Does it make sense to use a uniform prior when no additional information is given?
 The justification of uniform priors is situationdependent and may be subjective; the point is that there is often no clear notion of "uniform distribution". Even in the discrete case, it's not always clear how to define the set of possible worlds. A world may be divided into multiple worlds when a new feature is given. Furthermore, all priors imply some assumed knowledge about the parameters; they are never "unbiased" in the statistical sense.
