Home › Forums › 2015 International Workshop on Quantum Foundations › Retrocausal theories › Retrocausality is intrinsic to quantum mechanics
July 4, 2015 at 8:30 pm #2477
Below I comment on three papers relevant to time symmetry in quantum mechanics.
First, there is my paper “A reasonable thing that just might work”, in the IJQF online John Bell Workshop 2014. The second paragraph of Sect. II of that paper (p. 6) claims that retrocausality is intrinsic to quantum mechanics. Let three observers – Alice, Bob and Jim – share an ensemble of triplets of spin-1/2 particles in the Greenberger, Horne and Zeilinger (GHZ) state. Let Jim have the special role of the “jammer” who chooses whether to put the particles held by Alice and Bob into a product state or an entangled state. Even if Alice’s and Bob’s measurements precede Jim’s by timelike separations, Jim can determine – by his choice of what to measure – whether the correlations measured by Alice and Bob will turn out to violate Bell’s inequality or not.
Somewhat to my surprise, this claim has aroused criticism. I considered devoting a paper to retrocausality in quantum mechanics, until an e-mail from Emil Zemirman alerted me to two closely related papers: A paper by A. Peres, “Delayed choice for entanglement swapping”, J. Mod. Optics 47, 139 (2000), presents a thought experiment, also involving three observers, that is different from mine but is equally retrocausal. And then a paper by X.-s. Ma et al., “Experimental delayed-choice entanglement swapping”, Nat. Phys. 8, 47 (2012), realizes the Peres thought experiment using entangled photons. X.-s. Ma et al. aptly characterize this retrocausality as “quantum steering into the past”.July 5, 2015 at 5:58 pm #2495
thanks for interesting papers.
Regarding the first one, from eq.(4) for the ‘collective’ observables B and B’, it does not seem obvious which kind of measurement of those observables is assumed–on every pair separately [and then calculating the B and B’] or some kind of collective [direct] measurement. Hence my question: whether your conclusions might change due to (a tacit assumption) on contextuality of measuring B and B’?
MiroljubJuly 7, 2015 at 11:09 am #2510
Dear Mirojlub Dugic,
Thank you for your question. I definitely intend for Bob’s measurements of B and B’ to be collective. In the classical limit, which we assume here, there is no contextuality just as there is no complementarity.
DanielJuly 7, 2015 at 5:42 pm #2512
Dear Daniel [if I may?]
could you please describe the effects with certain collective variables such as the macromolecules interferometry experiments? It appears to me that in such experiments there is no dependence of the collective-observable (the molecules’ center-of-mass position) standard-deviation on the number of the constituent particles (atoms in the molecules). If I am correct, does this mean irrelevance of such effects for your theory?
MiroljubJuly 8, 2015 at 8:29 pm #2542
Hi Daniel, I don’t agree with the claim that “retrocausality is intrinsic to QM” — if that means that any viable quantum theory has to involve backwards in time causation. I made some comments about this over in the “Bohm’s theory” forum (somewhat directly in response to your comments above and the earlier paper they refer to, but also as a way of answering Max Schlosshauer’s prompt there):
I will definitely be interested to hear your thoughts.July 9, 2015 at 12:50 am #2543Ken WhartonMember
As you know, while I agree that your arguments are indeed indicative of a hidden, underlying retrocausal account of entanglement experiments, I was one of the people who pushed back on your claim that “retrocausality is intrinsic to quantum mechanics”. As I see it, standard QM has gone out of its way to avoid the need for retrocausality. (Too far, in my opinion!)
Reading these other two papers you cite, it looks like you probably would have also gotten push-back from both Peres and Ma et al. on this point, but for a different reason. The issue Peres raises is that “being entangled” should be thought of as a state of knowledge, not a state of reality. Ma et al. also raise this ontic/epistemic issue, and they note that the apparent retrocausation disappears if you don’t think of “entanglement” as a state of reality. If “being entangled” is not a state of reality, their argument effectively goes, then it can hardly be used to judge causation.
I think you have a potential case to make against this argument, starting with the question: If entanglement is not a state of reality, then what is? That question puts you right into the hidden-variable ontological framework in which Bell inequalities are surprising, and arguably indicative of retrocausality in their own right.
Still, I didn’t read this as your argument. I read that you were implying that *standard* QM is retrocausal. But for this to go through, there has to be something ontological about the past that is different for different future decisions. In standard QM, the full wavefunction is the implied ontology, and no matter what future measurements are made, the standard wavefunction *before* those measurements is always the same. So there’s no “variable” on which retrocausality could even potentially act.
