TEACHER MODEL FOR AN EDUCATIONAL ADAPTIVE HYPERMEDIA THAT SUPPORTS THE TEACHING OF DECISION-MAKING

This article analyses how a Teacher Model not only allows for the sharing of knowledge but also supports the emulation of teaching styles to recommend examples of network topologies for the teaching of decision-making skills in Network Design.
The data stored in a Teacher Model can be used as a basis to chose and recommend examples which characteristics are similar to the characteristics of the examples that a teacher uses. The examples used by the teacher illustrate the way she thinks and the way she teaches. We modelled the teaching style as the type of examples that a teacher uses most frequently.
Designers create computer-based educational environments, but it is the teacher who adapts these environments for teaching courses according to their particular interests. A Teacher Model is the information about human teachers and their personal manners of teaching a subject. A Teacher Model keeps a structured representation of the teacher’s knowledge about how to teach the expertise to solve problems in a knowledge domain.
More specifically, we use a Teacher Model to provide more suitable problem-solving support in a Web-based system or Educational Adaptive Hypermedia (EAH). An EAH needs to have teachers’ profiles to allow adaptation to needs, limits, and pedagogical goals of different teachers. The purpose of this system is to provide problem-solving support options such as the generation, testing and recommendation of examples and the classification and sharing of teaching materials created by different teachers.
If a teacher has never interacted with the Web-site, the system does not have information about her teaching style, and then, to solve this “start up” or cold-start problem, the system uses a collaborative filtering method that takes into account the similarities among teachers with similar characteristics. The system believes that teachers with similar characteristics can share teaching styles and it infers that examples used by the teachers in the selected group can also be used by the new teacher.
Based on a Teacher Model, we are using a recommender that reduces the complexity of finding an appropriate example for a particular teacher. Our system uses a probabilistic recommender supported by a naïve Bayesian classifier (a supervised learning method) to classify teacher’s characteristics and assigns the new teacher to the category in which the teacher’s data are most suitable. Once the teacher belongs to a category, the most frequently used class of examples in this category is used to recommend sessions and examples to the new teacher.
The recommendation in our system is also based on the automatic generation and testing of examples according to teachers’ preferences. If the teacher does not like an example recommended from a case base, the system can generate a new example according to the teacher’s teaching style. The generation, testing and recommendation of examples, that is a type of automatic creation of courseware material, depends on the information contained in the Teacher Model. Our system has a particular module, the Generator, which creates examples (network topologies). By means of a Bayesian classifier, this module verifies that any generated topology fits in the class of examples used by the teacher (the teacher’s teaching style). If the generator creates an example that is out of the scope that the teacher prefers, the example is aborted and a new one is generated, tested and recommended.