Why is there no consensus about the meaning of probability in MWI?

[jbergman]

You don't do that. It is a gross misrepresentation of how MWI works. For a start you don't strictly count "worlds" at all. You count orthogonal microstates. And you don't introduce "additional states to give the needed number of worlds to get the right probabilities". You introduce additional decoherence until the number of microstates is big enough for the next step in the argument to be valid. It's not even a mathematical sleight-of-hand. It is a physical process which continues indefinitely but is complete, for all practical purposes, in less than an attosecond for ordinary objects.

I don't think it can be a gross misrepresentation because I've already said I don't follow this step in the argument for assigning probabilities in Zurek's paper. Also, what you described sounds exactly what I was attempting to describe with possibly cruder wording. I was planning to start a separate thread to discuss this on more detail.

I think it would also be helpful to actually spell this out clearly with an example if you feel you have a grasp on this step.

 

[PeroK]

But no observer can ever measure this. So on this view nobody ever measures the probability of outcome A. Which makes this definition useless.

Let's assume MWI. We are going to do a spin 1/2 measurement. You claim that:

The probability of getting spin up is 1. And the probability of getting spin down is 1.

We do the experiment and get spin up. What happened to the certainty we would also get spin down?

If that's all QM was about, then it's hard to see that MWI fails on some definitional technicality. Whether you allow that to be called a probability or not doesn't matter. In that simple case MWI reproduces the results of QM - including what appear to be probabilities.

Moreover, if MWI could (by some counting mechanism or otherwise) reproduce the Born rule, then it would predict the outcome of all QM experiments. And, likewise, whether you allow the term probability to encompass this mechanism is not relevant.

 

[pines-demon]

There is a book called, "Many Worlds?: Everett, Quantum Theory, & Reality" in which a whole section is for critical replies, and his essay is included titled 'Probability in the Everett Picture', which you can find some of their talks here if you'd rather listen: https://vimeo.com/user1742588 (David Albert's own talk on probability is here: )

Or, check out https://philsci-archive.pitt.edu/22650/ page 15 he gives citations for both defenders, and critics.

Thanks for this! Here is a better point to start this thread from Wallace*** discussing MWI:

The probability problem remains largely intact when decoherence is considered, and has been the main locus of controversy about the Everett interpretation in 21st-century discussions. But decoherence does clarify what the conditions are by which the mod-squared amplitudes have the right formal properties to act as probabilities, in particular when they can be used to define transition-probabilities in a consistent way. (In the ‘consistent histories’ [28, 29]or ‘decoherent histories’ [22, 30] approach to decoherence, this so-called ‘consistency condition’ is taken as a formal definition of decoherence.) And so it shows that the problem is essentially philosophical: given that some quantity in our mathematically-stated theory has the right mathematical properties to count as probability, what more needs to be said to justify the claim that it is probability. A range of arguments has been advanced as to why there is indeed a problem here; defenders of the Everett interpretation have variously argued
(i) that formal properties are sufficient and that any further issue is part of the general problem of understanding probability and is not specific to this context [31, 32, 33],[17, ch.4], or
(ii) that the Born rule can be derived in unitary quantum mechanics by various frequency-based [34], symmetry-based [17, ch.4], decision-theory-based [35, 36, 17], entanglement-based [37] or other [38, 39, 40]methods. (For criticism see, e.g., [41, 42, 43]).

Sadly it does not discuss much beyond that... I will try to watch the video or check Ref.s 42–43.

 

[gentzen]

Here is a better point to start this thread from Albert discussing MWI:

That was a quote from David Wallace's paper, not from Albert.

 

[kered rettop]

We do the experiment and get spin up. What happened to the certainty we would also get spin down?

Obviously, in MWI, there is a spin-down world which is not accessible to us. I was puzzled how this leads into the rest of your post. Do you just mean that, since both outcomes do occur and the observer's experience is always of just one outcome, you need something to explain the stochasticism?

If that's what you meant, I think it needs to be stated, otherwise thickos like me will struggle.

 

[Fra]

Indeed. Which is the root problem I'm trying to unravel here - there seems to be no agreed meaning. I'm afraid I tend to say "If it looks like probability and quacks like probability then is probably is probability" but others seem to insist on a more restrictive definition.

Mathematically if a measure satisfies certain rules, it can be said to be a probability.

