Splitting Debate up into Two Subsystems — LessWrong