The Future of Forecasting Featured Pattern: P0779 May 2015

Recent efforts show the very different ways that researchers are trying to develop foresight methods for the twenty-first century.

Abstracts in this Pattern:

Two notable efforts investigate new forecasting approaches: the US Intelligence Advanced Research Projects Activity (IARPA; Washington, DC) Aggregative Contingent Estimation (ACE) Program and the Value at Political Risk (Vapor) model, which is the result of a collaboration between risk adviser and insurance broker Willis Group Holdings (London, England) and consultancy Oxford Analytica (Oxford, England).

IARPA's ACE Program has sought to improve forecasting of geopolitical events by exploiting recent research from the social sciences, insights from behavioral economics, and experience with prediction markets. Five teams from universities and research centers around the United States built forecasting systems, enlisted experts, and competed to provide the most accurate forecasts about events the IARPA identified as of particular interest. The winning team from the University of Pennsylvania (Penn; Philadelphia, Pennsylvania) succeeded in part because it developed a training program for team members that included education about the biases and flaws in reasoning that normally trip up forecasters, it gave team members constant feedback about the accuracy of their previous forecasts, and it allowed team members to work in collaborative groups. The Penn team was also able to identify several attributes that make for better forecasting, including open-mindedness, general cognitive ability, and patience in assessing data.

The Vapor political-risk model pushes risk analysis and forecasting in a completely different direction. Vapor is a large simulation that aims to identify major variables in global and national events and the character of their interactions, to assign probabilities to those events, and to estimate the financial cost of various scenarios. In effect, Vapor adapts some of the tools developed by the reinsurance industry—which are concerned with quantifying the effects of phenomena such as climate change and terrorist attacks—to political-risk analysis.

Although these two efforts go in very different directions, they both explore basic questions about forecasting and challenge basic assumptions in futures. These efforts suggest that although predicting the long-term future may be impossible, accurately forecasting near-term events may be more feasible than practitioners thought it was.