An Intelligent Tutoring System Based On Learning Style And Probabilistic Inference

Intelligent tutoring system (ITS), as a descendant of earlier generation of Computer-aided Instructional system (CAI), has been one of the hottest topics in web-based education area. Most of the early CAI systems are very simple, displaying static learning materials one by one without any adaptive assistance according to the different learning preferences and knowledge levels of students. This situation lasted long until the emergence of ITS. The most valuable feature of ITS is the ability to perform different teaching strategies to teach students after identifying their knowledge level. This feature brings the tutoring system up to a higher level. However, another important aspect of education has been ignored: students have different strengths and preferences in the ways they take in and process information, which means they have different learning styles. When the learning styles of the students in a class and the teaching style of the instructor are seriously mismatched, the students will become bored and inattentive, or even drop out of school.In this thesis, I addressed this problem and proposed a B/S (Browser and Server) architecture intelligent tutoring system, called ITLP, based on the learning style theory and Bayesian network. Since students have different learning styles, it will be meaningful to develop a system that can intelligently identify the individual's learning style, and customize its user interface to present the style-matched learning materials for students to learn more efficiently and positively. Besides learning style theory, I also incorporate Bayesian network technique to track the knowledge states of students for performing adaptive teaching strategies. With the combination of the abovementioned mechanisms, the system can be more practical in teaching a student how to learn thoroughly and efficiently. I discuss how to use the learning style instrument to identify the learning style of the students and how to incorporate JavaBayes, a modeling tool, to build a Bayesian network to infer their knowledge states. Moreover, I detail the design architecture of ITLP and how to implement it. To decide whether ITLP is efficient in improving the learning experience of the students, the experiment was carried out on CS1 and 2 level students, and the result validates my hypothesis: it is a useful tool, in some sense, to teach CS1 and 2 level students the C++ programming language.