Fra said:
Indeed there is a justifiction for the assymmetry - the fact that QT and the standard model of particle physics is by definition, formulated to describe a small subsystem, from the perspective of a controlled classical environment, where processes can be prepared and repeated in a controlled and fapp unlimited manner. Noone has yet constructed a quantum theory without a classical background.
The conclusion would have to be, that once GR certainly has to be modified (given the singularities) and to modify both at the same time leaves us completely without any guidance, to accept that we need a classical background too.
Fra said:
The problem is that one can argue that this justification breaks down for cosmological observations;
What follows from the Copenhagen-QT-inherent conflict between the classical and the quantum part depends on the interpretation. Different interpretations handle cosmology differently. The realist interpretations extend the classical part to everything - they introduce a continuous configuration space trajectory ##q(t)\in Q## into the quantum part too.
Is there, similarly, also a wave function of the universe? A consistent epistemic interpretation would deny this. Can such a consistent epistemic interpretation describe the whole universe completely? Formally not, there has to be that part containing the incomplete information about the system. But that "incomplete information" we somehow have to know completely, assuming we have only incomplete information about that incomplete information smells paradoxical.
For what happens in practice, we have the example of thermodynamics, statistical mechanics in the Bayesian approach, where entropy describes no physical property of a system but our incomplete knowledge of the trajectory of that system. The picture is the same - the observer is outside the system. Can we do thermodynamics for cosmology? Obviously. But we would have to handle the internal observer on that planet Earth as something external. Could we handle Earth as completely internal too? That's possible too, simply use some far away alien civilization as the observer, and take out that alien planet somewhere in Andromeda instead of Earth. Both no essentially nothing about each other, beyond some general approximate information (say about the actual properties of the background radiation), so FAPP this is completely harmless. But the cut between observer and observed is there, and remains there, and I see no chance to get rid of it.
Once the conceptual problems related with this cut can be nicely considered based on thermodynamics, it makes sense to study them first in thermodynamics instead of quantum theory, and then to apply the results of what necessarily follows (say, that such a cut is unavoidable anyway) to quantum mechanics too. Simply it would not make sense to require from QT more than from statistical mechanics. If (or better once) we have a satisfactory situation in statistical mechanics, all we need is a consistent Bayesian approach to QT. Which exists, Caticha's entropic dynamics (ignoring with its existence, following the traditions of dBB theory, a lot of impossibility theorems like the many variants of PBR). It interprets ##|\psi(q)|^2## as required by the Born rule and the phase in terms of the entropy of the system in dependence given the information available outside the system. But it has also the continuous configuration space trajectory ##q(t)\in Q## which defines the ontology.
Fra said:
Does mean that non-classical observers can not do science, or are cosmology not science? Well maybe not. Or maybe there is something wrong with the question we ask - I argue there is. But this is a much more painful insight, further adding to the assymmetry.
I see no potential here for a serious conflict. In Caticha's interpretation we objectively have a trajectory ##q(t)\in Q##. The aliens from Andromeda can apply QT to describe that trajectory based on the incomplete information they have. In this sense, we are "quantum observers", and I see no problem with this.
Fra said:
If you adhere to my second paragraph above, one can similary conclude from start that "QT must be wrong". Or rather than "wrong", correspond to a a special limiting class of observers, that makes us unable to scale the framework to arbitrary observer frames required for full unification.
I see here only a problem with the "full unification" idea, which runs into the same problem already in the Bayesian variant of statistical mechanics where everything else is well understood. Nothing comparable with GR singularities in seriousness.