But the general issue in physics is to attach the mathematics of probability to something that can be quantified in terms of observations (ie samplings) or placing bets (actions taken).

There is several ways to do this, and none is more correct than the other I think, but the question is what we can to USE "probability" for?

I like to think of it as having two purposes

descriptive - this means probability is usually associated with a relative frequency of a hypothetical infinite repeat of some identiacally prepared experiment. The challenge with this, is that the relation to "reality" relies on hypothetical infinite repeats of processes, something that is realistic for some atomic scale experiments, but not realistic in the most general case. So to "verify" if a probability is "correct" you rely on a fictive infinite statistics. Still what is the USE for a descriptive probability, unless we think that a description of the past, gives as an edge into the future! Here the problem is that any real past, is always truncated.

normative - the probability is associated with some ideal odds, one how to play your bets. So this is not a statement about a remote fictive ensemble of futures, it is a statment about our best guess of the future. The challenge with this is that there is not "hard" way to verify if such a "subjective" probability is "correct". Instead all you can to is - enter the game - and one idea is that, the agents that up with good guesses will stay in the game; with time, agents making "bad" guesses are unlikely to stay in business. But there is still the challenge of how to define the set "possible" normative probabilities, and how we can bring order in a subjective mess. If one likes the normative interprettaion AND the frequentists interpretaion, I think they can conceptaully be combined if you count not hypothetical futures, but pieces of evidence (that exists now), that together defines a normative measure. Wether this is consistent with the frequency of possible futures in a repeat scenario, is irrelevant to the justification of the normative measure. It's purpose is not to predict the future, but to predict the actions of the agent.

Both perspectives have their own "problems", and this seems not specific to mwi though. I don't know what mwi adds to all this though? Does mwi provide any conceptual handles on something? Not having a clear notion of what the probability measure is, seems to be a far bigger problem than the issue of wether borns rule is postulated or explained in terms of other postulates.

/Fredrik

 

[kered rettop]

I think it would also be helpful to actually spell this out clearly with an example if you feel you have a grasp on this step.

Could you be more specific about what you want an example of? I'm not being evasive I'm trying to pin your question down so I can answer it before the last proton decays!

 

[kered rettop]

Both perspectives have their own "problems", and this seems not specific to mwi though. I don't know what mwi adds to all this though? Does mwi provide any conceptual handles on something? Not having a clear notion of what the probability measure is, seems to be a far bigger problem than the issue of wether borns rule is postulated or explained in terms of other postulates.

/Fredrik

Good post. If only everyone agreed! I'm all for telling people not to offload their classical problems onto MWI. They presumably blame MWI because they don't notice the problems until some crazy Many-Worlder starts talking about parallel worlds and multiple copies of "me". And because they've only just seen it - in MWI - it stands to reason that it only afflicts MWI.

 

[PeroK]

Obviously, in MWI, there is a spin-down world which is not accessible to us. I was puzzled how this leads into the rest of your post. Do you just mean that, since both outcomes do occur and the observer's experience is always of just one outcome, you need something to explain the stochasticism?

If that's what you meant, I think it needs to be stated, otherwise thickos like me will struggle.

To clarify what I meant. Both spin up and spin down have a certainty of being part of the wave function after the measurement. But, do not have a certainty of being measured in a particular world.

The branching of the wavefunction and the theory that observations can be made in a meaningful way within a branch, introduces uncertainty - albeit a different type of uncertainty from orthodox, single-world QM.

 

[PeterDonis]

Let's assume MWI. We are going to do a spin 1/2 measurement. You claim that:

The probability of getting spin up is 1. And the probability of getting spin down is 1.

Yes, because both outcomes occur.

We do the experiment and get spin up.

No, we do the experiment and both results occur, each in its own world. If you are using the MWI, any language you use that ignores that is simply wrong.

What happened to the certainty we would also get spin down?

It was incorrectly ignored when you misused ordinary language to make it seem as though only one outcome occurred, when actually both occurred.

 

[PeterDonis]

The branching of the wavefunction and the theory that observations can be made in a meaningful way within a branch, introduces uncertainty

No, it doesn't. The branching is completely deterministic. There is no uncertainty anywhere.

albeit a different type of uncertainty from orthodox, single-world QM.

No, something that should not be called "uncertainty" at all, unlike the case of single-world QM.

