined: You will be put into a superposition of writing "Dead" and writing "Alive". To write "Both" would violate the linearity of quantum mechanics.

 

[stevendaryl]

Personally I am with Gell-Mann who thinks decoherent histories is just MW without the many worlds and doesn't really understand why some want to invoke the actual existence of many worlds.

Here's what I don't understand about decoherent histories. Given an initial state $|\psi\rangle$, a hamiltonian $H$, a sequence of times $t_1, t_2, ...$ and a sequence of projection operators $\Pi_1, \Pi_2, ...$, then the Born rule can give a probability associated with the history:

The proposition corresponding to $\Pi_1$ is true at time $t_1$ and the proposition corresponding to $\Pi_2$ is true at time $t_2$, etc.

To me, this is double-many worlds. For every sequence of $N$ propositions, you have $2^N$ possible sequences of yes/no answers. So have a different world for each choice of a sequence of questions, and for each possible sequence of answers.

Now, you can pare down the many worlds to just one world evolving nondeterministically by fixing the sequence of questions and allowing the history to grow nondeterministically according to the Born Rule. However, is the sequence of questions an additional structure? Who determines the sequence of questions (projection operators)?

 

[A. Neumaier]

You will be put into a superposition of writing "Dead" and writing "Alive". To write "Both" would violate the linearity of quantum mechanics.

and to write either Dead or Alive, too.

 

[bhobba]

Who determines the sequence of questions (projection operators)?

Well in decoherent histories QM is considered the stochastic theory of all history's. It just like anything a theorist can define.

My issue with it is it seems to be skirting a number of key issues such as exactly what is an observation and trying to define your way out of trouble. It may work - but seems somewhat artificial - still its not a completed program yet - some key theorems are missing.

Thanks
Bill

 

[stevendaryl]

Well in decoherent histories QM is considered the stochastic theory of all history's. It just like anything a theorist can define.

But to turn it into a stochastic theory, you have to specify the projection operator used. How does that get determined?

 

[bhobba]

But to turn it into a stochastic theory, you have to specify the projection operator used. How does that get determined?

A projection operator is an operator such that P^2 = P. A sequence of such operators is called a history. Its well defined. Its supposedly the stochastic theory of all such sequences. Obviously some will have zero or so close to zero probability so as to be of no value. There is a general theorem built from the Born Rule that gives that probability.

Added Later
Some histories are not stable and it seems decoherence places certain conditions on what histories actually can occur and those that cant. Its been a while since I went deep into it but managed to find the following theses on it:
http://www.cs.ox.ac.uk/people/bob.coecke/Roman.pdf

Thanks
Bill

 

[stevendaryl]

A projection operator is an operator such that P^2 = P. A sequence of such operators is called a history. Its well defined. Its supposedly the stochastic theory of all such sequences. Obviously some will have zero or so close to zero probability so as to be of no value. There is a general theorem built from the Born Rule that gives that probability.

Yes, I know what a projection operator is. I asked how the projection operators are chosen.

 

[bhobba]

Yes, I know what a projection operator is. I asked how the projection operators are chosen.

I am confused - as I said there is no restriction - all are considered. Some are of no relevance, are inconsistent or for other reasons not worried about. The linked article gives some of the details eg about division into families and their consistency - it's part of the other name consistent histories.

Thanks
Bill

 

[stevendaryl]

I am confused - as I said there is no restriction - all are considered.

I know, I'm saying that without such a restriction, it isn't a stochastic theory. A stochastic theory tells the probability of a future state given the current state. Consistent Histories doesn't do that.

 

[bhobba]

Of course there are restrictions in order for a history to make sense eg they need to have a practicable probability, they are grouped into families and families must be consistent, have a Boolean structure that is classically logical etc. The exact detail I at one time knew more about - that was a bit ago now, and I can't recall many of those details, but the thesis I gave the link to seems to give the full details and reasoning on how they are used, as well as issues.

