While theories of cognition can be traced to antiquity, modern cognitive science is very much a child of the computer age. One powerful approach to studying cognition is to construct a computer model that embodies a theory, and to use it to explore the consequences of that theory. This approach is more accessible to objective review and critique than introspective accounts of cognition, and free from many of the difficulties (ethical and otherwise) involved in conducting experiments with actual people. Of course, at the end of the day, computational models must be aligned with both introspective and experimental results in order to be meaningful to the broader psychological community.

These two edited volumes provide convenient points of entry into the wide range of research currently underway in computational cognition. Though produced by the same editor, they are very different and complementary volumes.

Cognition and multi-agent interaction examines the relation between individual cognition and social behavior. Much of our intelligence (some would say all of it) is both driven and consumed by interactions with other cognitive beings. However, as the introductory chapter makes clear, much social simulation is done with black-box agents without a clear underlying cognitive model, while much cognitive research focuses on the individual agent rather than on inter-agent exchanges. The book’s fundamental hypothesis is that these two areas of research should be integrated with one another. The next three chapters introduce three examples of cognitive architectures: ACT-R with its integration of neural and symbolic reasoning, Soar with its chunking capability, and the editor’s own CLARION architecture, with its distinction of implicit and explicit cognition. The next ten chapters are standalone papers describing various experiments with interacting agents built on these and other platforms, illustrating the range of ways that cognition is involved in sociality. The last five chapters synthesize the issues raised in the specific systems that the previous chapters have presented.

The Handbook is much broader, dealing with the full range of cognitive modeling without restriction to its relation to social interaction. After an introductory chapter, five chapters present a range of cognitive modeling paradigms. These are not the high-level architectures in the previous book, but rather the underlying mechanisms: connectionist networks, Bayesian reasoning, dynamical systems, declarative logic, and constraint-based reasoning. The next 18 chapters, the bulk of the book, are devoted to discussing a wide range of different cognitive functions, for example, learning, categorization, social simulation, explanation, visual information processing, and motor control. Where the chapters in the previous book were descriptions of individual systems or experiments, these are review chapters, written by experts in each cognitive function but covering the range of research on that function, and not just the authors’ work. The book ends with two chapters synthesizing the role of computer modeling in cognitive science.

The two volumes share a common commitment to consider a broad range of models. Here one can find agents driven by quantitative neural models alongside implementations of high-level symbolic belief-desire-intention (BDI) cognition (and in some cases, such as ACT-R, integrations of these two extremes in a single system). Like most edited volumes, the bibliography of each volume is distributed across the chapters, not collected at the end. Cognition and multi-agent interaction has a very brief index, but the Handbook has a much more thorough subject index and also a detailed author index. Both volumes will be invaluable to researchers in bringing together and organizing a wide range of research.