A team of researchers from Kairos Research (Dr. Amy Summerville, Dr. Cara Widmer, Dr. Brandon Minnery, and consultants Dr. Ion Juvina and Dr. SubashiniGanapathy) has published an article in a special issue of the journal Decision focused on judgment and decision research involving “wisdom of crowds.” The paper presents research conducted as part of the team’s work on IARPA’s Hybrid Forecasting Competition (HFC) program, which explored ways to support human forecasters with AI tools that improve the accuracy of predictions about geopolitical events (e.g., the outcome of an election, the value of a stock index, or the number of conflict-related deaths in a region of interest).
Previous research has found that frequent forecasting updates (e.g., in response to events in the news) are critical for making accurate predictions. However, making regular updates is a costly use of human forecasters’ time. In their paper, the team explored having human forecasters make both initial forecasts and simple update rules to automatically update forecasts as a function of other measurable indices (e.g., if the exchange rate for the yen falls by X%, decrease the predicted closing value of the NIKKEI stock market by Y%). In addition to improving the overall accuracy of forecasts, the team found that making forecasting rules improved the accuracy of the initial forecasts, and that crowd-aggregated forecasts also benefited from automated updates. The paper is available as a preprint here and on the publisher’s website.
