the classic multiverse (or "many-worlds") concept, in it's scientific guise, originated from Everett's formulation of QM, where he set out to re-formulate QM without its dependence on observers' special role "collapsing" the wavefunction. iirc, in his formulation, Everett does away with the collapse of the wavefunction to a determistic eigenstate post-"measurement"; instead he considers observer as being part of the system tied to a particular eigenstate that remains a mixture even after the act of measurment. in other words, a measurement does not somehow magically "collapse" a previously probabilistic world to a deterministic one, but rather the system (with observers being part of that system) is split into two or more orthogonal "worlds". so the whole concept arose from interpretation of the formalism (result wise the two formalism, classical and Everett's, are essentially equivalent). its truth value is really matter of faith. whether the interpretation is something falsifiable, i'm not sure. at least, i'm not aware of any experimental setup to test it.