Forecast Aggregation

Forecast aggregation aims to combine predictions from multiple sources to improve accuracy and reliability, particularly when the underlying information structures of the predictors are unknown. Current research focuses on developing robust aggregation algorithms that minimize worst-case error, exploring the impact of non-Bayesian reasoning in experts (such as base rate neglect), and analyzing the sample complexity required for effective aggregation, especially in high-dimensional settings. These advancements have implications for diverse fields, improving the accuracy of predictions in areas ranging from economic forecasting and climate modeling to machine learning ensembles and risk assessment.