Embedded On-Line Assessment for Intelligent Tutoring Systems

A great deal of data show that diagnostically-tailored instruction is superior to untailored instruction. Thus, an important way in which automated instructional systems differ is in the degree to which their instruction is modified by an inferred model of the student’s internal representation of the subject matter. Intelligent Tutoring Systems (ITS), the newest generation of automated instructional systems, maintain a dynamic model of the student’s internal representations. ITS use this student model to diagnostically tailor instruction for the current student. The resulting automated instruction is often more effective or more efficient than alternative forms of instruction. In this paper, we will describe student modeling techniques used by ITS developers to accomplish on-line diagnostic assessment in support of individually tailored. automated instruction. The basis for these student modeling techniques is a set of knowledge representation schemes that allow the ITS to (a) represent the knowledge or skill to be taught, and (b) use students’ on-line test and performance data to infer evolving mastery levels. Finally, we will discuss possible applications of these techniques for diagnostic or evaluative assessment in support of instructional methods other than ITS.