Another cooperative rationality exercise — LessWrong