I understand that you’re looking at Alice and Bob’s measurement outcomes, not the wavefunction, but in standard QM those outcomes result from the wavefunction, which is in turn unchanged by anything a future Jim might do. In the standard QM account, Jim’s remaining piece of the wavefunction tells him how those earlier outcomes were correlated (if Jim makes the proper measurement to extract the information). Standard QM implies that Jim’s not *causing* an earlier correlation, he’s being *informed* about it, at least if he chooses to make the right measurement.
Like you, I think there’s a more natural retrocausal resolution of these sorts of experiments (esp. given the similarity to the case where Jim measures first). But without some hidden past variable that can be “caused”, I guess I just don’t see how *standard* QM has any room for retrocausation.
KenJuly 9, 2015 at 11:41 pm #2587Michael B. HeaneyParticipant
One could argue that the equivalence of the Stueckelberg-Feynman interpretation to the *standard* QM is evidence that retrocausation is a viable explanation.July 11, 2015 at 1:12 am #2623David MillerParticipant
Hi Daniel Rohrlich
Firstly and with respect, I am very enthusiastic about your idea of using “the existence of a classical limit” as an axiom. Apart from the fact that you have shown it to be fruitful, I think it strengthens physicists’ “grip on reality” which seems to be being progressively loosened.
I have a few questions/comments about your retrocausality argument which I hope you might consider responding to.
(1) In your GHZ example, Jim steers a bipartite system into a mixture of either product or entangled states. In an EPRB experiment, Jim can steer a single quantum system into a mixture of polarization states of his choice. Does the involvement of entangled states in the GHZ case add anything to the argument? It “feels” like it does (the paper of Timpson and Brown may be relevant here).
(2) In the DCES case, Jim steers into states which span the Hilbert space while in the GHZ case, Jim steers into a subspace. Is there a material difference?
(3) In our submission to this forum, Matt Farr and I try to draw a distinction between “correlation” and “causation”. In our terms, “correlation” applies when joint probabilities are changed but marginals are not. “Causation” is reserved for changes in the marginal probabilities. Relativistic causality means we are talking about “correlation” in this sense in the relevant cases. Perhaps your suggestion about retrocausality in A reasonable thing that just might work attracted criticism because the term unconsciously implies in the minds of readers that a (retro)change in marginal probabilities is involved. Perhaps a better and less alarming term is “retrocorrelation”. But if the marginal probabilities are not changed can there be a distinction between “correlation” and “retrocorrelation” (except by bringing in the direction of causation when the marginals are changed)?July 14, 2015 at 9:31 am #2723Nathan ArgamanParticipant
I agree that retrocausation is a good axiom to use. At least it’s negation – the assumption that the causal arrow of time applies to microscopic degrees of freedom – should not be used.
But I think you’ve overstated your point. From your description, it sounds like there’s a potential causal loop – couldn’t Jim ask Alice and Bob whether they’ve observed a violation of Bell’s inequality, and if they have, decide to jam it? The answer is, as you know, that even if Jim “steers” them into an entangled state, they must know the results of his measurement in order to bin their data in a way that would exhibit violations of the inequality. I think that even in a brief description, you need to mention that, in order to avoid more controversy than necessary.
To me, this is indeed an indication that the wavefunction is epistemic rather than ontic. In a stochastic theory where future as well as past boundary conditions affect the probabilities of microscopic degrees of freedom, and at the same time a low-entropy past condition is imposed, there is a chance that information causality will be a consequence, and that all of QM will follow. In such a theory, there will be a need for something like the wavefunction to represent the known (epistemic) probabilities up to some time, and it will not be surprising that its complexity is exponential in the number of particles (like the Liuoville equation in classical many-body phase space). I do find it surprising that the exponential complexity issue is hardly ever brought up as another indication in this direction.
All the best, Nathan.July 16, 2015 at 11:27 am #2791
Thank you all for your comments and questions! I will try to do justice to them.
1. Dear Miroljub, the only variables I considered were those taking values ±1 and averages of many such variables. The goal was to investigate simple correlations subject to the Bell-CHSH inequality. I did not attempt to investigate anything beyond such correlations, such as the center of mass of an ensemble of “the world’s smallest soccer balls” [p. 336 of M. Arndt, O. Nairz, A. Zeilinger, “Interferometry with macromolecules: quantum paradigms tested in the mesoscopic world” in Quantum (Un)speakables: From Bell to Quantum Information, eds. R. A. Bertlmann and A. Zeilinger (Berlin: Springer), 2002, pp. 333-550]. At least at this stage, where I consider only post-quantum correlations, I don’t need to derive the classical limit of macromolecules: quantum mechanics dictates their classical limit, and it is doubtless consistent with relativistic causality, because quantum mechanics is.
