Abstract:

An intelligent tutoring system (ITS) is an educational teaching and training software, which has the task to – abstractly speaking – transfer knowledge from a database to a human learner. For this purpose, it needs to copy some of the behaviour of the teacher, to filter the knowledge so that it matches the student’s needs and to explain it in a way that is easy to understand. If we focus on the knowledge an ITS has to have before it can adapt to a student, we have the expert knowledge, as the knowledge that we want to transfer, and the pedagogical knowledge, which explains how to do it best for this certain student. The expert knowledge is described very detailed in other papers, the pedagogical knowledge is a newer component of ITS which is sometimes lightly added to the expert knowledge or even ignored. Whereas the expert knowledge is a – usually fact and rule based – amount of knowledge, the pedagogical knowledge can be seen as the semantic annotation of the expert knowledge, i.e. adding context, additional rules and rationale to the facts and rules. With this additional knowledge, an ITS can “work” with the expert knowledge: it can structure the content according to case-based training, to diagnostic reasoning, but also for example to develop a logical coherent lecture. The latter is an example, where the ITS uses a pedagogically founded sequentialization of the expert knowledge to create lecture material. The problem is that the pedagogical knowledge is bound to the concrete expert knowledge and therefore, usually the reuse of the pedagogical knowledge is not allowed. Knowledge engineers have to work with teachers to add these annotations to existing expert knowledge databases without merging them to one. That means many hours of manual labour to create this bound pedagogical knowledge. This knowledge engineering is the classical bottleneck in all ITS – and the main reason, why the pedagogical knowledge is missing in quite a lot modern ITSs.

Our approach tries to fill the gap between and to overcome the bounded restrictions of a combined expert- and pedagogical knowledge model. After analyzing the pedagogical knowledge component in existing ITS, we developed the idea to reduce the amount of work by automating the process of creating bound pedagogical knowledge. This can be done by giving the ITS a more abstract and general understanding of pedagogical knowledge. With our background in didactics and instructional design, we investigated a lot of pedagogical rules. The result was, that the pedagogical knowledge could be constructed by developing rules with which annotations can be created. The rules have a scope, which limits them to a domain or subdomain, but they are not fixed to a certain expert knowledge database. To use such rules in an automated way to create bound pedagogical knowledge and be exchangeable they have to be formally defined. We have investigated our approach in some examples and have been able to show, that the general model is working. In the paper, the model, a software prototype and first insights in applying the method to examples will be shown.

Keywords:

Pedagogical knowledge, Automation.