Advanced Targeting Through an Integrated, Adaptive, and Instructive Decision Aid Architecture

We propose an innovative combination of biologically inspired and other modern artificial intelligence (AI) techniques in the design of an integrated and evolving decision aid architecture. By drawing on our experience developing machine learning techniques and modeling decision processes, we have devised an eclectic approach that draws on a number of technologies. Specifically, the proposed system will offer a significant improvement in operation effectiveness by standardizing the implicit assumptions utilized by multiple decision aids and integrating them into a single unified and extensible architecture. This task will be accomplished through the application of object oriented techniques developed in conjunction with the Internet Softbot. Further, we will take advantage of a natural synergy among genetic algorithms, "programming by demonstration," and decision process modeling techniques to create an environment that can learn appropriate actions and procedures by observing experienced users. This learned information can then be used to assist less-experienced users of the decision aids as well as provide the capability to make decisions autonomously. We will prove the feasibility of our approach by developing and demonstrating a proof-of-concept prototype in Phase I.