2. Dear Travis Norsen, I thank you for your critique of my claim via Bohmian mechanics. Let me try to summarize your argument in terms of quantum mechanics with nonlocal hidden variables. We assume that each particle in each shared triplet comes with hidden instructions how to respond to possible measurements on the part of Alice, Bob and Jim. In the scenario of interest, Jim measures last. At an intermediate moment – after Alice and Bob have made their measurements – what is left of the nonlocal hidden variables are instructions to Jim’s particles regarding the result (±1) of each of his two possible measurements. Then when he measures what he measures, all he does is create facts in his own laboratory. Nothing retrocausal here, right?
I claim that this analysis is valid but not exhaustive. Here is a different analysis: Alice and Bob announce that they are about to perform the first absolutely loophole-free test of Bell’s inequality. To their laboratories flock physicists and journalists, who are given complete access to every last detail of the measurements and results. The same goes for Jim’s measurements, the results of which are available in real time. His only decision concerns what to measure on all his particles, i.e. whether Alice and Bob will succeed or fail. Of course, the success or failure of our efforts is often unpredictable. But if Alice and Bob succeed, they succeed, by all accounts, because they made their measurements on a mixture of entangled states, which they could later bin appropriately. And if they fail, they fail, by all accounts, because it is pointless to try to test Bell’s inequality on product states. The accounts in this case differ fundamentally from seemingly similar accounts in which we say, “In retrospect, if we had only known…” or “What made me believe that…”). Here it is not just new information or a new perspective that affects our account of the past, but incompatible facts (entangled states or product states) that depend on Jim’s later decision.
But – thanks to your analysis – I am considering whether my claim assumes a particular interpretation of quantum mechanics. See also my reply to Ken (next).
3. Dear Ken, I agree that Asher Peres had a different aim, but his thought-experiment suits me fine (as does the fact that it went from thought to laboratory).
You write that the apparent retrocausation disappears if we don’t think of “entanglement” as a state of reality. I prefer the safer terminology of preparation of a state and measurement of a state – what defines the state is its preparation. Then, regardless of interpretation, we can account for different results, confirming quantum predictions, obtained from measurements on one state or another. And then you write as follows: “I understand that you’re looking at Alice and Bob’s measurement outcomes, not the wavefunction, but in standard QM those outcomes result from the wavefunction, which is in turn unchanged by anything a future Jim might do.” It is precisely here that I disagree with you: Jim’s decision of what to measure is part of the preparation of the state that Alice and Bob measure!
I can already hear your objection: “No, it is Alice and Bob who prepare Jim’s state!” But that is a circumlocution when my way of putting it is apt. It will catch on. I am reminded of a beautiful NYT article about an evangelical Christian college that had constrained the teaching of evolutionary theory on its campus. The article ends with an interview with the president of the college, who is quoted as saying, “But this is Bryan College, and this is something that’s important to us. It’s in our DNA. It’s who we are.”
4. Dear Michael B. Heaney, I’m sorry that you didn’t spell out your comment at length. I understand that path integrals offer the possibility of imposing arbitrary initial and final boundary conditions, i.e. pre- and post-selection, which, in the approach of Yakir Aharonov et al., allows “weak” measurements to yield “weak” values. Backward (in addition to forward) causation emerges very naturally in this approach.
5. Dear David Miller, first, I thank you for your comment about physicists’ “grip on reality”!
In response to your questions: First, what my GHZ example has to add to EPRB steering is nonlocality, i.e. the possibility of testing Bell’s inequality. If all we have is EPRB steering, in which Jim measures in the future light cone of measurements already made on the systems paired with his, what is the case for retrocausality? I could only make what Travis rightly calls a “silly” claim. Second, DCES is…is…aha! Delayed Choice Entanglement Swapping! I suppose that Jim’s additional options in the DCES case would allow variations on my argument but would not add to nor detract from its essence. Third, my reply to Travis explains that the retrocausality I have in mind is not retrocorrelation (a good idea and a good word) – if it boils down to correlations, we are back to nonlocal hidden variables.
6. Dear Nathan, thank you (again) for your comments in the past and in the present. You wrote a sentence that needs lots of elaboration: “I do find it surprising that the exponential complexity issue is hardly ever brought up as another indication in this direction.” Who should elaborate it, if not you, and if not now, when?
