A BDI-Based Methodology for Eliciting Tactical Decision-Making Expertise

Abstract

There is an ongoing need to computationally model human tactical decision-making, for example in military simulation, where the tactics of human combatants are modelled for the purposes of training and wargaming. These efforts have been dominated by AI-based approaches, such as production systems and the BDI (Beliefs, Desires, Intentions) paradigm. Typically, the tactics are elicited from human domain experts, but due to the pre-conscious nature of much of human expertise, this is a non-trivial exercise. Knowledge elicitation methods developed for expert systems and ontologies have drawbacks when it comes to tactics modelling. Our objective has been to develop a new methodology that addresses the shortcomings, resulting in an approach that supports the efficient elicitation of tactical decision-making expertise and its mapping to a modelling representation that is intuitive to domain experts. Rather than treating knowledge elicitation, as a process of extracting knowledge from an expert, our approach views it as a collaborative modelling exercise with the expert involved in critiquing the models as they are constructed. To foster this collaborative process, we have employed an intuitive, diagrammatic representation for tactics. This paper describes TEM (Tactics Elicitation Methodology), a novel synthesis of knowledge elicitation with a BDI-based tactics modelling methodology, and outlines three case studies that provide initial support for our contention that it is an effective means of eliciting tactical decision-making knowledge in a form that can be readily understood by domain experts.