University of Siegen (GERMANY)
About this paper:
Appears in: EDULEARN18 Proceedings
Publication year: 2018
Pages: 6790-6800
ISBN: 978-84-09-02709-5
ISSN: 2340-1117
doi: 10.21125/edulearn.2018.1610
Conference name: 10th International Conference on Education and New Learning Technologies
Dates: 2-4 July, 2018
Location: Palma, Spain
In our modern world, ‘knowledge’ has become the key to development. Knowledge transfer methods play a considerable role in enabling learners to utilize the knowledge they receive. Traditional teaching methods are still widely used, relying on rigid learning processes. These methods focus on providing knowledge to a group of learners at the same time, which takes into consideration only a small portion of the individual student learning needs. With the wide spread of learning materials provided online; and the growing diversity of learner needs based on the diversity of the work market demands, a personalized learning approach may provide benefits beyond traditional teaching methods, by adapting to each learner’s needs. An important approach for this adaptation is linking each learner with the most suitable learning materials. The recommendation of certain learning material can be influenced by several sources, such as the student’s personal judgment of other learning materials, the assessment of other similar students of this material, the preferences of teachers and experts, or the requirements of the domain of knowledge itself.

The objective of this research is then to answer the question, how to fuse and balance different recommendation approaches and perspectives to give students a suitable recommendation on which topics to learn in a domain of learning. The assessment of recommendations and their influence on the student is essential to ensure that each individual learner is linked to the suitable recommending source, which in turn will increase the acceptance of the recommendation and the trust that the learner has in the recommender itself.

In this paper, three recommending approaches are being implemented, analyzed and compared in order to assess the final recommendation proposed to the learner based on his/her individual needs. The first recommender is based on the student’s learning preferences. These are acquired through a conversational engine that handles the interaction with the student. The second recommender reflects the preferences of the teacher, incorporating the expert opinion into the recommendation. It is designed to provide the student with the set of teachers’ preferences and influence the conversation between the learner and the recommendation system. The third recommender is built upon a domain network. It represents the internal links between the topics in the learning domain, which in turn affects the recommended learning path for each student.

The three recommendations are presented to the student, where the student’s feedback is assessed in order to weigh each approach and the level of satisfaction it offers. This assessment is intended to enhance the personalized experience of the learner and increase the overall recommendation acceptance. An evaluation process is designed and combined with the recommender in one framework. This framework is intended to close the loop of the recommender system and provide relevant information for assessing different recommendations.

Since the interaction process between the student and the recommender system may affect the feedback required from the learner, the conversational feedback loop provides students with a human-like interaction that enhances the experience of the learner, encourages him/her to provide more information and ensures a user friendly environment that helps the student to focus on the recommendation itself rather than the software interface.
Recommender Systems, Technology Enhanced Learning, Learner and Teacher Preferences, Educational Domain Models, Collaborative Filtering.