An Attempt at Preference Uncertainty Using VNM — LessWrong