Does quantum mechanics need a conscious observer?

[atyy]

Textbooks like Landau and Lifshitz avoid the observer having to be conscious, by saying a measurement is the interaction of a quantum system with classical apparatus to produce a classical result.

However, there is an analogy that is often drawn that wave function collapse is like statistical updating or Bayesian conditioning. If one accepts this analogy, doesn't the observer have to be conscious, since it is the observer's knowledge that is being updated?

 

[StevieTNZ]

Some people (including physicists) advocate that view that it is indeed a conscious observer that performs the measurement. I personally subscribe to that interpretation.

However, its very hard to tell which interpretation is correct with the lack of experimental means of testing which interpretation is correct over others. It all depends whether standard QM is correct after all, and doesn't need modifying (i.e. conforming to GRW theory, which includes collapse as part of its equations; effectively a physical collapse of the wave function).

 

[bhobba]

The question is - does QM NEED a conscious observer.

The obvious answer is no - since many many interpretations such as my ignorance ensemble interpretation do not require it.

Remember this idea came out of Von-Neumann's famous textbook when he showed the Von-Neumann cut can be placed anywhere. But when you trace it back the only place different is human consciousness - so that's where he placed it. But nowadays with decoherence we know a place that is different - just after decoherence. Place it there and all this silly conscious observer created rubbish goes out the door.

The other high priest of conscious causes collapse, Wigner, when he heard of some early work by Zurek on decoherence, did a total 180% about face and totally rejected it.

So the answer is - why bother - simply place it just after decoherence. I think SOME (not all, but some) people that hold to it have a penchant for deliberately wanting the world to be weirder than it actually is. Also some new age touchy feely types use it as justification for whacko silly stuff like What The Bleep Do We Know Anyway that would undoubtedly make a confirmed rationalist like Von-Neumann wince.

BTW - I hasten to add - its a valid interpretation some people hold to even today - but they are well in the minority.

Thanks
Bill

 

[atyy]

The question is - does QM NEED a conscious observer.

The obvious answer is no - since many many interpretation such as my ignorance ensemble interpretation do not require it.

Remember this idea came out of Von-Neumann's famous textbook when he showed the Von-Neumann cut can be placed anywhere. But when you trace it back the only place different is human consciousness - so that's where he placed it. But nowadays with decoherence we know a place that is different - just after decoherence. Place it there and all this silly conscious observer created rubbish goes out the door.

The other high priest of conscious causes collapse, Wigner, when he heard of some early work by Zurek on decoherence, did a total 180% about faceand totally rejected it.

So the answer is - why bother - simply place it just after decoherence.

Thanks
Bill

I should have said, "Does QM need a conscious observer if wave function collapse is interpreted as a kind of statistical updating" If it is statistical updating of knowledge, doesn't it need a conscious observer, since only such observers have knowledge?

 

[bhobba]

I should have said, "Does QM need a conscious observer if wave function collapse is interpreted as a kind of statistical updating" If it is statistical updating of knowledge, doesn't it need a conscious observer, since only such observers have knowledge?

Hmmm.

Expressed that way it's a more interesting question.

The same could be said of probability theory in general.

Personally I think no. Remember it's a model - and models are something only understood by rational entities. Its part of the model that this updating occurs - its purely a theoretical construct.

Thanks
Bill

 

[.Scott]

Textbooks like Landau and Lifshitz avoid the observer having to be conscious, by saying a measurement is the interaction of a quantum system with classical apparatus to produce a classical result.

However, there is an analogy that is often drawn that wave function collapse is like statistical updating or Bayesian conditioning. If one accepts this analogy, doesn't the observer have to be conscious, since it is the observer's knowledge that is being updated?

If you require a "conscious observer", you would then need to define what exactly is a "conscious observer".

 

[Jano L.]

I should have said, "Does QM need a conscious observer if wave function collapse is interpreted as a kind of statistical updating" If it is statistical updating of knowledge, doesn't it need a conscious observer, since only such observers have knowledge?

