This sounds interesting. Alternatively, you could have the procedure-makers not know what questions will be forecasted, and their procedures given to people or teams with some stake in getting the forecast right (perhaps they are paid in proportion to their log-odds calibration score).
After doing enough trials, we should get some idea about what kinds of advice result in better forecasts.
This sounds interesting. Alternatively, you could have the procedure-makers not know what questions will be forecasted, and their procedures given to people or teams with some stake in getting the forecast right (perhaps they are paid in proportion to their log-odds calibration score).
After doing enough trials, we should get some idea about what kinds of advice result in better forecasts.