 

[PeroK]

No, it doesn't. The branching is completely deterministic. There is no uncertainty anywhere.


No, something that should not be called "uncertainty" at all, unlike the case of single-world QM.

Why not turn the argument round? Assume MWI is the orthodox, deterministic interpretation of QM. Where the Born rule is the problematic counterpart of the measurement problem in collapse interpretations.

Now, the collapse interpretations are ruled out because they involve probabilities. Whereas, orthodox QM involves no probabilities.

 

[PeterDonis]

Why not turn the argument round?

Because no interpretation of QM can rule out another interpretation, since they all make the same experimental predictions. It doesn't matter which one you consider "orthodox"; that's just a matter of historical accident and personal opinion. Even if the MWI had been discovered first, it would still have all the same issues, and other interpretations would still have the same issues they have in our actual world.

But that doesn't change the fact that in order to talk about any interpretation, you have to correctly describe what it says, and you have to avoid making assumptions, especially implicit ones, that are false within the context of that interpretation. An MWI proponent who criticized a collapse interpretation in terms that included an assumption of multiple outcomes would not be making a justified criticism, even if such terms were part of the MWI proponent's ordinary manner of speaking.

 

[gentzen]

Because no interpretation of QM can rule out another interpretation, since they all make the same experimental predictions.

Well, when I read what you wrote about MWI in this thread, I get the impression that MWI would not make the same predictions as other interpretations, if your statements concerning probability in MWI were accepted.

 

[PeterDonis]

when I read what you wrote about MWI in this thread, I get the impression that MWI would not make the same predictions as other interpretations, if your statements concerning probability in MWI were accepted.

The MWI makes the same predictions, but its explanation of why those predictions are made is very different from other interpretations. My statements are just giving the MWI skeptic's viewpoint on how valid the MWI's explanations are. But of course MWI proponents think they are valid, however outlandish they seem to skeptics.

 

[PeroK]

The MWI makes the same predictions, but its explanation of why those predictions are made is very different from other interpretations. My statements are just giving the MWI skeptic's viewpoint on how valid the MWI's explanations are. But of course MWI proponents think they are valid, however outlandish they seem to skeptics.

If your posts on this thread aren't ruling out MWI on the grounds that probabilities are essential to QM and that MWI cannot under any circumstances produce probabilities, then I don't follow what you are saying?

How can MWI make the same (probabilistic) predictions if it cannot make probabilistic predictions in the first place. It can either make a 50-50 spin up.or spin down prediction or it can't.

Several people on this thread have said they think MWI can make a probabilistic prediction in that simple case. And you have repeatedly and categorically told us we are wrong.

Now, it seems, MWI does make that 50-50 prediction?

 

[PeterDonis]

If your posts on this thread aren't ruling out MWI on the grounds that probabilities are essential to QM and that MWI cannot under any circumstances produce probabilities, then I don't follow what you are saying?

I am trying to correct misstatements about what the MWI says. That is a valid effort even if it cannot rule out the MWI (which can't be done since no QM interpretation can be ruled out at our current state of knowledge--the guidelines for this subforum discuss this). I am also giving skeptical counter arguments to proposed concepts of probability in the MWI, to make clear that those counter arguments exist and are considered valid by at least some portion of the physics community. If nothing else, that is relevant to the original thread topic, which is why there is no consensus about the meaning of probability in the MWI.

 

[PeterDonis]

How can MWI make the same (probabilistic) predictions if it cannot make probabilistic predictions in the first place.

MWI proponents say it can make probabilistic predictions. MWI skeptics say it can't. Neither can convince the other because we are talking about QM interpretation questions that cannot be resolved by experiment. So it all comes down to personal opinions on either side. That is to be expected in discussions in this subforum, as the subforum guidelines make clear.

 

[PeterDonis]

you have repeatedly and categorically told us we are wrong.

And you can simply choose not to accept my counter arguments. There's no way I can convince you if you do that, because there are no experiments I can appeal to to decide the issue. Nor can you convince me that I'm wrong, for the same reason. Again, this is the QM interpretations subforum, where we can't resolve disputes by experiment the way we can in other subforums.

 

[pines-demon]

Again, this is the QM interpretations subforum, where we can't resolve disputes by experiment the way we can in other subforums.

What do you think is the best way to handle this kind of discussions?

 

[PeterDonis]

What do you think is the best way to handle this kind of discussions?

