Why I Prefer the Copenhagen Interpretation(s)

post by dadadarren · 2020-10-31T21:06:02.500Z · LW · GW · 44 comments

This is a link post for https://www.sleepingbeautyproblem.com/9-perspective-based-reasoning-and-quantum-interpretation/

Contents

  Perspectives as Fundamentals
  PBR and the Copenhagen Interpretation
  Reasons to Prefer This Idea
None
44 comments

I have (relatively) recently finished Sean Caroll's Something Deeply Hidden. I'm also under the impression that the Many-Worlds Interpretations (MWI) is favored by many in the lesswrong community. However, the more I think about it, the Copenhagen Interpretation (and other recent variants like RQM and QBism) becomes more appealing. I am not a physicist by any means. Just sharing my amateur thoughts for discussion.

Perspectives as Fundamentals

Thomas Nagel suggests we come to the notion of objectivity through the following steps[1]:

  1. Realize (or postulate) that my perceptions are due to the actions of things upon me, though their effect on me.
  2. Realize (or postulate) the same property that caused actions upon me can also lead to actions on other things.
  3. Realize (or postulate) the property can exist without causing any action at all, which leads to our conception of a “true nature”, independent of any perspective.

Thomas Nagel calls this “the view from nowhere”, Bernard Williams the “absolute conception”. It represents our ordinary understanding of reality which science should strive to describe.

I think maybe Step 3 is one step too far. Step 1 can be regarded as the realization that I am experiencing the world from a given perspective. Step 2 can be seen as the realization that my perspective is just one of many, that it is not particularly special. Both steps put perspectives in a fundamental position. Then Step 3 suggesting objectivity being perspective-independent seems to be quite a jump.

By rejecting that last postulate, properties are no longer fundamentals but derived from actions upon a perspective center. The reality, instead of an absolute conception, needs to be described from some given perspective. Objectivity property is not perspective-independent but rather perspective-invariant. It refers to qualities, methods, and rationales that remains true with perspective changes. I call this idea the Perspective-Based Reasoning (PBR).

PBR and the Copenhagen Interpretation

PBR explains troubling concepts in the CI including the “observer”, “measurements”, and quantum nonlocality. It can be summarized as the following:

  1. It starts with a given perspective and its center.
  2. From here, any information or knowledge about the world is based on actions upon the perspective center.
  3. These actions are the fundamental subject matter. The natures and properties of the actions’ counterparties (other objects) are derived based on them.
  4. The actions are empirically discrete and indeterministic.
  5. Quantum mechanics describes the behavior of these actions. We can, for example, use it to predict future actions based on past ones, or give a description of its counterparties.

What qualifies as a perspective center? The perspective center is a logical starting point. It is either primitively identified (in the case of our natural first-person perspective) or postulated (when we choose to reason from other thing’s perspective). It is never derived. In principle, any physical systems can be proposed as the perspective center E.g. an electron, a Geiger counter, a person, etc.

What is an “observer” in the Copenhagen interpretation? The “observer” is the perspective center. Since the subject matter is actions upon it, quantum mechanics do not describe the observer itself. Therefore the observer cannot be explained by reductionism using quantum mechanics. This does not mean whatever the perspective center is, say a Geiger counter, is physically unanalyzable. It only means the analysis cannot be performed from its own perspective. From a different perspective, e.g. when “the observer” is a person, it can be perfectly explained by quantum mechanics.

What is a “measurement”? A measurement is an action upon the perspective center. Due to their stochastic nature, they cannot be perfectly predicted by quantum mechanics, only probabilistically using the Born rule. It is why the wavefunction is updated when an action happens.

How does it explain quantum non-locality? It interprets quantum mechanics as local. Any notion of non-locality is either caused by not taking a consistent perspective or aperspectival (view from nowhere) reasoning. In the EPR paradox, the supposed non-locality manifests itself when Alice measures the spin of the entangled particle. Bob would instantaneously have a confirmed measurement outcome, no matter the distance separating the two. However, one is an action upon Alice and the other upon Bob. The non-locality is not present from either one’s perspective.

