The keystones of traditional intelligent tutoring systems (ITSs) have been complex procedures for student diagnosis and adaptive instruction based on diagnostic data. While some of these systems have been shown to be effective, they are also very expensive to develop. This paper describes another class of ITSs, non-diagnostic ITSs, which do little or no student diagnosis, and concentrate their intelligence in other areas. Intelligent features of non-diagnostic ITSs include: modeling of experts' reasoning processes and cognitive representations (often using graphic displays), comparison of student and expert performance, and replays and summaries of student performance. While traditional, diagnostic ITSs are usually intended to be used in a stand-alone fashion, non-diagnostic tutors are designed to facilitate collaborative learning among students and between teachers and students.

The non-diagnostic approach to ITS development offers either a low-cost alternative to traditional ITSs or a way to expand the educational capabilities of traditional systems. This paper presents a framework for comparing the features of non-diagnostic and diagnostic tutors. A number of non-diagnostic and diagnostic ITSs are described, and data on the costs and educational effectiveness of each type of ITS is presented. Finally, obstacles to wider use of non-diagnostic ITSs are discussed.