Skip to content. | Skip to navigation
The merits of computerised adaptive testing for computer assisted learning
by Theo Eggen and Bernard Veldkamp, University of Twente
Abstract:
A major goal in computerized learning systems is to optimize the learning process of a student, while in computerized adaptive tests (CAT) the efficient measurement of the proficiency of students on well defined domains is the main focus. There seems to be a common interest to integrate computerized adaptive item selection in learning systems and testing. Item selection is a well founded building block of CAT. However, there are a number of problems that prevent the application of a standard approach, based on item response theory, of computerized adaptive item selection to learning systems. In this presentation attention will be paid to two unresolved points in this: the calibration of the items and the item selection in such systems. Traditional cats assume that the item parameters are as known. So in order to apply Cat techniques to learning systems a time consuming item calibration process has to be conducted. An alternative approach for that would be to predict item parameters instead of estimating them. In the presentation a regression tree approach to tackle this will be introduced and illustrated. In traditional CATS, item selection is based on maximum (Fisher) information, aiming at the lowest contribution to the estimation error of the measured ability. Being optimal for efficient estimating the ability, it is questionable whether this selection method is also optimal for establishing learning which can be measured by growth in the ability of the learner. In the presentation, alternative selection methods will be presented and compared.
