Using Constraints to Develop and Deliver Adaptive Tests

This paper shows how the techniques used to develop knowledge-based systems can be applied to the construction of adaptive tests. It reports on an experimental approach to the construction of adaptive tests, and it draws on work in the fields of Intelligent Tutoring Systems, Expert Systems and Constraint Logic Programming. The distinctive features of the work are:

• The use of expert emulation as a basis for the design of tests.

• The use of logic programming and constraints for two purposes. Firstly, to support knowledge acquisition from an expert tutor during the development of a test; and secondly, to support the delivery of the test.

The paper, after reviewing the approach to adaptive tests in Intelligent Tutoring Systems research, presents a case for expert emulation under those circumstances in which statistically based testing procedures are inappropriate. The paper then describes how knowledge about the content and design of an adaptive test can be facilitated by computer support. The software developed for this task makes use of a constraint solver embedded in a Prolog system. In the subject area of the adaptive tests, namely elementary arithmetic, it is shown that this software can be used for several purposes. It can be used to: describe classes of problems, describe arithmetic skills, describe student responses to problems and to generate problems. The paper concludes with a discussion of a procedure for exploring a student’s knowledge of a syllabus.