Abstract:

Personalised learning requires a tailored system that helps a learner to learn according to his/her profile. Personalised learning objects, learning activities, learning methods, etc. play an important role for learning personalisation. One of the possible solutions for personalized learning is to integrate a pedagogical agent into the learning environment. In traditional educational approaches, the concept of pedagogical agent is closely related to the role of a teacher, or the role of a learning companion, as well as to the so-called animated pedagogical agents, which are the part of intelligent tutoring systems. Researchers select different criteria for the agent’s functionality. Numerous studies have evaluated the impact of pedagogical agents on students’ learning outcomes. For example, there are pedagogical agents, which provide personalisation according to the learning style, learning activities, etc. However, so far, there is insufficient knowledge regarding the use of these agents in the real educational settings.

In this paper, we define the pedagogical agent as an intelligent component acting within the learning environment. The role of the agent is to collect and analyse data retrieved through learning processes and to form recommendations for user (learner or teacher). We propose the two-level model of a pedagogical agent. The function of the agent is to support the learner’s decisions in obtaining the most relevant paths in the process. The basis of our model is the use of the educational data mining techniques. The data is collected through learning processes while user is acting in the learning environment. We define criteria for data mining processes.

One group of criteria include
(i) sequences of content chunks;
(ii) learning pace taken from the real learning paths in the data collection.

The other part of the criteria includes (i) metadata, which forms annotations of learning objects and place these annotations into the personal generative library. We use a set of the known data mining techniques for educational purposes, including clustering, decision trees, etc. We have tested our model by implementing case studies in the real setting.

Our model ensures:
(i) an effective formation of the relevant sequences of personalized learning paths and
(ii) gives suggestions for upgrading existing annotations (as a part of metadata) in the personal generative library.

Integrating of the proposed pedagogical agent into STEM-driven intelligent learning environment for Computer Science education is our future work.

Keywords:

Pedagogical agent, pedagogical agent model, personalised learning, personalized learning path, educational data mining techniques.