PRELIM: Predictive Relevance Estimation from Linked Models
Final rept. 11 Jul-14 Oct 2014
SOAR TECHNOLOGY INC ANN ARBOR MI
Pagination or Media Count:
PRELIM Predictive Relevance Estimation from Linked Models draws on semantic models provided by some members of the PDS team to generate forecasts and estimates of value of information VOI for use by other members of the team. The central challenge in proactive decision support is to anticipate the decision and information needs of decision-makers, in the light of likely future developments in a scenario. In many cases, this anticipation does not require sophisticated forecasting, but can be triggered by simple decision rules e.g, If a typhoon is moving toward the Philippines, identify the state of Humanitarian AssistanceDisaster Recovery resources within a 1000 mile radius. Such an approach corresponds to automation of elements of recognition-primed decision-making. However, some cases may go beyond patterns that are readily learned from past experience, and require a forecasting mode of reasoning that in people is called explanation-based reasoning. Several conditions that require proactivity and prediction, including The need to preposition assets, materiel, or information because of intrinsic timelags Indications of unlikely but high value future events that cannot yet be perceived the typical indications and warnings scenario Estimation of second and third order responses in developing Blue COAs. This document summarizes the technical work done on Contract N00014-14-P-1185 to advance the PRELIM vision. It includes a short narrative that describes how someone would benefit from using the PRELIM technology as part of a Proactive Decision Support System a technology roadmap that describes the science technology that needs to exist to realize the PRELIM vision a description of the science research andor theory that needs to be done to create make these technologies real.