Abstract:

Learning analytics (LA) and educational data mining (EDM) are young, but rapidly developing research areas at the intersection of learning theory, data science, computer science, and statistics. Main goals of LA and EDM are to understand learners, understand and optimize learning and the environments in which it occurs, and support decision making in education. In the early days, LA and EDM research was cross-disciplinary – viewing learning from the perspective of data analytics and data mining. Soon it was realized that a multidisciplinary approach is needed, combining approaches from both learning theory and data science. And the field is steadily moving towards more interdisciplinary and transdisciplinary approaches, integrating and synthesizing knowledge and methods from constituent disciplines, and enriching them in the process. Drawing researchers from different backgrounds necessarily brought together different epistemologies and methodologies. Recent research by Baker and Gašević identified within the learning analytics four communities with different philosophical orientations.

The idea of communities of thought in science dates back to 1935 and Fleck’s book Genesis and development of a scientific fact. More recently, in 1980’s, Wenger promoted idea of communities of practice, an informal network of people who co-create, share, and communicate knowledge and methods of dealing with a common set of problems.

Due to the rapid pace of development, there is a need to gain deeper understanding of the evolution, current research and applications of LA and EDM. Newcomers to the fields would benefit if they had access to thematically and methodologically organized overview of these research areas. We believe that this could be achieved by identifying the LA and EDM communities of practice.

In this preliminary research we focus on four thematic areas: resources, use cases, ethical issues, and self-regulated learning. Scope of the research is further narrowed to the use of LA and EDM in higher education with hybrid or blended learning practices.

Main objectives of our research are to identify clusters of papers sharing similar problems and approaches to their solving, and to analyse co-authorship networks emerging from these papers. Combined results of these analyses provide information about common problems areas, methodologies, and cooperation networks, which are main characteristics of communities of practice. Latent Dirichlet allocation (LDA) for topic modelling is used to create clusters of papers. Cooperation patterns within the identified communities of practice are further analysed using social network analysis (SNA).

Keywords:

Learning analytics, educational data mining, communities of practice, text mining, topic modelling, social network analysis, higher education.