Mitchell,

there is another argument speaking for many-worlds (indeed, even for all possible worlds - which raises new interesting questions of what is possible of course - certainly not everything that is imaginable): that to specify one universe with many random events requires lots of information, while if _everything_ exists the information content is zero - which fits nicely with ex nihilo nihil fit :-)

Structure and concreteness only emerges from the inside view, which gives the picture of a single world. Max Tegmark has paraphrased this idea nicely with the quip "many words or many worlds" (words standing for high information content).

Max's paper is quite illuminating: Tegmark, Max. 2007. The Mathematical Universe http://arxiv.org/abs/0704.0646

So we could say that there a good metaphysical reasons for preferring MWI to GRW or Bohm.