The criticism of CI being utterly instrumental. This criticism is basing on the notion that physics should be describing reality instead of merely predicting experiment outcomes. Yet if reality cannot be described by "a view from nowhere", if it has to be described from a given perspective as suggested by PBR, actions upon the perspective center become the basis of such descriptions.

Reasons to Prefer This Idea

Discarding the aperspectival absolute conception has past success. Relativity rejects absolute space and time. It treats spacetime as perspective-dependent. Nonetheless, we still regard it as an objective theory since it can be applied to different perspectives while remains accurate. It is conceptually similar to the PBR. Even though the description of reality is perspective-dependent, the method outlined by quantum mechanics remains constant for different perspectives, it is still an objective (perspective-invariant) theory.

The absolute conception (or a view from nowhere reality) is a good approximation for the perspective-based interpretation in classical settings. Actions caused by a microscopic entity are few and far between. For any given moment, it might be acting upon a certain object while not affecting another. The effect of perspective choice is prominent as they would give significantly different descriptions. In contrast, a macroscopic object is constantly interacting with its environment. Any practical perspective choice would continually experience numerous actions affected by it. The limiting factor of describing macroscopic objects is not the complete lack of actions but the ability to perceive and process them. As such, different perspectives would give nearly identical descriptions. One can safely use descriptions from other perspectives if it does not wish to capture and analyze the actions upon itself. The absolute, perspective-independent objectivity is an energy-saving shortcut that works well in our daily scales.

It has been suggested that, in terms of basic ingredients, the Many-Worlds Interpretation (MWI) is as simple as it gets. There are no additional hidden variables nor ad hoc collapse mechanisms. So the MWI should be considered the “default” interpretation unless there is new physical evidence suggesting otherwise. The same argument can be made to support this perspective-based interpretation. It actually involves even fewer assumptions than the MWI because it rejects the commonly accepted postulate in Step 3 above. In this sense, the Copenhagen Interpretation should be regarded as the “default” solution instead.

[1] Thomas Nagel put the realization that properties can exist without causing action as the second part of Step 2. So in his original formulation Step 3 only involves the realization of the absolute conception. I changed the grouping for ease of expression.

44 comments

Comments sorted by top scores.

comment by Paul Hayes (paul-hayes) · 2020-11-02T18:43:06.248Z · LW(p) · GW(p)

So the MWI should be considered the “default” interpretation unless there is new physical evidence suggesting otherwise.

 

comment by TAG · 2020-11-01T18:50:14.535Z · LW(p) · GW(p)

Nonlocality is more than than one thing, BEIMT1T. EPR type experiment s are not causally non local, in the sense that you can use them for FTL signalling...but they are also not causally local, in the sense that a local hidden variable theory is viable. They are correlationally non local, because the results are statistically correlated, in a way that can't be explained by local causality.

Replies from: dadadarren
comment by dadadarren · 2020-11-01T22:14:13.731Z · LW(p) · GW(p)

Yes, EPR does not suggest non-locality as FTL signalling. I was saying the correlational non-locality requires first accepting the absolute (a view from nowhere) objectivity. I.E. it is meaningful to think about Alice’s measurement and how it correlates with the space-like separated Bob’s measurement. This reasoning is directly examining the perspective-independent reality. If we reject Step 3, therefore making perspectives fundamental then this correlational non-locality could not exist. For example, from Alice’s perspective, Bob’s measurement can only affect actions upon her when its light cone reaches her, which would definitely happen within the light cone of her own measurement, making any correlation local. I’m under the impression that QBism argues against quantum non-locality with similar logic.

Replies from: TAG
comment by TAG · 2020-11-02T19:20:46.528Z · LW(p) · GW(p)

What correlates are two sets of measurements. It's a statistical property . Alice can note that Bobs observations are correlated with hers , and vice versa -- neither needs a view from nowhere.

