The book is a collection of relatively well-connected papers on reasoning about opponents. This is a topic of growing interest within the artificial intelligence and game playing community, and the book is a welcome contribution to a growing field. The aspects that are emphasized in the book are related to adversarial reasoning in complex military operations.
The 13 chapters are organized into three major areas: intent and plan recognition, deception, and strategy formulation. The coverage includes some more theoretical topics, such as human factors, dealing with imperfect information, and reinforcement learning, as well as some practical examples in military games.
The chapters on intent and plan recognition cover topics such as Bayesian networks, belief-desire-intention architectures, and modeling of human cognitive processes. The chapters on deception are heavily focused on detecting deception in military applications. The chapters on strategy formulation are a bit more theoretical, ranging from handling partial and corrupted information to learning.
Relatively limited coverage is given to game theory. Nash equilibria are mentioned in one of the chapters on deception, but most of the chapter deals with examples of air strike games rather than theory. Modeling a war game as a nonlinear risk-sensitive stochastic control system is discussed in the chapter on robustness against deception. Nonmilitary games, specifically poker, Texas hold’em, and Kriegspiel chess, are covered in the chapter on the role of imperfect information. Network security is used as an example in the chapter on plan recognition, and financial fraud is covered briefly in the chapter on deception as a semantic attack.
In conclusion, the book brings together a valuable collection of material on an important and timely topic. The book is best for people interested in understanding the state of the art of adversarial reasoning for military applications. Readers who are interested in adversarial reasoning for multiplayer games such as poker or RoboCop will not find much specific material, although the theory and algorithms presented are of value.