The workshop on “Architectures for Intelligent Interfaces: Elements and Prototypes,” held in March 1988 in Monterey, California brought together people who had working implementations of intelligent interfaces. The papers from the workshop were reviewed by peers, and a subset were chosen for expansion and revision for this volume. As a result, the papers are of high quality.

What is an intelligent interface? It is a user interface that incorporates some kind of artificial intelligence. This answer may not be satisfying to some, and the editors themselves note that it is often difficult to describe where the AI is in a particular system. The notion of an intelligent interface remains cloudy, but an enumeration of the main sections of the book give a flavor of what was considered at the workshop: “Multimodal Communication,” “Models, Plans, and Goals,” “Dynamic Presentation Design,” and “Knowledge-based Tools for Interface Design.”

The papers in this volume are professional, well-written, thoughtful, and carefully prepared. They are not vaporware presentations, and many of the systems described have become the basis of ongoing work that is still in the forefront of human-computer interaction research. Reading the book more than four years after the workshop itself, however, one gets a feeling of datedness. Many of the systems described, such as Nephew (in Szekely’s paper), UIDE, (Foley et al.), Peridot (Myers), and APEX (Feiner), have been significantly extended or redesigned by now. The newer systems are described well in accessible publications such as the CHI proceedings. The workshop papers are of historical interest, but no longer represent leading-edge human-computer interaction architectures. In addition, many systems that postdate the workshop and represent important architectural advances, such as BOZ [1], ITS [2], Unidraw [3], and ACE [4] are not represented in this volume.

An aspect of human-computer interaction research that has become standard in today’s work is reporting on user experience with a system. Again, the feeling of this volume is dated, as few of the papers describe user experience. This omission is a product of the time when the papers were written, and not the fault of the authors. Human-computer interaction research has simply changed in character. A noteworthy exception to this generalization is Cypher and Stelzner’s interesting paper on the SIMKit system, which reports the experiences of more than 100 SIMKit users.

The papers are quite technical, plunging deeply into the architectural details of each system. This approach is laudable, but I think readers would have appreciated a bit more editorial commentary, such as can be provided by introductions to each section of a book. A good introduction appears at the beginning of the volume, but no commentary of any kind follows. The breadth of topics covered (discourse models, automatic programming, graphical user interfaces, and so on) demands some higher-level discussion, comparison, and context-setting.