As Bhobba says, it is the same as with probability in general. But the answer is no. I am sure that there are cases where computers take data as they come in and use them to calculate corresponding probabilities. One could program computer that monitors results of sequential Stern-Gerlach experiment and calculate the same things physicist would do.

 

[Jano L.]

If you require a "conscious observer", you would then need to define what exactly is a "conscious observer".

Good point! I guess most people would say machine is conscious if it were capable to respond in an impressive way to inputs they provide. Perhaps machine capable of doing calculations with $\psi$ and predicting results of atomic experiments would qualify as conscious at least in some technical sense:-)

 

[StevieTNZ]

I define conscious the easiest way possible: what I perceive, hear, touch, in my mind. Without getting too philosophical on what those terms mean (especially 'mind'). What I'm seeing right now is in my mind, for example, is the easiest approach for me.

 

[atyy]

Hmmm.

Expressed that way it's a more interesting question.

The same could be said of probability theory in general.

Personally I think no. Remember it's a model - and models are something only understood by rational entities. Its part of the model that this updating occurs - its purely a theoretical construct.

Thanks
Bill

As Bhobba says, it is the same as with probability in general. But the answer is no. I am sure that there are cases where computers take data as they come in and use them to calculate corresponding probabilities. One could program computer that monitors results of sequential Stern-Gerlach experiment and calculate the same things physicist would do.

If you require a "conscious observer", you would then need to define what exactly is a "conscious observer".

Good point! I guess most people would say machine is conscious if it were capable to respond in an impressive way to inputs they provide. Perhaps machine capable of doing calculations with $\psi$ and predicting results of atomic experiments would qualify as conscious at least in some technical sense:-)

How about if I asked whether the rational entity must be in the model itself? Say if I distinguish two layers of model: (1) reality and (2) the model of reality. In a model of such as Newtonian mechanics, it does seem that I must put a rational agent like myself in (1), but not necessarily in (2). So could I refine the question to ask, that while all models require a rational agent in (1), do Bayesian conditioning models of the wave function collapse require that the rational agent also be in (2)?

As you say, it is really a question about probability. Based on a Bayesian interpretation, I believe Scott Aaronson gives one definition of a "rational person" http://www.scottaaronson.com/blog/?p=822: "As for what “rational” means, all we’ll need to know is that a rational person can never assign a probability of 0 to something that will actually happen.'

Does a frequentist interpretation of probability (eg. Durr, Goldstein, Zanghi's version of dBB http://arxiv.org/abs/quant-ph/0308039) avoid a rational agent in (2)? Actually, is there even such a thing as frequentist updating? Or is frequentist updating necessarily irrational?

 

[strangerep]

QM is not only about the association between a (modified) notion of probability and squared matrix elements.

It's also about spectra of the observables.

Consider angular momentum. If the generators are not represented as Hermitian operators on a Hilbert space, then you can't derive the well-known half-integer spectrum.

So one might as well ask: "is an electron still an electron if there's no conscious observer?"

Or: does the hydrogen atom collapse (electron falling into the nucleus) if there's no conscious observer"?

(Duh)

 

[Jano L.]

... I believe Scott Aaronson gives one definition of a "rational person" http://www.scottaaronson.com/blog/?p=822: "As for what “rational” means, all we’ll need to know is that a rational person can never assign a probability of 0 to something that will actually happen.'

That is a strange definition. There were events that had predicted probabilities 0 and still happened. One king won an island in the game of dice, because he managed to throw 7. The probability of such event is rationally 0, but nevertheless it happened (the die split into two pieces while rolling and these ended up on 3 and 4). Still, the prediction that ' "the die will end up showing 7" has probability 0' was rational.

The same thing applies in any application of probability theory - events that we deem irrelevant have probability 0, but no theorizing can prevent them from happening.

I'm sorry, but I do not understand your two layers. I do not see how introduction of rational entity into theoretical scheme helps in any way.

