Adaptive techiniques in web-based education

Abstract

This dissertation proposes the use of artificial intelligence methodologies and techniques for providing personalization of web-based courses. Almost all web-based systems try to carry out personalization in order to be more useful, more attractive, and more efficient in fulfilling user’s needs. However, personalization has a cost in terms of background operations, for instance in educational systems in terms of teacher’s effort. What is a reasonable compromise between sofisticated personalization methodologies and their realization? Is it possible to provide personalization with an acceptable effort for the domain expert responsible for contents producing? This dissertation focuses on these questions, considering in particular systems for web-based education, and proposes a methodology that aims to carry out a reasonable compromise between an effective personalization from the student’s point of view and from the teacher’s point of view, or, in general, from the user’s point of view and from the domain expert’s point of view. From the student’s point of view, personalization is provided on the basis of student’s knowledge and learning styles and guiding the student during the fruition of the course, like a teacher could do: proposing a sequence of contents suitable for the student at the beginning of the course and performing recovery strategies, during the fruition of the course, if the study does not proceed as it should. From the teacher’s point of view personalized courses are generated automatically, on the basis of the student model. The teacher is required to specify few metadata, necessary for characterizing learning materials, such as prerequisite relations and suitability of contents for a given type of student. The teacher is helped by a graphical interface, allowing a global vision of the course, and he can express didactic preferencies, such as the level of the course, according to the Bloom’s Taxonomy. The effort required to the teacher is as near as possible to his “way of thinking”: prerequisite relations are generally defined, even if implicitly, when a course is arranged; learning materials are tagged, according to the Felder and Silverman’s learning styles model. Ad-

iv herence to this model is not, however, a strict constraint for the teacher: the estimate of student’s learning styles is in fact updated taking into account the teacher’s tagging of the learning materials. In this way, relevance is given to the matching between the teacher’s tagging of the learning materials and the student’s way of learning: if the student studies a given material with success, the material is considered suitable for the student and his learning styles are updated towards the weights given by the teacher to that material. On the basis of the above-mentioned methodologies the LS-Plan system has been proposed. LS-Plan provides educational hypermedia with adaptivity; it has been integrated in the Lecomps educational hypermedia in order to carry out evaluations both from the student’s and from the teacher’s point of view. A layered and an as a whole evaluation, together with evaluations of teacher’s functionalities, have been performed and have shown positive results. Different approaches have been proposed in the literature for curriculum sequencing, that is “help the student to find an “optimal path” through the learning material”. According to the aim of providing support for teachers in performing personalization, a suitable system, LS-Lab, for comparing different algorithms has been proposed. LS-Lab provides a uniform environment in which several algorithms can be compared using the same input, i.e. the same set of didactic materials, the same sample student models and the same learning objective. The subjective comparison, made by teachers or domain experts, is supported by some metrics and by the visualization of the produced sequences. According to the necessity of providing an easy-to-use personalization for background actors, the personalization methodologies proposed for the educational domain have been applied also for cultural visits personalization. Analogies and differences between course personalization and cultural visits personalization have been detected and the framework for course personalization has been adapted and enhanced taking into account visitor’s interests.