Computational models hold considerable promise for planning and prediction in applied domains such as military operations. But such models face a knowledge acquisition bottleneck when applying them to complex real-world problems. For example, agent-based models are typically implemented in an object-oriented language such as Java or Python, and their reasoning processes must be expressed in terms of objects, classes, instance variables, and other artifacts familiar to software engineers. The users of the model, operational decision makers, are experts in their domain but will not understand the details of a computer program. This mismatch between the conceptual worlds of problem stakeholders and software engineers is a major obstacle in advancing models beyond toy scenarios to those that require serious contributions from operational experts.

This slender volume provides a concrete example of one solution to this problem for agent-based models in the domain of military planning, a technique called the tactics development framework (TDF). TDF is an extension of the Prometheus methodology for multiagent systems, which itself is built on the proprietary JACK framework for Java-based agents with high-level cognitive capabilities (the so-called belief-desire-intention (BDI) agents that reason in terms of plans and objectives and communicate with semantically complex messages). Prometheus supports constructs meaningful to system engineers, such as system goals, message protocols, agent descriptors, and formal plans. TDF adds constructs appropriate for military operators, such as scenarios, teams, tactics, strategies, percepts, and actions. Like Prometheus, TDF is supported by a graphical design tool that allows users to manipulate its various constructs and then generates JACK code templates. The overall vision is that domain experts can construct meaningful system descriptions in TDF, and then generate code templates that embody their system vision for engineers to flesh out.

Chapter 1 motivates the importance of dynamic decision making in domains such as military operations, and argues for the BDI style of agents. Chapter 2 introduces the main constructs in TDF and the iconography that supports them. Chapters 3, 4, and 5 outline the three stages of the TDF methodology: requirements analysis, architecture, and behavior definition. These stages correspond closely to the Prometheus stages of system specification, architectural design, and detailed design. Chapter 6 illustrates the overall process with an integrated tutorial around a scenario involving undersea warfare. Chapter 7 describes the software tool, while chapter 8 concludes the book. There is also an integrated bibliography of 73 references, but no index.

This volume is an exposition of a particular methodology for a specific domain, not a general treatment of dynamic decision making. It will certainly be of interest to those who are modeling military scenarios, or scenarios that can be described in terms of tactics, strategy, and interactions, both adversarial and cooperative, and to users of the Prometheus methodology and the JACK framework. But it deserves much broader attention as a detailed example of how domain experts can be directly involved in specifying and designing complex models, and it sets a high bar for other modeling technologies to approach in providing similar support for their users.