I’m skeptical in general that having people give their confidence levels leads to better aggregate predictions/outcomes, given that most people are terrible at confidence calibration. At the least, this would underweight the opinions of well-calibrated people while overweighting those of overconfident people.
I’d guess the best approach here should be basically copied from whatever the prevailing view is in the literature around consensus finding and aggregating opinions.
I’m skeptical in general that having people give their confidence levels leads to better aggregate predictions/outcomes, given that most people are terrible at confidence calibration. At the least, this would underweight the opinions of well-calibrated people while overweighting those of overconfident people.
I just meant they’d indicate how confident they are (1-9).
I’d guess the best approach here should be basically copied from whatever the prevailing view is in the literature around consensus finding and aggregating opinions.