Like its predecessor [1], this volume is a strong, valuable exception to the general tendency of proceedings (and other early publications in a new research discipline) to lack coherency and to disorient and discourage readers. The editors have carefully selected 26 high-quality papers from the 54 workshop submissions that exemplify the best work being done worldwide; organized them coherently into sections that orient the reader to the particular approaches being taken; and successfully encouraged the authors to relate each of their papers to other contributions in the volume. In addition, they have prepared an annotated bibliography of 355 studies, organized around the same lines as the papers. The result, while still reflecting the richness of a multi-author volume, has a coherence and order that will enable even nonspecialist readers to find their way into the field.

The original volume organized its essays into three sections, corresponding to the topics of the workshop: theories, architectures, and languages. These topics offer a felicitous set of stepping stones from more abstract models to resources for implementation. The current volume has refined this analysis into seven sections that increase the coherence among the contributions in each section.

The six papers in Part 1, “Theories,” are devoted to logical formalisms for reasoning about agents. In addition to the perennial subjects of temporal reasoning and how to represent beliefs, desires, and intentions in a tractable manner, there is one contribution on reasoning about agent failure, and another using deontic logic to characterize the legal rights and obligations of agents.

Part 2, “Agent Capabilities,” contains only two studies, one on how an agent can proactively learn through its interactions with the world, the other on conflict resolution among collaborating agents. These essays, while well done in themselves, do not fit as well into the overall framework of theories, architectures, and languages as the other contributions to the volume.

Some of the more sophisticated agents being implemented today can maintain models of one another and use these models to guide their interactions. The four essays in Part 3, “Agent Modeling,” show various ways that this challenge has been addressed. They may be viewed as a special case of the “Theories” theme, focusing on formalisms intended for use mainly by agents rather than by human researchers.

The “Architectures” theme has two important components: the architecture of an individual agent, and system architectures within which multiple agents can interact with one another. The three papers in Part 4, “Agent Architectures,” address the first of these issues, while the four contributions to Part 5, “Multi-Agent Architectures,” address the second. Explicit recognition of the latter area is important, since it tends to be ignored in research derived from the single-agent orientation of classical AI.

The original implementation focus of the “Languages” theme has also been expanded into two parts. Part 6, “Languages,” contains four essays, including the best short introduction to the Knowledge Query Manipulation Language currently available. The three papers in Part 7, “Agent Testbeds and Evaluation,” give further guidance to practical implementation issues that are not related to languages in a strict sense.

As is customary in proceedings, each contribution has its own bibliography, but the editors’ annotated bibliography provides a central source for the most important contributions to the field. In addition, the editors have provided an index (containing over 250 entries) to the volume as a whole, increasing its usability. Through their careful work, they have produced a volume that will serve not only as a record of a single workshop, but as a credible introduction to agent research for serious students.