An intelligent tutoring system based on self-organizing maps - Design, implementation and evaluation

This work presents the design, implementation and evaluation of an Intelligent Tutorial System based on Self-Organizing Maps (neural networks), which is able to adapt, react, and offer customized and dynamic tuition. The implementation was realized in web environment (and technology). On the instructional design, the content, source of knowledge to be learned, has been modeled in an original way and is adequate to neural control. At the evaluation, two user groups have been compared. The first one (the control group) moves freely in the content, while the other group (the experimental group) is guided by the decision of neural networks previously trained from the most successful free interactions. Therefore, the control group serves not only as reference but also as source of good examples. Statistical techniques were employed to analyze the significance of sample differences between the two groups. Results from the interaction time have shown significant differences in favor of the guided tutor. All users guided by the intelligent control have performed as well as the best ones which had freedom to navigate through the content.