Spymasters in search of forecasting software to enhance intell analysis
Reseachers seek better way to aggregate intelligency assessments from dispersed analysts
The intelligence agencies’ research arm is looking to procure software prototypes that will help improve forecasting.
Intelligence agencies often rely on the judgments of their analysts for estimates because there isn’t enough quantitative data to generate statistical forecasts, according to the Intelligence Advanced Research Projects Activity (IARPA) organization. Research indicates that taking the mathematical aggregation of many individual judgments is more accurate than those produced by individual analysts, according to IARPA.
At the same time, incorporating data about forecasters, such as previous performance, cognitive style and their level of risk aversion, could improve the accuracy of those aggregated forecasts, the intelligence agencies’ research arm said in a Broad Agency Announcement dated June 30. However, IARPA said there hasn’t been any research that has optimized aggregation methods using information about large numbers of predictors and their judgments, and little has been done to test ways to come up with conditional forecasts.
To fill that gap, IARPA has created the Aggregative Contingent Estimation (ACE) Program, and it wants organizations to propose research to develop methods that provide more accurate, precise and timely intelligence forecasts. ACE would lead to well-tested methods in software prototypes that support the creation of "conditional probabilistic forecasts" for a broad range of events and tools for to analyze distribution of forecast data, IARPA said.
Wanted: A better system for intelligence collection
IARPA said the software prototype should be able to be used on the judgments of more than 1,000 geographically dispersed analysts. IARPA says methods developed through the ACE Program would be used to:
Aggregate relevant knowledge from analysts across intelligence agencies
Come up with "probabilistic forecasts" for different events
Identify where additional analysis and collection may be required
Monitor changing forecasts that provide early warnings
Assess analysis’ accuracy.
IARPA plans for the program to consist of a 12-month base period followed by three 12-month option periods. Multiple awards are anticipated for the program and funding for option periods will depend on performance during the base period, option periods and program priorities, IARPA said.
Only U.S. organizations can be prime contractors and submit proposals for ACE, and overall 20 percent of the principals of a team must be from U.S. organizations or institutions, according to the announcement.
Proposals must be received by Aug. 18.
Ben Bain is a reporter for Federal Computer Week.