entropy1 said:
So what I am wondering about is if the spin is defined as a certain outcome of a certain measurement, or that it is regarded as something ontological real, in which case the measurement doesn't necessarily has to represent the value of it.
I am aware that these measurements don't commute, but if one took the correlation between two measurements on the same spin to be real, but not de measurements themselves*, then we might have less of a problem for instance in considering that a two dimensional operator yields two real worlds in MWI, because the measurement outcomes are not real*.
Observables are generally defined as a quantified phenomenon that can be measured. This is true for all of physics, also in the realm, where classical physics is applicable as an approximation.
If you have two observables which are not compatible, i.e., if their representing self-adjoint operators do not commute, it's generally impossible to prepare states, where both observable take determined values. In your example: If you prepare a state, where the spin component ##s_z## of a particle is determined, spin components in other directions are necessarily indetermined.
It's also easy to understand, why this is the case in this example: To prepare a spin component you can use the Stern-Gerlach setup, i.e., you let the particle run through a magnetic field with a large homogeneous part in the ##z## direction and some gradient also in the ##z## direction (then, because of ##\vec{\nabla} \cdot \vec{B}=0## it necessarily has also a gradient in some other direction, but that you can almost neglect, as will become clear below). Then the spin rapidly precesses around the spin direction, and thus almost only the force in ##z## direction due to the field gradient is relevant for the motion of the particle. This leads to an almost perfect entanglement of the ##z##-component of the particle's position with the spin component ##s_z##, i.e., if you use an appropriate beam of particles this beam splits into ##s## partial beams, each of which contains particles with (almost) determined spin-##z## components.
Now it's already clear even from these qualitative considerations only that in this way you can only determine the spin-##z## component, while all others are necessarily pretty indetermined. So if you decide to determine the spin-##z## component you cannot determine another spin component at the same time. To determine the other spin component you'd have to use a magnetic field in its direction rather than the ##z## direction. If you send a particle with determined ##z## direction, prepared using the corresponding Stern-Gerlach apparatus, through another Stern-Gerlach apparatus to determine, e.g., the spin-##x## component this apparatus randomizes the spin-##z## component again, i.e., you destroy the preparation of the spin-##z## component necessarily if you want to determine the spin-##x## component instead. For each individual particle with a determined spin-##z## component you cannot say which value the spin-##x## component you will get when running it through the Stern-Gerlach apparatus for the spin-##x## prepration. All you know from the preparation in the state with determined spin-##z## component are the probabilities to end up in any of the possible spin-##x## states.