Let the Hamiltonian $H$ be invariant under the symmetry $U$ such that: $$U H U^\dagger = H \, .$$ Consider a bias of ground states $\{\left|GS,n\right>\}$. It is often said that if $\left|GS,n\right>$ are not invariant under the symmetry $U$ then the system chooses one of these and hence spontaneous breaks the symmetry (e.g. ferromagnets and the Higgs).
But why do we emphasis a specific bias and which basis do we emphasis in general? Would it not be possible to write a basis consisting of a superposition of the states $\{\left|GS,n\right>\}$ which is invariant under the symmetry - would this not be a natural choice?