Thanks
Bill

 

[AlexCaledin]

What ontologically have to exist, it's some different "sorts of reality", according to Gell-Mann

 

[bhobba]

What ontologically have to exist, it's some different "sorts of reality", according to Gell-Mann
View attachment 233111

Your guess is as good as mine - we do not discuss what 'some sort of reality means' here - if that is what interests you go over to philosophy forums. Aside from the video I linked to previously Gell-Mann did another related one that may shed some further light on the issue:


Thanks
Bill

 

[stevendaryl]

I know, I'm saying that without such a restriction, it isn't a stochastic theory. A stochastic theory tells the probability of a future state given the current state. Consistent Histories doesn't do that.

I really can't make any sense of Consistent Histories as an interpretation. To me, it's basically Copenhagen bumped up a notch---rather than talking about single measurement outcome whose probability is given by the Born Rule, it shifts focus to an entire history whose probability is given by a more sophisticated version of the Born rule. I don't see how it's an interpretation, though.

 

[Carsten Milkau]

There is a really interesting recent article on MWI in quanta magazine. It shows a fundamental flaw which is really hard to grasp. I think the author failed to realize that this flaw is not unique to MWI, it is just most obvious there. The basic problem is that most interpretations say that observations are random - but how do you tell something is random once you have observed it? If you know all fair dice rolls that have ever happened, how would you asses whether dice rolls are random? The basic model of randomness doesn't say anything about particular outcomes, only about "a tendency for many repetitions" and you cannot repeat the universe.
Well MWI is special on that in its simplest form it doesn't talk about probability but superposition amplitude. This has the same problem though: if everything that can happen does happen, even absurd (extremely improbable) things do happen in some "parallel world". But our world looks normal, not absurd. And it doesn't help to say absurd parallel worlds have a small amplitude because a parallel world cannot feel its own amplitude, so that doesn't explain why ours isn't absurd.
There are ways out of this. Kolmogorov complexity is a way to define what randomness can mean for a finite, unrepeatable sequence of observations. Maybe some process in nature suppresses all simple patterns in observations (except those that are always present - the laws of nature). The Boltzmann brain thought experiment suggests there should be - otherwise it's more likely to be a Boltzmann brain than to be a sentient being in a world with other sentient beings.

 

[entropy1]

Could one interpret MWI as the assertion that a measurement puts the observer in a superposition of worlds that are all to a certain degree real?

For example: if the measurement yields $\frac{1}{\sqrt{5}}|w_{0}\rangle|W_{0}\rangle+\frac{2}{\sqrt{5}}|w_{1}\rangle|W_{1}\rangle$, could one say that $\frac{2}{\sqrt{5}}|w_{1}\rangle|W_{1}\rangle$ is 'more real' than $\frac{1}{\sqrt{5}}|w_{0}\rangle|W_{0}\rangle$?

 

[DarMM]

Could one interpret MWI as the assertion that a measurement puts the observer in a superposition of worlds that are all to a certain degree real?

For example: if the measurement yields $\frac{1}{\sqrt{5}}|w_{0}\rangle|W_{0}\rangle+\frac{2}{\sqrt{5}}|w_{1}\rangle|W_{1}\rangle$, could one say that $\frac{2}{\sqrt{5}}|w_{1}\rangle|W_{1}\rangle$ is 'more real' than $\frac{1}{\sqrt{5}}|w_{0}\rangle|W_{0}\rangle$?

I've seen the phrase "more real" used, so it would agree with how some see it.

Another explanation used in modern derivations of the Born rule in MWI is that there are an infinity of worlds and the $\frac{2}{\sqrt{5}}$ represents the volume in world space of all the worlds with the $|w_1, W_1\rangle$ outcome, i.e. 80% experience it.

 

[bhobba]

The basic model of randomness doesn't say anything about particular outcomes, only about "a tendency for many repetitions" and you cannot repeat the universe.

Oh yes it does - its called the Kolmogrov axioms and the strong law of large numbers that follows from those axioms:
https://en.wikipedia.org/wiki/Law_of_large_numbers

You can't say anything about a single observation, other than the possible outcomes you can get - but in a large ensemble of systems prepared exactly the same way (technically the same state) the same observation performed on them all will yield results as per the strong law of large numbers.,

And yes that you cannot observe the universe is an issue with interpretations of QM.

Thanks
Bill

 

[stevendaryl]

You can't say anything about a single observation, other than the possible outcomes you can get - but in a large ensemble of systems prepared exactly the same way (technically the same state) the same observation performed on them all will yield results as per the strong law of large numbers.

