Work in Progress: Practical Computerized Adaptive Assessment based on Bayesian decision theory

This work reports on the development of a novel software tool, namely Module for Adaptive Assessment of Students or MAAS for short, for adaptive multi-user Web-based assessment. The framework of Bayesian decision theory has been used for sequential mastery testing to classify students as masters or non-masters based on their responses to adaptively selected test items. MAAS applies a naive Bayesian decision model to adaptive testing, assuming many levels of student performance and taking into account not only right/wrong, but also blank answers. MAAS is embedded in a software platform, namely Platform for Adaptive and Reliable Evaluation of Students or PARES. MAAS is currently in a pilot use both for formative assessment and self-assessment. A preliminary experiment of using MAAS under real world conditions is described