Automated calibration training for forecasters
Abstract
In two studies, we investigated the effectiveness of an automated form of calibration training via individualized feedback as a means to improve calibration in forecasts. In Experiment 1, this training procedure was tested in a realistic forecasting situation, namely, predicting the outcome of baseball games. Experiment 2 was similar but used a more controlled forecasting task, predicting whether competitors would bust in a modified version of blackjack. In comparison to a control group without training, participants provided with calibration training had reduced confidence levels, which translated into reduced overconfidence and better overall calibration in Experiment 2. The results across both studies suggest that an automated form of individualized performance feedback can reduce the confidence of initially overconfident forecasters.
Document Details
- Document Type
- Pub Defense Publication
- Publication Date
- Jun 28, 2023
- Source ID
- 10.1002/bdm.2334
Entities
People
- Annie H. Somerville
- Cory K. Costello
- Eric R. Stone
- Jason Luu
Organizations
- Intelligence Advanced Research Projects Activity
- Wake Forest University