 

[atyy]

@Strangerep and @Jano L, the question is in the context of the many attempts to understand collapse as a form of Bayesian conditioning. It's a very old idea. bhobba often mentions it casually - it was really his common use of the analogy between collapse and statistical updating, as well as his saying that the wave function doesn't need a conscious observer ("rational" might have been a better word choice) that lay behind my question, since they weren't obviously compatible to me. The analogy is hinted at in Cohen-Tannoudji, Diu and Laloe's text, and it is mentioned in the text of Wiseman and Milburn.

Some recent attempts include:

http://arxiv.org/abs/quant-ph/0106133
Quantum probabilities as Bayesian probabilities
Carlton M. Caves, Christopher A. Fuchs, Ruediger Schack
Phys. Rev. A 65, 022305 (2002)

http://arxiv.org/abs/1107.5849
Towards a Formulation of Quantum Theory as a Causally Neutral Theory of Bayesian Inference
M. S. Leifer, R. W. Spekkens
Phys. Rev. A 88, 052130 (2013)

Bayesian coherence is a formal theory of rationality, and one of the requirements is that the prior must assign non-zero probability to the truth, which is why I believe Aaronson used that as his definition of a rational agent - one is using a Bayesian framework.

A similar line of query is what the nature of hidden variables behind quantum mechanics can be. Depending on the relationship between the hidden variables, the wave function can be interpreted as "psi-ontic" or "psi-epistemic", where I think the latter means something like "belief". One popular set of definitions is in:

http://arxiv.org/abs/0706.2661
Einstein, incompleteness, and the epistemic view of quantum states
Nicholas Harrigan, Robert W. Spekkens
Found. Phys. 40, 125 (2010)

Although maximally psi-epistemic interpretations seem to have been ruled out http://arxiv.org/abs/1208.5132 , the existence of psi-epistemic interpretations is said to have been proved in:

http://arxiv.org/abs/1201.6554
Distinct Quantum States Can Be Compatible with a Single State of Reality
Peter G. Lewis, David Jennings, Jonathan Barrett, Terry Rudolph
Phys. Rev. Lett. 109, 150404 (2012)

http://arxiv.org/abs/1303.2834
Psi-Epistemic Theories: The Role of Symmetry
Scott Aaronson, Adam Bouland, Lynn Chua, George Lowther
Phys. Rev. A 88, 032111 (2013)

I don't know whether a psi-epistemic interpretation necessarily requires a Bayesian interpretation, but I believe the search for the possibility of a Bayesian interpretation is a motivation for psi-epsitemic interpretations.

Interestingly, there is a form of many-worlds in which Bayesian probability plays a role:

http://arxiv.org/abs/quant-ph/0312157
Quantum Probability from Subjective Likelihood: improving on Deutsch's proof of the probability rule
David Wallace

 

[strangerep]

the question is really in the context of the many attempts to understand collapse as a form of Bayesian conditioning

The notion of collapse is not necessary to derive the correct angular momentum spectrum.

But... I guess you already know that I consider the whole wave-function-collapse-to-eigenstate-after-measurement thing to be a load of nonsense.

 

[atyy]

The notion of collapse is not necessary to derive the correct angular momentum spectrum.

But... I guess you already know that I consider the whole wave-function-collapse-to-eigenstate-after-measurement thing to be a load of nonsense.

Assuming your view about collapse is more or less in line with Ballentine's book, let's agree to disagree about that specific issue in this thread.

 

[martinbn]

Why only conscious observers can have knowledge, and not any classical or classical-like object? A classical measuring device records the result of the measurement and that is its knowledge.

 

[atyy]

Why only conscious observers can have knowledge, and not any classical or classical-like object? A classical measuring device records the result of the measurement and that is its knowledge.

