Abstract:

Nowadays, Computer-Based Instruction environments and eLearning Systems are gaining in importance. Among these systems we will pay special attention to Intelligent Tutoring Systems (ITSs). An ITS merges eLearning tools with Artificial Intelligence (AI) techniques in order to obtain the most suitable pedagogical strategy for each individual student. To accomplish this, the ITS makes use of a component called Student Model, which allows the system to estimate a student's cognitive state and, as a result, to react according to a Pedagogical Model. The key element within this framework is the Diagnosis Process, which allows the student's cognitive state to be inferred from observable data. As a consequence, both the Student Model’s quality and the ITS's adaptation capability depend on the quality of the Diagnosis Process. Currently, there are many proposals based on different theoretical frameworks to model the Diagnosis Process. Among the existing techniques, we propose the use of Bayesian Networks (BN) for their ability to represent and manage the inherent uncertainty of the real world. In this paper, we present our proposal of a Bayesian Student Model, analyse the influence of each element in the BN during the Diagnosis Process, and characterise the behaviour and sensitivity of the model under different settings. Finally, we show the proposal’s application in the “Programming Fundamentals” subject, which is part of the CS1 course at the Computer Science and Engineering Faculty of the University of Castilla-La Mancha.

Keywords:

Student Model, Diagnosis, Bayesian Networks, Intelligent Tutoring System.