I take the position that information is exchanged, but does not involve a superluminal communication. Yakir Aharonov's time symmetric interpretation of quantum mechanics provides the explanation.
Please visualize two particles that are quantum entangled moving apart in opposite directions. At some distance from their common origin, Alice measures the spin of one of the particles and finds that the spin is in the up direction. In traveling from the point of origin to Alice, we may understand the particle's wave function to have, in a probabilistic sense, taken all possible paths and to possesses all possible states consistent with the initial boundary condition of the system at the origin. With TSQM we must now visualize a time-reversed wave function which proceeds backwards in time from the occurrence of Alice's experiment to the time and point of origin for Alice's particle. This backward in time wave function would also, in a probabilistic sense, take all possible paths and possesses all possible states consistent with three constraints: (i) the time evolution of the wave function is backward in time; (ii) the time-reversed wave function is bounded by the initial state of the system at the origin and (iii) the time-reversed wave function is also bounded by the particle location and spin information arising from Alice's experiment. It should be noted at this point that due to conservation of momentum the direction of spin manifest in Alice's time-reversed wave function will be opposite to the spin direction that Alice measured; and identical to the spin Bob will find when his measurement later occurs. In any event, Alice's time-reversed wave function may be understood to carry the spin information arising from Alice's experiment to the time and location of origin for the entangled particles. Here, the information contained in Alice's time reversed wave function may be understood to "bounce" forward in time in a state that is entangled with Bob's particle. Please note that weak measurements of Bob's and Alice's particles immediately prior to the occurrence of their respective ideal measurements will show that each particle has remained entangled with the other.
My conclusion from the foregoing is that TSQM reintroduces a classic-like causality, and locality, to quantum mechanics that I believe has very broad implications. This interpretation based on time reversal is far from original. As early as in 1983 Costa de Beauregard gave a formulation of the EPR setting that allowed a time-reversed EPR.
J. W. Moffat in his paper “Quantum Measurements, Nonlocality and the Arrow of Time” (See:
http://arxiv.org/pdf/gr-qc/9710019) proposes an absorber wave function reduction process to resolve the EPR paradox that is based on the retarded (forward-in-time) and advanced (backward-in-time) waves that John Cramer proposed in his transactional interpretation of QM.
The TSQM approach, which I favor, is presented in a paper by Yakir Aharonov and Jeff Tollaksen titled New Insights on Time-Symmetry in Quantum Mechanics (see
http://arxiv.org/PS_cache/arxiv/pdf/0706/0706.1232v1.pdf
Additionally, Dr. Henry Stapp in a private communication I catalyzed has stated:
“If one considers an EPR-Bohm-Bell correlation experiment, then during some interval in Process Time the initial (singlet) state of the two particles will be created.
Over an interval in Process Time this singlet state will grow out in an expanding V-shaped region of spacetime, toward the two far-apart detection regions. At some Process Time a detection will occur. At that moment in Process Time the state of the universe in the space-time past of the associated space-like surface will suddenly change, relative to what it was at the earlier moments in Process Time. In the V-shaped region of spacetime the state will suddenly jump from a singlet state of the two diverging particles to a state in which, for example, one particle is polarized in one specific direction, specified by the orientation of the device in one of the two regions, and the particle traveling along the other wing of the V is polarized in the opposite direction. The correlation between the parts in the two wings will be fixed instantly (in Process Time) over the entire V-shaped region in spacetime. The effective transfer of information about the choice of polarization direction, which choice was seemingly made by the agent/observer in one region, is made via the V-shaped region that extends backward in time: the [apparent] faster-than-light transfer of information is made by an effective transfer first backward in time to the region where the two particle interacted (or originated), and then forward along the other wing of the V.”
