Designing and Implementing Decision Aids for a Complex Environment Using Goal Hierarchies

Abstract

In dynamic tactical environments such as Anti-Submarine Warfare (ASW), operators must continuously balance goals for safety with mission goals and objectives as the tactical situation evolves. These high-level goals introduce specific constraints on operator actions that often conflict with one another. Being able to assess and balance conflicting constraints is thus a key element of the operator's cognitive processes. This design effort extends the COGNET (COGnition as a NEtwork of Tasks) modelling methodology in order to identify and resolve goal/constraint conflicts as they arise in the tactical context. COGNET is a meta-model, or architecture, for building and embedding models of human operators in real-time, multi-tasking systems. It integrates the GOMS notation for building operator task models with the blackboard architecture for building a knowledge representation, and production rules that link the tactical situation with the changing knowledge representation. An extension to the model was added in the form of AND/OR tree diagrams that represent the relationship between high level goals and lower level constraints on operator actions. The goal/constraint trees were modelled for the Submarine Operational Automation System (SOAS) operator using expert knowledge elicitation techniques. Results of the modelling effort indicate that operators must often keep track of many conflicting constraints in making decisions on ship course, speed, and depth, and often do not consider the complete set of constraints because of the limits on working memory. This finding highlights the need for decision aids for managing goals and goal conflicts. Applications of the goal/constraint hierarchies with a COGNET model framework to the design and implementation of SOAS decision aids are discussed.