Multi-Segment Computerized Adaptive Testing for Educational Testing Purposes

Abstract

Computerized adaptive testing (CAT) was initially developed as a psychometric tool for the efficient estimation of the ability of a student. Because of technological and psychometrical developments, CAT can now meet practical conditions and can also have other purposes than the mere estimation of ability. It can be applied in not only summative but also formative settings. In this paper, attention is given to the goals of testing in education and the different CAT algorithms serving these purposes. In particular, the approach of multi-segment adaptive testing is described. A multi-segment CAT consists of a number of segments, each with its own algorithm and branching rules. In this approach, practical constraints can be implemented in CATs. Furthermore, having a different testing purpose per segment for possibly different parts of the population is possible. The method will be illustrated with a CAT that has been developed as a part of an operational student monitoring system, which is a spelling test of Dutch words.