I wouldn't say "will yield results..." I would say "will probably yield results..." (with probability approaching 1).

 

[bhobba]

To me, it's basically Copenhagen bumped up a notch---rather than talking about single measurement outcome whose probability is given by the Born Rule, it shifts focus to an entire history whose probability is given by a more sophisticated version of the Born rule.

What it tries do do is as detailed at the start of the thesis I posted:
Its from 'Pioneering work by Griffiths, extended among others by Omnes, Gell-Mann and Hartle, has led to a new formulation of quantum mechanics in which no observer or quantum-classical divide is required and wave function collapse does not occur. This interpretation is known as ‘consistent histories’.

It is supposed to fix the issue in Copenhagen of what is a measurement - everything is quantum from the start - no collapse - just a quantum state. A book explaining it at the intermediate level has kindly been made available by Griffiths:
http://quantum.phys.cmu.edu/CQT/index.html

I read it about 10 years ago now, liked it but it seemed a bit too contrived so didn't go deep into it. That's the thing about all these interpretations - you can't really know the deep detail of them all, only a general outline. I did study the Emergent Multiverse in its entirety and its a bit more advanced, and interestingly actually similar in many ways to Decoherent Histories

Thanks
Bill

 

[eloheim]

In addition, one should look to more recent work, like the Pusey-Leifer theorem concerning operational time symmetry, which show Many Worlds would have to be fine tuned to avoid disagreeing with QM.

Just wondering to which result this is referring?
https://arxiv.org/abs/1607.07871 "Is a time symmetric interpretation of quantum theory possible without retrocausality?" maybe?

I think in that result they say it doesn't apply to Everett (MWI) or Bohm (although I may be misreading it, or we're talking about different things here...it gets confusing to me). If that is the paper you're referring to I may just need to study it more.

 

[stevendaryl]

What it tries do do is as detailed at the start of the thesis I posted:
Its from 'Pioneering work by Griffiths, extended among others by Omnes, Gell-Mann and Hartle, has led to a new formulation of quantum mechanics in which no observer or quantum-classical divide is required and wave function collapse does not occur.

I don't think it actually accomplishes that, though.

 

[DarMM]

Just wondering to which result this is referring?
https://arxiv.org/abs/1607.07871 "Is a time symmetric interpretation of quantum theory possible without retrocausality?" maybe?

I think in that result they say it doesn't apply to Everett (MWI) or Bohm (although I may be misreading it, or we're talking about different things here...it gets confusing to me). If that is the paper you're referring to I may just need to study it more.

So a symmetry can show up in three ways in a theory let's say.

1. Ontologically, that is it really is a genuine fundamental symmetry of the theory
2. Dynamically, where the symmetry is not naturally present but becomes effectively true because of some effect of evolution over time. An example might be where some liquid freezes into a crystal. The fundamental molecular laws don't have lattice displacement symmetries, but the crystals do.
3. Fine-tuning, where the initial conditions are chosen so that the symmetry looks like it is true, but they aren't. An example would be a pencil being perfectly set up to balance on its tip. This situation is perfectly circularly symmetric, but the laws and the pencil itself are not by default. The "tuning" part refers to how the symmetry would be destroyed by the slightest perturbation of the situation

So there is a symmetry in QM called Operational Time Symmetry. What they prove in that paper is that if you don't have retrocausal or acausal theories, then Operational Time Symmetry can't be Ontological, i.e. can't be a fundamental symmetry. Now this is an observed symmetry in experiments, so you do have to explain it. By dropping Retrocausality then, you only have recourse to showing it arises dynamically or via fine-tuning.

If you choose dynamically you really have to prove this occurs and you would be letting in the fact that before the relevant dynamical process took place (e.g. the freezing of the liquid above) the symmetry would not be present. Since it is a symmetry of QM, this would mean Many-Worlds would disagree with standard QM at some point in the early history of the universe and hence cosmological observations could disprove it.

Choosing fine-tuning arguments is considered artificial, as basically the multiverse in MWI would have to have started off in precisely the correct state to make it look as though Operational Time Symmetry was a fundamental symmetry, when it isn't.

They discuss Bohmian Mechanics and Many-Worlds in the last section of the paper.