If I understand .Scott and Jano L correctly, a conscious observer is probably a classical object. So in Landau and Lifshitz, it is the irreversible reading on the classical measuring apparatus that is intuitively specified. In the Bayesian analogy, it is the rational agent that is intuitively specified (let's use "rational" instead of "conscious", since that seems closer to standard Bayesian terminology). So the question then is, what would be a more precise specification of "rational" agent? Presumably it would do "Bayesian" updating, and have a "rational" prior. In the framework of Bayesian coherence, I think it can be shown that any prior which assigns non-zero probability to the truth will converge to the truth, and so is rational (at least that's how I understand Scott Aaronson's definition of a rational agent http://www.scottaaronson.com/blog/?p=822).

 

[StevieTNZ]

Why only conscious observers can have knowledge, and not any classical or classical-like object? A classical measuring device records the result of the measurement and that is its knowledge.

Everything is quantum, though (in principle). So in reality there is no classical apparatus.

I could go further to state that the fundamental equation of QM, the Schrodinger equation, only predicts probabilities, and that you never end up with a definite state of knowledge if there exists Classical devices - if a quantum system interacts with a Classical device.

 

[atyy]

Everything is quantum, though (in principle). So in reality there is no classical apparatus.

But if everything is quantum, eg. many-worlds, then there is no wave function collapse. However, apparent collapse must still hold for an observer in many-worlds. In some forms of many-worlds that observer uses decision-theoretic principles or Bayesian inference. So in that context we can talk about the question: if collapse is analogous to Bayesian updating, must quantum mechanics contain a "Bayes-rational" agent?

http://arxiv.org/abs/0906.2718
A formal proof of the Born rule from decision-theoretic assumptions
David Wallace

 

[StevieTNZ]

But if everything is quantum, eg. many-worlds, then there is no wave function collapse.

Exactly. But it doesn't stop us from seeing a definite reality.

So maybe we need to re-think what 'wave function collapse' actually means if everything is quantum. Is there a wave function collapse per say that causes a definite reality to arise, or does the Schrodinger equation still hold when we see a definite reality - but we're just seeing one component of the superposition, where the superposition principle still holds true?

 

[atyy]

Exactly. But it doesn't stop us from seeing a definite reality.

So maybe we need to re-think what 'wave function collapse' actually means if everything is quantum. Is there a wave function collapse per say that causes a definite reality to arise, or does the Schrodinger equation still hold when we see a definite reality - but we're just seeing one component of the superposition, where the superposition principle still holds true?

That's a question for a different thread. I am only interested here in interpretations in which collapse or apparent collapse are considered analogous to Bayesian updating (or any other form of statistical updating that may exist).

 

[StevieTNZ]

That's a question for a different thread. I am only interested here in interpretations in which collapse or apparent collapse are considered analogous to Bayesian updating (or any other form of statistical updating that may exist).

The following book may be of help to you - "Quantum Information Theory and the Foundations of Quantum Mechanics" (http://ukcatalogue.oup.com/product/9780199296460.do). I've inter-library loaned it and should actually be at my doorstep right now.

 

[bhobba]

So maybe we need to re-think what 'wave function collapse' actually means if everything is quantum.

That's the big advantage of my ignorance ensemble interpretation. There is no assumed classical world - everything is quantum from the get-go. An observation occurs once decoherence has occurred. This means, for example, when a few stray photons, decoheres a dust particle to give it a definite position (ie is in a mixed state) an observation has occurred, even without an actual observational apparatus, human consciousnesses etc etc. One puts the Von-Neumann cut just after decoherence. It leads to a very common sense world, and everything is quantum.

This is the exact reason why I prefer this interpretation over Ballentines. It'is a slight variant, but to me conceptually much cleaner.

Thanks
Bill

 

[atyy]

That's the big advantage of my ignorance ensemble interpretation. There is no assumed classical world - everything is quantum from the get-go. An observation occurs once decoherence has occurred. This means, for example, when a few stray photons, decoheres a dust particle to give it a definite position (ie is in a mixed state) an observation has occurred, even without an actual observational apparatus, human consciousnesses etc etc. One puts the Von-Neumann cut just after decoherence. It leads to a very common sense world, and everything is quantum.

