Involving the Learner in Diagnosis – Potentials and Problems

A crucial issue in building adaptive systems is maintaining models of users, which allows computers to tailor their behaviour to the needs of particular individuals. When a computer system has educational goals, modelling a learner’s cognitive capacity is essential for the system to provide individualised instruction and adaptive interaction (Self, 1999). Allowing the user to have some control over the diagnostic component and to inspect the model which the system has made of him has been suggested as one of the potential techniques in user modelling (Kobsa, 1990) and building adaptive systems (Brusilovsky, 1996). In the area of student modelling, such an approach is favourable not only because it allows modelling of the dynamics of student behaviour (Self, 1990) but also because it has potential educational gains in providing means for reflective learning (Bull 1997). A recent trend in student modelling focuses on building open and inspectable student models (see Morales, Pain, Bull and Kay, 1999). These projects reveal new computational aspects that need further investigation, such as maintaining a diagnostic interaction with/between human agents, externalising portions of the learner model and maintaining different views about the learner model. The aim of this paper is to present a comparative study of learner modelling systems that involve the learner in diagnosis and open the learner model for inspection, change and discussion. We outline potentials of this approach and refer to some problems that need further investigation. As far as the interactive diagnosis concerns student modelling in intelligent educational systems, this issue is briefly outlined in section 2. The following section discusses the approach where diagnostic systems involve learners in diagnosis. A review of research projects that apply this approach is presented in section 4. Finally, a summary that discusses some achievements and open issues is given.