The diagnosing behaviour of intelligent tutoring systems

Abstract

Intelligent Tutoring Systems (ITSs) determine the quality of student responses by means of a diagnostic process, and use this information for providing feedback and determining a student’s progress. This paper studies how ITSs diagnose student responses. In a systematic literature review we compare the diagnostic processes of 40 ITSs in various domains. We investigate what kinds of diagnoses are performed and how they are obtained, and how the processes compare across domains. The analysis identifies eight aspects that ITSs diagnose: correctness, difference, redundancy, type of error, common error, order, preference, and time. All ITSs diagnose correctness of a step. Mathematics tutors diagnose common errors more often than programming tutors, and programming tutors diagnose type of error more often than mathematics tutors. We discuss a general model for representing diagnostic processes.