The Curse Of The Counterfactual — LessWrong