An Approach to Logically Updateless Decisions — LessWrong