Confidence Levels in Forecasts and Psychological Surveys — LessWrong