Replies from: dadadarren
comment by dadadarren · 2020-11-02T19:52:48.806Z · LW(p) · GW(p)

If not employing a “view from nowhere” then the non-locality has to be formulated with a perspective switch. E.g. Alice considers her own measurement then imagine Bob’s measurement from his perspective (as opposes to find out Bob’s measurement when it affects actions upon her). This non-locality does not apply to any perspective.

Another way to look at it: QM is local. But the conceptual perspective switch is instantaneous/FTL. Thus the non-local statistics.

Replies from: TAG
comment by TAG · 2020-11-02T21:17:49.241Z · LW(p) · GW(p)

Alice considers her own measurement then imagine Bob’s measurement from his perspective.

No. They just need to write down their observations on two strips of paper, and bring them together .

Its an important feature of QM that measurements are pretty much objective and classical once you have made them.

Replies from: dadadarren
comment by dadadarren · 2020-11-03T15:45:30.591Z · LW(p) · GW(p)

They can write their result on two pieces of paper and bring them together. But as long as we treat perspective as fundamental, there still won’t be any non-locality. Because the spins are perceived by actions upon a perspective center. And those actions cannot be spacelike separated. For example, if the two pieces of paper are sent to me, from my perspective all I can say is something like “since I got Alice’s result as spin up, safe to say I will get Bob’s result as spin down.” Non-local correlations like “Alice measured up, so Bob will measure down” is either a perspective switch between the two or a view from nowhere statement.

I agree that measurements are objective once made (in the sense of perspective-invariant not perspective-independent). But Alice and Bob are not analyzing the same set of measurements. Each of them from their respective perspective analyzes actions upon him/herself. Their deduction about the spins would have opposite causal arrows. But as long as we don’t switch perspectives, things would be local.

Replies from: TAG
comment by TAG · 2020-11-09T15:30:30.108Z · LW(p) · GW(p)

They can write their result on two pieces of paper and bring them together. But as long as we treat perspective as fundamental, there still won’t be any non-locality.

Nonocality is more than one thing. The correlation I spoke of continue to exist even if your metaphysics say they can't.

Replies from: dadadarren
comment by dadadarren · 2020-11-10T16:48:09.197Z · LW(p) · GW(p)

That's fair. I think it would be beyond my expertise to criticize other notions of correlational non-locality. I just wanted to point once treating perspectives as fundamental, the non-locality cannot be formulated in QM. Somewhat similar to QBism's treatment of nonlocality.

comment by Signer · 2020-11-01T00:17:22.383Z · LW(p) · GW(p)

It actually involves even fewer assumptions than the MWI because it rejects the commonly accepted postulate in Step 3 above.

I don't see why we need any of the three. MWI assumes only wavefunction. PBR assumes perspective, actions, indeterminism and still wavefuntion, that describes behavior of actions.

Replies from: TAG, dadadarren
comment by TAG · 2020-11-01T18:56:14.844Z · LW(p) · GW(p)

PBR assumes perspective, actions, indeterminism and still wavefuntion

That's a list of nouns rather than of ontological posits. Indeterminism is arguable a non-unicorn, an absence of something. Perspectives are hard to dispute, since we've all got one. Wave functions are explicitly regarded as a kind of map, rather than something out there.

Replies from: Signer
comment by Signer · 2020-11-01T21:45:07.567Z · LW(p) · GW(p)

I assumed at least laws themselves don't change in PBR so we still need some things to be deterministic in addition to indeterministic things. Not sure if it requeries additional ontology, but still seems to result in more complex theory?

Perspectives in MWI can be derived.

I guess "actions behaving according to wavefuntion" in PBR do replace "wavefunction", but would't then laws of behavior became more complex to include that translation from wavefunction to actions?

Replies from: TAG
comment by TAG · 2020-11-01T21:59:28.609Z · LW(p) · GW(p)

I assumed at least laws themselves don’t change in PBR so we still need some things to be deterministic in addition to indeterministic things.

