stevendaryl said:
So if you want to insist that Alice's measurement of the electron's spin (along the z-axis, say) always produces the result of Alice having measured spin-up or the result of Alice having measured spin-down, then that means that we have to be able to interpret a superposition of macroscopically different states as being either in one state or the other.
Okay let's look at Spekkens toy model. We have a system with an electron, a measuring device and then the lab environment. We'll model each with a qubit.
In the model a qubit has four ontic states ##1, 2, 3, 4##. However there is an epistemic limit so you can only resolve the ontic space to half the maximum limit. This means there are six maximum knowledge epistemic states:
$$
|\uparrow\rangle = \{1,2\} \\
|\downarrow\rangle = \{3,4\} \\
|+\rangle = \{1,3\} \\
|-\rangle = \{2,4\} \\
|i\rangle = \{2,3\} \\
|-i\rangle = \{1,4\} \\
$$
So for instance ##|\uparrow\rangle## is an epistemic state indicating the ontic state is ##1## or ##2##. Superposition of two epistemic states like
$$|\uparrow\rangle + |\downarrow\rangle = |+\rangle = \{1,3\}$$
can then be seen to not be "or" from Kolmogorov probability but a bilinear mapping from maximum knowledge epistemic states to maximum knowledge epistemic states where one ontic state from each is present in the pair of the superposition.
We also have states of non-maximal knowledge, mixed states, like:
$$\frac{1}{2}\mathbb{I} = \{1,2,3,4\}$$
Now let us consider an entangled state:
$$|\uparrow\uparrow\rangle + |\downarrow\downarrow\rangle = \{1.1, 2.2, 3.3, 4.4\}$$
where ##a.b## means the first particle is in ontic state ##a## and the second is in ontic state ##b##.
As you can see this means that the first particle is in one of the states:
$$\{1,2,3,4\}$$
which is a mixed state.
By gaining more knowledge of the entire system, i.e. that the two particles are always in the same ontic state, I know less about a single particle. I don't what state a single particle is in at all and thus have non-maximum knowledge of it, a mixed state.
Now consider the three particle state:
$$|\uparrow\uparrow\uparrow\rangle + |\downarrow\downarrow\downarrow\rangle = \{1.1.1, 1.2.2, 2.1.2, 2.2.1, 3.3.3, 3.4.4, 4.3.4, 4.4.3\}$$
Take the first particle to be the atomic system, the second to be the device and the third to be the lab environment. A superobserver outside the lab might use the above state. An observer within the lab, who doesn't track his lab environment (it's impossible otherwise he'd be a superobserver) might see his equipment read "up" in which case he knows that "atomic system + device" is in the state:
$$|\uparrow\uparrow\rangle = \{1.1, 1.2, 2.1, 2.2\}$$
which can easily be seen to be compatible with the above use of a superposition by the superobserver.
This is because states of maximum knowledge of the three particle system are consistent with states of maximum knowledge of subsystems that remove one of its "branches". This in turn is because one is reducing the systems tracked and thus one can increase knowledge of the subsystems.
This is essentially what happens in all epistemic models, but it's easier to see here in the toy model.
See Chapter 11 of Richard Healey's
"The Quantum Revolution in philosophy" for a longer exposition on it in the ##\psi##-doxastic case.