Supporting the Commander's information requirements: Automated support for Battle Drill processes using R-CAST

摘要

This paper discusses a novel approach that addresses the problem of supporting the Commander''s dynamic information requirements through automation of the Military Decision-Making Process (MDMP) for time-constrained environments and training purposes, as part of the Tactical Human Integration of Networked Knowledge (THINK) Army Technology Objective - Research (ATO-R) initiative. We demonstrate this capability with automated user support for the execution of battle drills. Our approach is based on adapting the R-CAST cognitively-inspired agent architecture towards a context-aware anticipation of information requirements. R-CAST is a computational model of the Recognition-Primed Decision (RPD) model, which models human decision making under time stress. R-CAST agents support and collaborate with human decision making teams as both “smart aids” and “effective teammates” by anticipating, investigating, seeking, and interpreting information relevant to decision making. A key feature of R-CAST is that the proactive sharing of information relevant to decision making is automatically generated by the computational RPD model. The fundamental research question being addressed is whether the inclusion of R-CAST in Army staff processes improves said staff understanding and execution of battle tasks. We adapted R-CAST to Battle Drill #26 (i.e., responding to an IED event) as a proof of concept for team decision making under stress and constant switching of modalities. We demonstrate that the use of R-CAST cognitive agents effectively assists the Battle Manager in the S3 cell with auto-filling certain forms required by doctrine in response to the dynamism of the current state of the environment, improving cognitive performance in this task. Our novel approach integrates relevant context in communication, information, and socio-cognitive networks, coupled with cognitive modeling. We report initial findings that we can use the R-CAST cognitive framew-n-nork to effectively and efficiently develop individual intelligent training tools that understand and support the dynamic information requirements of Commanders.