Book Review: Bayesian Models of Perception and Action: An Introduction by Ma, W. J., Kording, K. P., & Goldreich, D.
Ryan S. Smith · Perception · 2024
In this book, Ma et al. have the ambitious goal of introducing Bayesian models of perception and action at a level accessible to both university students and researchers in the cognitive and neural sciences.They aim to provide sufficient mathematical details and example applications that would, in principle, allow the reader to construct such models and apply them in their own research.While this will be of most practical use to researchers at the graduate level and beyond, the format also has a textbook feel in many places, and one could envision the book being used as the backbone of a semester course focused on the topic.Along those lines, each chapter includes many exercises and problem sets designed to facilitate consolidation and generalization, which could be assigned as coursework, but could also be done independently by practicing scientists trying to pick up new analytic techniques.The authors make clear efforts to avoid the assumption that readers already possess extensive mathematical expertise, in hopes of allowing the book to be useful to a broader range of psychology researchers.I found these efforts to be partially successful, and more so in some chapters than others.They state in the introduction that no background in probability theory is required and that most content should be understandable without a strong mathematical background generally; however, they are also clear that a working understanding of calculus will be necessary for some sections of the book to be fully accessible.I think the first few chapters should be highly accessible and engaging to the majority of perception researchers.They include many explicit efforts to explain notation, terminology, and even fairly basic mathematical details.This changes somewhat quickly in later chapters, however, where many examples and exercises begin to assume a more considerable mathematical background.To the authors' credit, they include multiple appendices that provide additional resources for readers to gain a greater understanding of necessary mathematical foundations.Making full use of these appendices will still require considerable motivation by the reader, as well as other resources external to the text, but I thought they served a useful purpose.The Introduction and Chapter 1 jointly provide a very clear overview of Bayesian models at a conceptual level, with several engaging real-world examples.Chapter 1 also provides some useful historical context.Certain choices for notation (and early use of some terminology) in these opening chapters could perhaps lead to minor initial confusion for the uninitiated, but