Book Review

Abstract

Bayesian networks found their applications in many fields: in medicine, for medical decision making (Lucas 2001), in Artificial Intelligence, for robots and object recognition (Schneiderman 2004), in ecology, for environmental modeling (Aguilera et al. 2011), and in educational assessment, for measuring students' abilities (Mislevy et al. 2000). The network method provides a logical and graphical structure for reasoning under uncertainty (Neapolitan 2003;Pearl 1988). For instance, Lucas previously applied these methods in medicine, where symptoms are entered in a network as observable variables. By formulating probability distributions and a network structure, the most likely disease and a fitting treatment for patients could be determined. In Artificial Intelligence, Schneiderman used Bayesian networks to automatically learn network structures with the goal to detect frontal faces, eyes, and the iris of the human eye. These applications motivate the use of Bayesian networks to assess uncertainty associated with (human) environmental interventions and classifications. Uncertainty around complex relationships is found in many fields and can also arise in educational assessment settings. Innovative educational assessments, such as interactive simulations, games or complex constructed responses, can benefit from Bayesian network methodology. The book "Bayesian Networks in Educational Assessment-The State of the Field" provides an exhaustive overview of applying Bayesian networks in educational measurement. A unique reference work has been written aimed at professionals, researchers and students in this field.The book consists of three main parts. Part I introduces the building blocks of Bayesian Networks in Educational Assessment. Firstly, the evidence-centered design (ECD) approach (Mislevy et al. 2004) is outlined and explained. Then, Part I introduces Bayesian probability and graphical networks, after which methods for efficient calculations, examples, and inferences by the model are described. Part II focuses on learning and revising the network and student models when actual student data has been observed. Different model parameterizations are described. These parameterizations will definitely contribute to the readers' understanding of the choices that could be made in construction and updating of the models. Estimating distributions of parameters, and approaches to model fit or model evaluation are also introduced. This part concludes with a complex example that illustrates the concepts that were discussed in the previous chapters. After dealing with the technical and mathematical issues of Bayesian networks, Part III is about application of Bayesian networks in assessment design. It is discussed in detail how Bayesian networks can be integrated in the various steps of the evidence-based design approach.The three parts of the book give an excellent overview of current developments and the status of a still developing and emerging field. In this way, it helps researchers in psychometrics and educational assess...