I think David Deutsche has this idea that the best explanations should be treated as true even if you can't test it.