Experiences in implementing constraint-based modeling in SQL-TUTOR

Abstract

The problem with most student modeling approaches is their insistence on complete and cognitively valid models of student’s knowledge. Ohlsson [10] proposes Constraint-Based Modeling (CBM) as a way to overcome intractability of student modeling, by generating models that are precise enough to guide instruction, and are computationally tractable at the same time. The paper presents our experiences in building SQL-Tutor, an ITS built upon CBM. CBM is extremely computationally efficient. State constraints, which form the basis of CBM, are very expressive; we have encountered no situations where constraints were unable to diagnose student answers. The time needed to acquire, implement and test a constraint is less than times reported for the acquisition of production rules. The initial evaluation of SQL-Tutor proved the validity of design and appropriateness of CBM.