With best wishes,
DanielJuly 16, 2015 at 1:09 pm #2795Aurelien DrezetParticipant
Dear Daniel, Your example remind me aslo the old paper by elitzur [http://arxiv.org/abs/quant-ph/0205182] where retrocausality is also involved. This is of course connected to the work by Pres. My question concerning your claim that QM is retrocausal goes like that: 1) what do you mean by QM 2) how can you prove that this version of QM is retrocausal.
To 1) I think you mean QM from the macroscopic observer point of view (i.e., Bohr?). such an observer dont care about the ontology of the real staff (like Bohmian could do instead). So, in this orthodox version of QM you will only collect clicks on your detectors and makes statistics. But then there is no proof of retrocausality since you can only observe the presence or absence of entanglement between Alice and Bob by making correlations with Jim. If Jim is missing data you will just observe a lot of uncorrelated points at A and B. We could thus claim that it is only your post analysis which showed that A and B were correlated even if it looks magical. Clearly, you cant use this scheme to transmit faster than light signal which could then be equivalent to backward causation and convince me. This is my answer to 2) It seems that you are thinking in a ontological way like Bohmian or Rod Sutherland do. Then it seems that you are going beyond the ‘macroscopic observer view’ of QM. Now if you take an ontological view like the one of Rod Sutherland you have a retrocausal theory from the ground (I still dont konw if the theory is correct or not but let admit it here). Still this retrocausality will stay hidden like in your example and I dont think that there is any example demonstrating retrocausality from the view of macroscopi observer otherwise this could look like a time machine?
I precise also that I am not convinced by the point of Travis Norsen. Bohmian mechanics neeed a special foliation to explain the delayed entanglement. This is not very elegant to go back to the old aether concept after 1 century of relativity isn’ it ? I would like to knwo the point of view of Travis on that point as well.July 17, 2015 at 7:14 am #2821
somewhat simplified, my point was: how does your conclusion fit with the fact that quantum observables typically admit smooth transitions from ‘quantum’ to ‘quasi-classical’. In other words: is the ‘quantum side’ implicit to your considerations–that’s what i’d like to learn about?
MiroljubJuly 17, 2015 at 12:46 pm #2828
Thank you again!
1. Dear Aurelien Drezet, it is very appropriate that you invoke Bohr in this connection since, in my story about Alice, Bob and the first loophole-free test of Bell’s inequality (above in my answers to Travis and Ken) I subject quantum phenomenology to a classical description, as Bohr did. But I don’t see that I have committed myself to any ontology. If we only talk about “preparation” and “measurement” of states, what ontological commitment have we made? (And I certainly don’t consider time machines or exceeding the speed of light.)
Likewise, in returning to Travis’s question (about whether my claim applies to all interpretations of quantum mechanics), it seems to me that interpretations are not the issue. The issue is the description of quantum phenomenology in natural language. Thus, on the one hand, my claim is stronger because it is not tied to any particular interpretation. On the other hand, it is weaker because I don’t prove that quantum mechanics is intrinsically retrocausal; I argue that it is more natural and apt to include Jim’s choice in the preparation of Alice and Bob’s state, than to be tied to any microscopic account.
2. Dear Miroljub, please be patient with me; I found your question a bit cryptic. If I misunderstood, please try again!
Neither “A reasonable thing that just might work” nor my earlier paper “PR-box correlations have no classical limit” mentioned Planck’s constant h. Your question suggests to me that it might be worth examining how the two limits h → 0 and N → ∞ (where N is the number of systems in an ensemble) fit together. Making the role of h explicit might clarify that the transition from “quantum” to “quasi-classical” (or “semi-classical”) in my proof is indeed smooth.
DanielJuly 17, 2015 at 1:31 pm #2832
“I certainly don’t consider … exceeding the speed of light.” I think that is the problem. The setup/scenario in question here is exactly the sort wherein Bell showed — making a few other assumptions (like that the future cannot influence the past, that experimenters’ choices about what to measure can be considered “free” in the relevant sense, etc.) — that faster than light causation is *required*. So … *of course* … if you just decide a priori, up front, to refuse to consider the possibility of faster than light causal influences, you will end up convincing yourself that there is backwards in time causation, or that free will doesn’t exist, or some such thing. But arguments with unadmitted premises are never good/convincing.