This is the exact reason why I prefer this interpretation over Ballentines. It'is a slight variant, but to me conceptually much cleaner.

Thanks
Bill

Let's discuss this elsewhere. I'd like to stay on topic with interpretations in which collapse is analogous to Bayesian updating, or some form of statistical updating. We can include Deutsch/Wallace's decision theoretic many-worlds.

My understanding so far is that Bayesian updating does postulate that a "rational" agent does the updating.

 

[.Scott]

If you require a "conscious observer", you would then need to define what exactly is a "conscious observer".

My thinking on this is what most people consider as "conscious observer" is someone or something such as themselves that has an "experience of the universe". But clearly, our experience of the world is driven by a Darwinian-driven design that is unrelated to rules of particle physics. Or are we to suppose that for billions of years particle physics was completely different before a human-like "conscious observer" evolved?

So if you are going to require a "conscious observer" (which I really don't), you need to define "conscious observer" as something that's plentiful throughout the universe and throughout most of time. Since I view "human consciousness" as a special case of a basic physical property "consciousness", I wouldn't have a problem with that sort of definition.

But to address the broader question, whether something like a "conscious observer" is required, I doubt it on these grounds: The term "conscious observer" is ill defined, but is clearly a reference to part of the human experience. The main thrust of requiring a "conscious observer" is to eliminate the possibility that a potential moment in the life of a "conscious observer", such as ourselves, could ever be "erased" by some sort of interference pattern or delayed QM affect. That seems to me to be way too human-centric to be a fundamental law of physics. It's not necessarily wrong, but it is wrong outlook.

For the purpose of turning outlooks, let me describe this possible implementation of the human brain: Within each of our skulls there are scores of "consciousness circuits" each competing to come up with a candidate "intention". All are scored by potential value. The one with the highest score "gets our attention" in the sense that it is logged in our memory - and potentially has a direct effect on our behavior. All the other those other consciousness circuits continue on in silence monitoring some other little part of our world.

 

[atyy]

@.Scott, I am only interested in definitions of conscious or rational agents appropriate to interpretations of quantum theory in which the collapse of the wave function is analogous to Bayesian updating of knowledge.

 

[.Scott]

@.Scott, I am only interested in definitions of conscious or rational agents appropriate to interpretations of quantum theory in which the collapse of the wave function is analogous to Bayesian updating of knowledge.

Are you expecting that the definition will have some sort of connection to the common definition/usage of "conscious" and "rational". If not, then why not call it a "collapsing agent" or "collagent" to make it clear that it is an entirely new construct?

 

[atyy]

Are you expecting that the definition will have some sort of connection to the common definition/usage of "conscious" and "rational". If not, then why not call it a "collapsing agent" or "collagent" to make it clear that it is an entirely new construct?

I didn't want to use "collapsing agent" because there are interpretations in which collapsing agents which are just classical measuring aparatus. I was interested specifically in a interpretations in which collapse occurs because the collapsing agent's knowledge is being updated, eg. as in Bayesian updating. Examples of such attempts at interpretation are given in my post #13 and StevieTNZ's post #22. There appears to be a technical definition of rationality or coherence in Bayesian thinking. I would also be interested in other forms of statistical updating besides Bayesian updating.

Here is an example of Bayesian "coherence" in standard statistical usage:
http://mlg.eng.cam.ac.uk/mlss09/mlss_slides/Jordan_1.pdf.
Are You a Bayesian or a Frequentist?
Michael I. Jordan

Here's an account of developments following one of the papers in #13:
http://arxiv.org/abs/1301.3274
Quantum-Bayesian Coherence: The No-Nonsense Version
Christopher A. Fuchs, Ruediger Schack
Rev. Mod. Phys. 85, 1693–1715 (2013)

 

[Dale]

It has been recommended that we close this thread in order to be coherent with the closure of other similar threads.