I dont see what you mean. What are the deterministic quantum laws?

but would’t then laws of behavior became more complex to include that translation from wavefunction to actions?

That would just be the projection postulate that everything else uses .

Replies from: Signer
comment by Signer · 2020-11-01T22:22:52.452Z · LW(p) · GW(p)

The results of the laws are indeterministic, but laws themselves are kinda not - you always get the same probabilities. So I figured you would need additional complexity to distinguish between deterministic and indeterministic parts of description of the universe.

Forgive my ignorance, but why do we need projection postulate in MWI?

Replies from: TAG, TAG
comment by TAG · 2020-11-02T20:24:19.971Z · LW(p) · GW(p)

Forgive my ignorance, but why do we need projection postulate in MWI

Because if this you are trying to calculate the probabilities of future events, you need to treat anything you have already observed as probability 1 .or equivalently , discard anything unobserved.

Replies from: Signer
comment by Signer · 2020-11-02T21:38:17.036Z · LW(p) · GW(p)

It sounds useful but I don't see any reason to include the way I treat anything into ontology. That wavefunction is nearly zero in all regions where Born statistics fails is just consequence, not postulate. Similarly you can derive that following Bayes rule will result in largest amount of spicemeasure for states where you know something. Whether you want this or not is purely ethical question and ethics today is as arbitrary as it was yesterday. You might as well only track uncertainty about wavefunction and not specific decoherence-path and decide to minimize worst ignorance or something.

You would need a postulate only if you want there to be some fundamental point-knowledge but there are no point-states in reality - everything is just amplitudes.

Replies from: TAG
comment by TAG · 2020-11-02T22:19:17.113Z · LW(p) · GW(p)

It sounds useful but I don’t see any reason to include the way I treat anything into ontology.

I didn't say it had anything to do with ontology, and in MWI it doesn't . In MWI , you disregard results from other branches that you haven't observed in order to predict future probabilities correctly , but you don't regard them as non existent.

comment by TAG · 2020-11-02T20:08:49.075Z · LW(p) · GW(p)

The results of the laws are indeterministic, but laws themselves are kinda not—you always get the same probabilities. So I figured you would need additional complexity to distinguish between deterministic and indeterministic parts of description of the universe

Under subjective interpretations like rQM, there arent deterministic and indeteministic parts of the universe. You can use Schrödinger's equation to model a part of the universe , and that will work until it stops being isolated -- until it interacts with something not in the model. All models are ultimately indeteministic because they are always based on incomplete information. And the process by which this becomes apparent, by which the limited model is invalidated, isn't anything special .

Observer 01 can model system A just fine until it interacts with system B, which they don't know anything about. If observer O2 is more knowledgeable , they might be able to model the AB interaction using the SWE.

comment by dadadarren · 2020-11-01T23:32:12.674Z · LW(p) · GW(p)

MWI suggests wavefuntion describes the objective world or wavefuntion is the objective world. That in itself assumes objectivity is perspective-independent. I.e. we can think about reality with a “view from nowhere”. I am arguing that is an assumption, that epistemically speaking, perspectives and actions are more fundamental than reality as an absolute conception.

Replies from: Signer
comment by Signer · 2020-11-02T00:14:28.752Z · LW(p) · GW(p)

Right, I confused epistemic assumptions with ontological assumptions. But does minimizing epistemic assumptions even makes sense? I mean we don't start with PBR anyway - we start with what happened to be in our brains. So what's the point then in selecting description of the universe that is epistemically nearest to our starting state as opposed to the least complex one? I guess it would be interesting if we actually could reach PBR QM without ever invoking complexity minimization...

Replies from: dadadarren
comment by dadadarren · 2020-11-02T17:11:05.075Z · LW(p) · GW(p)

I definitely see the value of choosing the least complex theory or Occam’s Razor. The problem is that it works really well in hindsight. But before the debate is settled it is hard to measure which theory is the simpler one.

