Compromising improves forecasting.

Area of Science:

• Decision Science
• Cognitive Science
• Social Psychology

Background:

• Historical reliance on supernatural and expert forecasting has evolved to collective intelligence.
• Current collective intelligence models primarily focus on individual forecast accuracy.
• A gap exists in understanding aggregated forecast performance.

Purpose of the Study:

• To investigate if compromise forecasts (group averages) outperform individual forecasts in predictive accuracy.
• To analyze the temporal dynamics of forecast accuracy as events approach.
• To propose a method for enhancing collective predictive intelligence.

Main Methods:

• Analysis of 5 years of data from the Good Judgement Project.
• Comparison of accuracy between individual forecasts and compromise forecasts.
• Examination of how forecast accuracy changes over time relative to event dates.

Main Results:

• Compromise forecasts demonstrated superior accuracy compared to individual forecasts.
• The accuracy advantage of compromise forecasts remained consistent over time.
• Forecasting error for both individuals and compromise forecasts began to decrease approximately two months before the event, contrary to monotonic expectations.

Conclusions:

• Aggregating forecasts through averaging (compromise forecasts) is an effective strategy to enhance collective predictive intelligence.
• The timing of accuracy improvement is not strictly monotonic, with a notable decline in error starting two months prior to an event.
• The proposed method offers a practical approach for improving forecast accuracy in real-world scenarios.