 

[entropy1]

Is a superposition itself ontologically consisting of multiple realities? Or is a superposition occurring in one and the same reality? (And is the separation into worlds the result of decoherence then?)

 

[cosmik debris]

I've seen the phrase "more real" used, so it would agree with how some see it.

What is the difference between real, more real, very real, and really real?

Cheers

 

[DarMM]

What is the difference between real, more real, very real, and really real?

Cheers

I honestly don't know what is meant by the phrase, I could never make sense of it, which is why I mentioned the world volume explanation of the coefficients, as that makes more sense to me.

 

[bhobba]

I honestly don't know what is meant by the phrase, I could never make sense of it, which is why I mentioned the world volume explanation of the coefficients, as that makes more sense to me.

The term reality should be used very sparingly and in very obvious ways in physics. Even then you rub into issues best discussed, not here, but in philosophy. As I have posted before it used to bore Richard Feynman so much during a class he took on it at MIT he purchased a little drill and spent his time in class drilling little holes in his shoes he was so bored. He had to write an essay on consciousness. In typical Feynman style he did experiments on himself while falling to sleep and wrote up his experiences. No discussion of what consciousnesses was - just some observations on it in regard to dreaming and being semi awake. We now know about various brain states etc like alpha state and certain practices like Tai Chi promote being in such a state. That's how Tai CHi masters do these unbelievable things eg:


They are in alpha state where you react quicker.

But what is consciousnesses? Who knows - we can do experiments, training etc - but ultimately who knows. Science is strange like that. You start out wanting to answer all these deep questions, but its a chimera - you end up answering something else that is interesting - but does not answer things like consciousness, reality etc.

Thanks
Bill

 

[eloheim]

They discuss Bohmian Mechanics and Many-Worlds in the last section of the paper.

Thanks for the detailed reply, and, of course, you're right about their conclusion in the paper. I was a little confused because they were saying they deal with those two theories at the end. Honestly I still find some of the arguments in that paper hard to follow, not with the math but just with all of the definitions they put forth and how those concepts are used to prove their points. But, anyway, that's another thread for another day, so I'll just leave it at that!

 

[Michael Price]

But it [MWI] does not explain why each observer sees only one alternative, and why different observers usually agree on the alternative seen.

MWI does explain why each observer sees only one outcome, and why all observers agree on all outcomes; this was the whole point of Everett's thesis. Each observer gets split by the outcome event, with each observer, in their own world, seeing just one of the outcomes. It is not an assumption but the consequence of the dynamics of quantum theory. Strictly undergraduate stuff - nothing advanced. And all observers, within each world, agree on their observations for the same reason.

 

[DarMM]

And all observers, within each world, agree on their observations for the same reason.

Renato Renner's group is currently investigating if that is actually the case, apparently different versions of MWI give different answers for this see:
https://www.video.ethz.ch/speakers/...oExmHKbHOT7fw62pctqqZJlz3Q9ZReWi89TiqKAO1UQzk

 

[A. Neumaier]

this was the whole point of Everett's thesis. Each observer gets split by the outcome event, with each observer, in their own world, seeing just one of the outcomes. It is not an assumption but the consequence of the dynamics of quantum theory.

No.

I had studied Everett's thesis. It contains nothing dynamical except for the dynamics of the wave function. All branches are there all the time; just their contents changes. When does a split happen, given only the Schrödinger equation? The Schrödinger equation does not distinguish between times with events and times without events - it does not even have a concept of event. These questions are swept under the carpet.

A proper logical definition would say what an observer is in terms of the deterministic time-dependent wave function (since there is nothing else), and what the stochastic dynamics of each observer's observables is, again in terms of the deterministic wave function. It would specify what it means that some part of the universe (an observer) observes another part of the universe (the observed).

 

[entropy1]

If you send a photon through a beamsplitter, it becomes in superposition of traveling several paths at once. Nevertheless there still is only one photon traveling those paths. So if we measure the photon, we detect it in one place and only once.

So I was wondering if the superposition of worlds leads to such a situation: the measurement is the detection of a quantum, so that it appears only once; in one of the worlds. The other worlds get probability 0 of detecting it.

So in that sense collapse is a result of quantization?

It seems strange to me that the quantum that gets measured suddenly multiplies itself to be able to be detected in ALL worlds!