The appeal of MWI is that it gives a very coherent explanation of quantum phenomenons (maybe not Born rule) without assuming anything extra. There is no additional collapse, there are no extra hidden variables. The existence of parallel worlds that many people find uncomfortable is not its assumption but the result of vigorous logical deductions. I would be lying by saying I don’t see the simplicity and beauty of it. However, I just want to point out MWI does need to assume perspective-independent objectivity. Which CI could (and I think should) go without. If Thomas Nagel’s steps are valid, then it can be argued that CI is less complex. But again, complexity is hard to compare, I have no problem if others find MWI simpler.

I also think we do start with perspective-based reasoning. Even thinking about what happens within my brain requires “a view from nowhere” (assuming perspective-independent reality, step 3), or an outsider’s perspective (step 2).

Replies from: TAG
comment by TAG · 2020-11-02T18:01:45.855Z · LW(p) · GW(p)

The existence of parallel worlds that many people find uncomfortable is not its assumption but the result of vigorous logical deductions

The existence of decoherent worlds is pretty hard to deduce from the SWE alone -- one pure state evolves into another. And a pure state doesn't decompose into a set of objective worlds without a preferred basis .

Replies from: dadadarren
comment by dadadarren · 2020-11-02T18:28:39.753Z · LW(p) · GW(p)

I think you are right. Preferred basis is another problem.

comment by ike · 2020-10-31T22:41:55.117Z · LW(p) · GW(p)

Copenhagen requires a collapse, you can't get there with fewer assumptions than MWI. At best you have agnosticity between MWI and alternate interpretations.

Replies from: dadadarren, TAG, shminux
comment by dadadarren · 2020-11-02T17:32:58.102Z · LW(p) · GW(p)

Maybe some variation of CI involves collapse, but strictly-speaking bare bone CI does not require it. This is typically not discussed by supporters of other interpretations. Even Sean Carroll says according to CI the world follows two distinct rules, one when you are not looking, a different one when you are making a measurement (collapse). However, that is not necessarily CI. For example, as PBR shows, there is only one rule in CI, which describes the behavior of actions. By rejecting “the view from nowhere”, there is no sense in examining how the world behaves when it's not affecting the perspective center. Here the wavefuntion is epistemic rather than ontic. There is no physical collapse needed.

Replies from: ike
comment by ike · 2020-11-02T17:52:39.903Z · LW(p) · GW(p)

Can you explain how you end up with fewer assumptions?

Under your terminology, I believe I can formulate MWI epistemically with the same number of assumptions, and if formalized I think MWI comes out slightly simpler. 

Replies from: dadadarren
comment by dadadarren · 2020-11-02T19:34:43.466Z · LW(p) · GW(p)

Because MWI needs perspective-independent objectivity, that it is meaningful to describe reality with “a view from nowhere”, as the universal wave function does. So it needs to accept the postulate in Step 3. CI could (and I argue should) do without. No matter how commonly accepted it is, Step 3 is still an assumption.

Replies from: ike
comment by ike · 2020-11-02T22:32:11.494Z · LW(p) · GW(p)

I deny that MWI requires that. In fact, all three of your postulates are incoherent, and I believe in a form of MWI that doesn't require any of them. 

Replies from: dadadarren
comment by dadadarren · 2020-11-03T15:43:21.466Z · LW(p) · GW(p)

Fair enough.

comment by TAG · 2020-11-01T12:25:46.066Z · LW(p) · GW(p)

Copenhagen as opposed to Objective Reduction doesnt require collapse to be an objective process. Unfortunately, most of the QM material on this site conflates the two.

Replies from: ike
comment by ike · 2020-11-01T14:22:31.706Z · LW(p) · GW(p)

If collapse is not objective then in what sense is that theory different than MWI?

Replies from: TAG
comment by TAG · 2020-11-01T15:56:33.197Z · LW(p) · GW(p)

It is not denied that collapse occurs. But it is not confirmed either.

Replies from: ike
comment by ike · 2020-11-01T17:35:58.919Z · LW(p) · GW(p)

So not at all different than MWI, using the exact same assumptions.