And by the way, if the surprising conclusion one does end up endorsing in this situation is that there are backwards-in-time (but non-super-luminal!) causal influences, one really does not thereby elude causal influences that exceed the speed of light, because one can always “chain” or “zig-zag” backwards- and forwards-in-time influences to produce a net (indirect) influence across spacelike separation. So, while there are admittedly some other ways of looking at this, the idea of trying to avoid nonlocality (in Bell’s sense) by endorsing sub-luminal but retro-causal influences, always struck me as silly. You don’t actually succeed, that way, in avoiding the thing one wanted to avoid.July 17, 2015 at 2:02 pm #2835
Dear Travis, you have misunderstood. What I meant by “exceeding the speed of light” was superluminal signaling, exactly in the sense of Aurelien’s comment, to which I responded: “Clearly, you cant use this scheme to transmit faster than light signal which could then be equivalent to backward causation and convince me.” I do not rule out faster-than-light causation that is nonsignaling.
Best, DanielJuly 17, 2015 at 2:38 pm #2836
Ah, sorry! Thanks for clarifying. But in any case, I still think there is some unacknowledged assumption in your argument that makes you think you’ve established retro-causation, when in fact faster-than-light (but perfectly forward-causal) influences (such as in orthodox QM with a preferred frame over which collapse occurs, or in Bohmian mechanics) can explain the phenomenology in your example perfectly well.July 17, 2015 at 3:34 pm #2840
Dear Travis, as regards the “unacknowledged assumption”, I refer you to my first reply to you and to my replies to Ken and to Aurelien, especially the part that begins “Likewise, in returning to Travis’s question”.
Shabbat shalom [I’ll be offline until tomorrow night]
DanielJuly 17, 2015 at 4:52 pm #2844
Thanks Daniel — Yes, I saw (and read with interest!) your earlier comments above. It seemed (from what you wrote in #2791) that you were still thinking of the argument as really strongly suggesting (if not quite rigorously proving) that there is retro-causation, but then also reconsidering this. I didn’t really have anything to say in response to that. Then in #2828 it seemed like you remained somewhat conflicted… on the one hand you were saying that the strength of your argument lies in the fact that it’s not tied to any particular interpretation, but on the other hand you acknowledge that the argument doesn’t actually prove retro-causation. I guess I just meant to be continuing to prod the part of you that still feels like the argument does show/suggest/whatever retro-causation in some interpretation-neutral sense. I just don’t think that’s right. I think you get to that conclusion basically by making various assumptions (something along the lines of locality, something along the lines of no hidden variables, etc.) that I would consider anything but interpretation-neutral.
But in so far as at least part of you acknowledges that, while this scenario *could* certainly be explained using retro-causation, it could also be perfectly well explained without any retro-causation, I’m happy and have nothing further to add. =)July 18, 2015 at 8:49 am #2853Aurelien DrezetParticipant
Dear Daniel, Thank you for your answer. The reason why I Invoke ontology is because as you just replied now ”I don’t prove that quantum mechanics is intrinsically retrocausal”. But if we dont prove and if you only suggest then you are going beyond facts? or not? Actually, my problem is the following David Hume already showed that causality can not be proven experimentally : you will never be sure that the succesion you observed once will come back again if you reproduce the experiment. You need a law and a postulate to fix that. Classical physics with Newton called that determinism and it relied on the study on an ontology describing the dynamics of the real world existing independently of the observer. If now you say that it is ‘natural’ to use retrocausality you somehow enter in the metaphysical universe of ontology like Bohmian do. This is beyond Bohr of course and I say that only because in my view you are hidding some (may be unconscious ) ontological preconceptions under your interpretation.July 18, 2015 at 8:55 am #2857
sorry for producing confusion. I’ll need some time to frame the subtleties i am interested in in a form appropriate for this forum.
MiroljubJuly 19, 2015 at 11:55 am #2898
Dear Travis and Aurelien, this time it seems most appropriate to address you together.
I agree that we have no proof that quantum mechanics is retrocausal, or for that matter that quantum mechanics is not retrocausal. Some interpretations via their definition leave no room for retrocausality, but then they don’t show that retrocausal thinking could not be helpful. That said, let me return briefly to the motivation for my claim in “A reasonable thing that just might work”. Aurelien, you are right! My interest in formulating quantum mechanics as a locally causal-retrocausal theory is that it might lead us to a quantum ontology that we can live with, free of “spooky action at a distance”. In trying to persuade others, but mostly myself, that quantum mechanics could be a locally causal-retrocausal theory, I came to the Alice-Bob-Jim thought experiment at the top of this page, and found it charming. To me, at least, it is just apt and natural to think of Jim’s choice as part of the preparation of the ensemble that Alice and Bob measure. After all, if variables A and B are correlated – as Jim’s choice is correlated with the kind of the correlations that Alice and Bob measure – and A is a human choice – isn’t it then that we say A caused B? And there is no paradox of causal circularity or free will.
Again, with many thanks to all,
- You must be logged in to reply to this topic.