R. G. Almond, R. J. Mislevy, L. Steinberg, D. Yan, and D. M. Williamson: Bayesian Networks in Educational Assessment
Technology, Knowledge and Learning
Performance-based, diagnostic assessment that captures and scaffolds individual learner's competency development is a critical element of learning environment design, especially when interactive and complex learning tasks are involved. Yet the design and implementation of evidence-centered diagnostic assessment is challenging because: (a) the design of measurable performance tasks based on the underlying proficiency model is tricky, (b) the recording and coding of the performance data of each
... arner of a large sample can be complicated and time-consuming in comparison with grading in traditional testing, and hence (c) conducting a real-time diagnosis of the performance data to facilitate dynamic learner support and instructional planning for components of the competency is difficult. The book by Almond and his colleagues helps to address the aforementioned issues by explaining the potential of using graphical models, Bayesian network models in particular, to accumulate observed evidence about the state of proficiency of individual learners. All authors are experts of educational assessment and well published in the fields of cognitive science, data mining, school and learning improvement. Base on prior research of the author team and the evidence-centered assessment framework, this book describes: (a) the basics of a Bayesian network model, (b) the multi-step process of constructing a Bayesian network that consists of the specification of the proficiency model at the construct/variable level and the development of evidence models at the task level, (c) the process of calibrating the network model to the assessment data, and (d) the potential of using the Bayesian network to predict learners' proficiency level and guide the future development of a cognitively diagnostic assessment.