Dynamic inconsistency of the inaction and initial state baseline — LessWrong