stevendaryl said:
I would say that it's definitely NOT that. I suppose there are different interpretations possible, but the way I read Bell's theorem is that the purely epistemic interpretation of the wave function is not viable.
Once again, I want to point out what are the implications of the claim the updating is purely epistemic. Again, we assume that both Alice and Bob have their filters oriented at the same angle, A. We ask what Alice knows about the state of Bob's photon. Immediately before measuring her photon's polarization, the most that Alice knows is: "There is a 50/50 chance that Bob's photon has polarization A". Immediately afterward, she knows "There is a100% chance that Bob's photon has polarization A".
It seems to me that if you want to say that the change is purely epistemic, then that means that the state of Bob's photon wasn't changed by Alice's measurement, only Alice's information about it changed. Okay, that's fine. But let's go through the reasoning here:
- After Alice's measurement, Bob's photon has definite polarization state A.
- Alice's measurement did not change the state of Bob's photon.
- Therefore, Bob's photon had definite polarization state A BEFORE Alice's measurement.
So it seems to me that assuming that measurements are purely epistemic implies that photons have definite (but unknown) polarizations even before they are measured. But that's a "hidden variables" theory of the type ruled out by Bell's theorem.
No, that's not what's implied although the "change of state" due to A's measurement is in my opinion indeed purely epistemic. Before any measurement, both A and B have simply unpolarized photons, which however are known to be entangled due to the preparation procedure in an entangled biphoton. Let's write down the math, because that helps here. I simplify it (somewhat too much) by just noting the polarization states of the photons and let's simplify it to the case that both measure the polarization in the same direction.
So initially we have the two-photon polarization state
$$|\Psi_0 \rangle=\frac{1}{\sqrt{2}} (|H V \rangle-|VH \rangle).$$
The single photon states are given by tracing out the other photon respectively, and both Alice and Bob describe it by
$$\hat{\rho}_{\text{Alice}}=\frac{1}{2} \mathbb{1}, \quad \hat{\rho}_{\text{Bob}}=\frac{1}{2} \mathbb{1}.$$
Now we assume that Alice measures the polarization of her photon and find's that it is horizontally polarized and nothing happens with Bob's photon which may be detected very far away from Alice at a space-like distance so that, if you assume that QED microcausality holds (which I think is a very weak assumption given the great success of QED). Then the state after this measurement is described (for Alice!) by the pure two-photon polarization state (which I leave unrenormalized to store the probability for this to happen conveniently in the new representing state ket; the state itself is of course the ray):
$$|\Psi_1 \rangle = |H \rangle \langle H| \otimes \mathbb{1}|\Psi_0 \rangle=\frac{1}{\sqrt{2}} |H V \rangle.$$
This, of course, happens with the probability
$$\|\Psi_1 \|^2=1/2,$$
which was already clear from the reduced state for A's single photon derived above.
Now, what Bob finds is with probability 1/2 H and with probability 1/2 V, because he cannot know (faster than allowed by the speed of light via communicating with Alice) what Alice has found. So Bob will still describe his single-photon's state as ##\hat{\rho}_{\text{Bob}}=1/2 \mathbb{1}##. Nothing has changed for Bob, and according to the usual understanding of relativistic causality he cannot know before measring his photon's polarization more about it if he doesn't exchange information about Alice's result, and this he can do (again using the standard interpretation of relativistic causality) only by exchanging some signal with Alice which he can get only with the speed of light and not quicker.
Alice knows after her measurement that Bob must find a vertically polarized photon, i.e., the conditional propability given Alice's result gives for 100% V polarization for Bob's photon. That this is true can be verified after exchanging the measurement protocols between Alice and Bob, given the fact that via precise timing it is possible to know which photons A and B measure belong to one biphoton. That's why Bob can "post-select" his photons by only considering the about 50% of photons where Alice found an H-polarized photon, and then finds 100% V-polarized ones.
This clearly shows that the notion of state is an epistemic one in this interpretation, because A and B describe the same situation with different states, depending on her knowledge. Note that there can never be contradictions between these two descriptions, because given that A doesn't know that B's photon is entangled in the way described by ##|\Psi_0 \rangle## A wouldn't ever be able to say that B would find V-polarization with 100% probability, if she has found H-polarization. This is what's stated in the linked-cluster theorem, and this of course holds for any local microcausal relativistic QFT. If on the other hand a theory obeying the linked-cluster theorem (which is the minimum assumption you must make to stay in accordance with usual relativistic causality) must necessarily be such a local microcausal relativistic QFT is not clear to me, and I've not seen any attempts to prove this (see Weinberg QT of Fields, vol. 1).
As a "minimal interpreter" I stay silent about the question, whether or not there is an influence of Alice's measurement on Bob's photon or not, as already mentioned by Ilja above. I only say this is the case in standard QED, which is by construction a local microcausal relativistic QFT. If there is an extension to QT where you can describe non-local interactions in the sense of a non-local deterministic hidden-variable theory that is consistent with the relativistic space-time structure, I don't know, at least I've not seen any convincing yet in the published literature. But what's for sure a "naive" instantaneous collapse assumption is for sure at odds with the relativistic space-time description and it is, as the above argument (hopefully convincingly) shows, not necessary to understand the probabilistic outcomes according to QT.