Deriving the Geometric Utilitarian Weights — LessWrong