To not expect to reach a resolution.

 

[kered rettop]

To not expect to reach a resolution.

Was that supposed to be helpful or a joke? If it was a joke then thank you, ha-ha. If it was intended to be helpful then no it wasn't. Nobody expects a resolution of the issues: that would be insanely optimistic. But I still nurse a faint hope of discovering what the issues are. Or what they might be.

 

[PeterDonis]

Was that supposed to be helpful or a joke?

It was a direct response to the question I was asked (not by you). It was not a joke, and was not intended to be either helpful or unhelpful, just a factual response.

Nobody expects a resolution of the issues

Perhaps you don't, but I'm not so sure that's true of the poster I was responding to. That's why I felt it necessary to make the point.

 

[pines-demon]

Perhaps you don't, but I'm not so sure that's true of the poster I was responding to. That's why I felt it necessary to make the point.

To clarify I agree with @kered rettop that nobody is expecting a resolution of the interpretation of QM (or to any open problem in physics) in a forum. I also think that the point is not to convince you or anybody of anything against what they think already, but to clarify what they mean.

I also felt your answer was unhelpful as a mentor, you said "not to expect a resolution" but did not explain how to find a conclusion. You can provide, even if it is in two words, what happens with this kind of threads. I hope we can find together what is the "consensus about the [lack of] meaning of probability in MWI?" (see title), and by consensus clearly is not the consensus of PF.

 

[PeterDonis]

I also felt your answer was unhelpful as a mentor

I'm not sure what else you expect me as a mentor to tell you, or to do. See below.

you said "not to expect a resolution" but did not explain how to find a conclusion.

A conclusion to what? A conclusion is a resolution. There might not be one.

You can provide, even if it is in two words, what happens with this kind of threads.

I can't do it in two words, but I've already done it with more than two words, in my post #229, which you quoted.

I hope we can find together what is the "consensus about the [lack of] meaning of probability in MWI?" (see title)

Again, there might not be one, either here at PF or in the physics community in general. It is certainly not in my power, as a mentor or otherwise, to just create one.

 

[PeterDonis]

nobody is expecting a resolution of the interpretation of QM

@kered rettop didn't just say that. He said he wasn't expecting a resolution of "the issues", which I take to mean the issues involved with finding a meaningful concept of probability in the MWI.

 

[PeterDonis]

I still nurse a faint hope of discovering what the issues are.

I'm not sure what more can be added to what has already been said in this now 7 page long thread. If anyone else has anything of substance to add, they are more than welcome.

 

[kered rettop]

I'm not sure what more can be added to what has already been said in this now 7 page long thread. If anyone else has anything of substance to add, they are more than welcome.

Well I have a lot to say. As the OP I didn't want to get too involved too early.

 

[PeterDonis]

Well I have a lot to say. As the OP I didn't want to get too involved too early.

I think 7 pages in is more than sufficient to avoid any hint of that.

 

[pines-demon]

I'm not sure what else you expect me as a mentor to tell you, or to do. See below.

Where do I find the responsibilities of a mentor? I was hoping a mentor could be asked how to handle a thread. Saying we cannot resolve it does not amount to anything. There might be examples of other conversations in this subforum that ended successfully, maybe with discussion of all points of view of the physics community on the difficulty of a topic.

 

[pines-demon]

@kered rettop didn't just say that. He said he wasn't expecting a resolution of "the issues", which I take to mean the issues involved with finding a meaningful concept of probability in the MWI.

The point being?

 

[PeterDonis]

Where do I find the responsibilities of a mentor? I was hoping a mentor could be asked how to handle a thread.

As a moderator (which is a mentor's job), the only option I have for "handling" a thread in general is to close it. Are you asking me to close it?

 

[PeterDonis]

There might be examples of other conversations in this subforum that ended successfully

What's your definition of "success"? Do people have to end up agreeing? Or just agreeing to disagree? Or...what?

 

[PeterDonis]

The point being?

The point being that you seem to be expecting something that the OP of the thread is not expecting. What's more, as far as I can tell at present, what you are expecting is beyond my or anyone's power to give you, since nobody can force a "resolution" to a topic that can't be resolved by experiment.

 

[Fra]

that would be insanely optimistic.

Nothing wrong with that though, a little insanity is likely even a requirement for work in qm foundations and unification.

/Fredrik