Learning analytics tools and practices are central to e-learning. These tools are of concern to both industry and research, as they offer the opportunity for log-based, behaviorally driven evaluations of learning. MOOCs (Massive Open Online Courses) and other online learning endeavors are based on online conversations, and enable rich logging data collection and data mining based on learners’ behavioral patterns. At the same time, the lack of face-to-face interaction in online learning calls for further analysis and research of online interactions. Social Learning Analytics focus on how communities of learners create knowledge together.
Despite the wealth of data in formal learning and social platforms for learning communities, the theoretical basis of Social Learning Analytics is still in its infancy. We propose to use social learning big data to explain and assess the online learning process, based on theories of learning, information and computer-mediated communication.
We present a theoretical based methodology for the quantitative analysis of online discussions in learning communities. We suggest using network analysis methodologies and tools to assess the learning processes of a community as a whole and we present initial results of using such methodology in online learning communities.
In deriving a network of interactions from an online discourse, we use the theoretical framework of Social Constructivism and Connectivism to define the set of the network's nodes. Social constructivism (Vygotsky, 1978) views the participants themselves as the building blocks of collaborative learning. Connectivism (Siemens, 2005) tries to reframe learning under the digital age eco-system, adding the notion of a hybrid network, where information items and devices act as nodes in addition to human agents. Accordingly, we use Interactivity Theory and Assimilation Theory to define the edges in the learning network. Interactivity Theory (Rafaeli, 1988) focuses on interactions between community members to explain knowledge forming processes. In turn, Assimilation Theory (Ausubel, 1968) emphasizes the relations between information items in the process of meaningful learning. We contend that knowledge cannot be conceived as constrained to a certain entity (a single object or the mind of a single individual), nor in a certain point of time. Knowledge construction is a continuous process which takes place in a hybrid network of human and information agents.
In addition to presenting our theoretical-framed methodology, we are also applying network analysis to specific networks derived from higher education learning communities. We examine the relation between the learning process and the community's network topology in real life settings. We set measurable parameters to quantify the illusive dynamics of the social learning process.
2. Research Questions
We propose an analysis of the dynamics in topological parameters of learning networks as indicators of the social learning process. We measure these parameters each week during three months academic semesters and try to explain how the change in network parameters (such as centrality) can explain the learning process dynamics. We thus investigate questions concerning the relation between network parameters and the learning process. Examples of such questions include: (1) Are there online learning tipping points? i.e., where in the growth of human and information networks (Rechavi & Rafaeli, 2012) do the topological measures reach inflection?; (2) How different is the growth pattern of the sub-network of 'passive' interactions from that of the sub-network of 'active' interactions (e.g., contributing posts)? (3) Do different learning designs and moderation styles of online discussions affect the learning process? (4) Can we analyze the classification of semantic relation types between information items, as given by the learners, in order to realize the type of discussion taking place during the learning process?
3. Data and Analysis
We examined data collected in several cases of online discussions conducted throughout semester-long higher education courses. Those communities used Ligilo (Kent & Rafaeli, 2015; Kent, Laslo & Rafaeli, 2016) , an online knowledge management tool for learning discussions.
The learning network as formulated from the communities' online discussions, is composed of two types of nodes and three types of edges (interactions). Nodes represent: (1) human agents and (2) information agents (posts contributed). Edges represent: (1) interactions between human agents and information agents (such as view, vote, relate to, write); (2) interactions among human agents (such as user follows user); and (3) interactions among information agents (such as 'example of', 'in contrary to', 'makes me think of').
Since the learning process can be depicted through the change in the structure of the learning network, we use SNA methodology to create a cumulative set of snapshots, showing the week-by-week change of the network. The cumulative map best describes the development of the learning network. We analyze the change and explain the results.
4. Initial Findings
Our initial findings include insights concerning the hybrid network, and the information items network. For example, we found that learners prefer to use semantic relations between posts using associative terms (“It reminds me”) and not functional terms (“Important for the TEST”). Another early finding is a quantitative tipping point at the third week, when the pace of new nodes and interactions grow dramatically. The network presents a preferential attachment process (Barabási 1999) where nodes are joining the network through well connected nodes.
Further analysis is still undergoing, and findings regarding the differences between learning communities design and moderation levels are still in progress these days.
To summarize, our paper suggests a methodological connection between learning theories and online social learning analytics, applied in real life settings. We offer a quantitative, log-based, formative assessment approach to learning, which assess social learning, as an alternative to the traditional individual learning assessment, such as manually grading assignments and exams.
Ausubel, D. P. (1968). Educational psychology: A cognitive view. New York: Holt, Rinehart and Winston.
Kent, C., Laslo, E., & Rafaeli, S. (2016). Associating interactivity in online discussions with learning outcomes. Computers & Education (In Press). Retrieved from http://www.sciencedirect.com/science/article/pii/S0360131516300537
Kent, C., & Rafaeli, S. (2015). Network-structured discussions for collaborative concept mapping and peer learning. IBM Journal of Research and Development, 59(6), 1–14.
Rafaeli, S. (1988). Interactivity: From new media to communication. In R. P. Hawkins, J. M. Wieman, & S. Pingree (Eds.), Advancing communication science: Merging mass and interpersonal processes (pp. 110–134). Newbury Park, CA: Sage.
Rechavi, A., & Rafaeli, S. (2012). Knowledge and Social Networks in Yahoo! Answers. In 45th Hawaii International Conference on System Sciences (pp. 781–789). IEEE Computer Society.
Siemens, G. (2005). A learning theory for the digital age. Instructional Technology and Distance Education, 2(1), 3–10.
Vygotsky, L. S. (1978). Mind in Society: the Development of Higher Psychological Process. Harvard University Pres.