Replies from: TAG
comment by TAG · 2020-11-01T18:58:28.108Z · LW(p) · GW(p)

MWI has to deny collapse, because otherwise branching would not occur. (And it has to be realistic about wave functions).

Replies from: ike
comment by ike · 2020-11-01T20:04:26.584Z · LW(p) · GW(p)

Can you be explicit about the set of assumptions you're using for MWI and Copenhagen? I can't make heads or tails of your comments here. Are you arguing that Copenhagen requires fewer assumptions? 

Replies from: TAG
comment by TAG · 2020-11-01T21:30:36.566Z · LW(p) · GW(p)

CI, when not confused with OR, is rather minimal. That's not necessarily a good thing, because, while you don't have to put a lot of assumptions into it, you don't get much clarity about what's actually going on out of it

Replies from: ike
comment by ike · 2020-11-01T23:55:35.995Z · LW(p) · GW(p)

It sounds like your version of CI is broad enough that MWI is a special case of it, or vice verse, or something. 

Replies from: TAG, dadadarren
comment by TAG · 2020-11-02T18:12:03.622Z · LW(p) · GW(p)

Minimal CI is rather broad. That explains how it how it became popular -- as opposed to the Physics community being a bunch of booby heads who need to be lectured about Bayes.

Replies from: ike
comment by ike · 2020-11-02T18:26:16.739Z · LW(p) · GW(p)

Do you have any sources explaining the assumptions that go into this minimal version?  

comment by dadadarren · 2020-11-02T17:44:46.158Z · LW(p) · GW(p)

I think no matter how board CI is can MWI be regarded as a special case of it. The two are very distinct. The MWI is top-down. It describes the world with the universal wavefuntion as objective reality from a view from nowhere. So its problem is explaining our individual experience and experiment results, e.g. the Born rule. The CI is bottom-up, it describes our experiments and measurements. So its problem is to provide a comprehensible explanation of reality. The unresolvable conflict between the two is the special role of the observer. CI recognizes it, MWI rejects it.

Replies from: TAG
comment by TAG · 2020-11-02T17:56:24.765Z · LW(p) · GW(p)

I think it's also important not to confiate CI with consciousness-causes-collapse theories. CI doesn't need to regard observers as ontologically special, it just asserts that macroscopic observers will make classical style , real valued observations

And that QM, nonetheless, isn't a classical theory. At the time if it's formulation, the main rivals were theories that tried to explain away the probablistic nature of QM as statistics applying to ensembles.

Replies from: dadadarren
comment by dadadarren · 2020-11-02T19:23:38.974Z · LW(p) · GW(p)

Yes. The special role of observers does not mean Wigner style consciousness-induces-collapse theories, which I am firmly against. That could potentially lead to some bizarre conclusions such as a dog could collapse the wavefuntion but a cat can’t. CI doesn’t say that (it didn’t say much at all). For example, PBR suggests the perspective center (the observer) can be a person just like it can be a Geiger counter.

That being said, PBR can be regarded as a consciousness-related theory if one pushes it. Our natural first-person perspective can be simple treated as primitively given or fundamental. But if there has to be an explanation to this perspective, “why is it apparent that I am the perspective center?”, then the answer ought to be because the only subjectivity available is of this particular human/physical system. By postulation, we can choose to reason from other perspectives (e.g. a Geiger counter’s). Now asking why the Geiger counter is the perspective center would potentially point to a subjectivity of its own. Granted, it does not suggest its subjective experience has got anything in common with humans’. Yet that in essence would lead to panpsychism.

Of course, all this has got nothing to do with actions upon the perspective center or anything physically detectable. So I don’t think they are related to the discussion of QM. Just some random thoughts.

comment by shminux · 2020-10-31T23:54:46.532Z · LW(p) · GW(p)

Every calculation requires a projection postulate, there is no way around it.

Replies from: TAG
comment by TAG · 2020-11-01T18:57:18.504Z · LW(p) · GW(p)

Projection:map :: Collapse: territory.