Calculate Learners' Competence Scores and Their Reliability in Learning Networks.

Intelligent tutoring systems rely on learner models in order to recommend useful learning resources. Learner models suffer from incomplete and inaccurate information about learners. In learning networks, learners share their knowledge and experiences online and collaboratively perform problem-solving tasks. In this paper, we present an approach that calculates learners' competence scores based on their contributions and social interactions in a learning network. We aim at more comprehensive and accurate learner models, which allow more suitable recommendations of learning resources. Competence scores range from 0 to 100 points, each associated with a confidence level representing the calculation's reliability. For evaluation, we conducted an experiment with 14 master students at university. The results show that our approach tends to underestimate competences, while it calculates 54% of the scores accurately. Student feedback suggests to apply our approach for recommending future courses as well as forming student groups.