This work focuses on Web-based intelligent instructional systems and research issues associated with the development of student modeling in an adaptive hypermedia system. The framework is iMANIC (intelligent Multimedia Asynchronous Networked Individualized Courseware), in which courses originating from existing video-taped lectures provide an initial set of slides, audio, and class notes. However, the existing course structure is initially linear, which, though usable, is not optimal for a WWW presentation. Web courses are used asynchronously and thus can provide a more individualized and interactive learning experience than can live courses. Therefore, we investigate ways in which personalized instruction can be delivered via the WWW. The domain organization used in iMANIC supports a non-linear, individualized course. However, once we introduce a non-linear topic structure, the “lost in hyperspace” problem might arise, in which students become confused about what to study next and how to remember where they have been. To combat these problems, adaptive navigation techniques are used to help guide the student through the course material. The original class material is presented so that each student sees the same content. This does not take into account learning differences of individual learners. However, iMANIC can consider those differences and adapt the information presented to each user. This adaptive content is achieved through a two phase approach which considers the user's level of understanding and the content that matches the user's preferences. A Naïve Bayes Classifier is used to learn the student's preferences by observing what type of content he chooses to see. An empirical study of the iMANIC system was conducted during 2000/2001 with 24 students learning Unix Network Programming. Results from this study show distinct differences in students' learning styles and provide evidence that using the same teaching strategies for each student cannot adequately support all students. This is demonstrated through two examples. The first shows that there is not a consistent direction for the correlation between time spent studying and quiz performance. The second shows that using the same parameters for the Naïve Bayes Classifier for every student results in poor overall performance of the classifier.
| Identifer | oai:union.ndltd.org:UMASS/oai:scholarworks.umass.edu:dissertations-3591 |
| Date | 01 January 2001 |
| Creators | Stern, Mia Keryn |
| Publisher | ScholarWorks@UMass Amherst |
| Source Sets | University of Massachusetts, Amherst |
| Language | English |
| Detected Language | English |
| Type | text |
| Source | Doctoral Dissertations Available from Proquest |
Page generated in 0.013 seconds