21 February 2012 I Valentina Garoia
Social Network Analysis and network visualisations: a case of eTwinning
Teachers’ online networks can be considered as a rich source of social capital, the key for such opportunities to emerge is collaboration and interactions among participants. Network theory and Social Network Analysis provide novel approaches to study teachers’ online collaboration networks. For this purpose, the data extracted from the eTwinning platform have been used as a case study. This section provides an introductory overview of such network theories and Social Network Analysis (SNA). The Tellnet reports, “Data management of large-scale lifelong learning data” and “Major trends arising from the network,” give an analysis of eTwinning through novel tools such as network theory, Social Network Analysis (SNA) and Information visualisation.Social Network Analysis
Social network represents interdependencies between the nodes and ties of which it is created. They can, for example, be friendship, a common interest and relationships of knowledge exchange. These connections can be created in a traditional face-to-face setting or with the help of Information and Communication Technologies and the Web 2.0 tools.
The history of social network analysis developed in the kinships studies in the 1950’s urbanisation studies in England [http://en.wikipedia.org/wiki/Social_network#History_of_social_network_analysis]. Through these relationships, a graph of the network can be created. By studying the network graph and its shape, the value of the network for the individuals who participate in it can be studied.
Rather than treating individuals as units of analysis, social network analysis focuses on how the structure of ties affects individuals and their relationships and how it can be used to better understand, for example, social influence and social capital [http://en.wikipedia.org/wiki/Social_capital], the value of social relations within the network, and in this case, a collaboration network for teachers’ professional development.
In eTwinning, an individual teacher or a school can be considered as a node and the various activities between them as network relationships, i.e. ties. Examples of ties that can be found in eTwinning are naturally the exchange that takes place between teachers who collaborate on an online-based project together. A social network created through project work is called project collaboration network, Figure 1 gives an example of visualisation of such network.
Figure 1: The first ever network visualisation of the eTwinning project collaboration network was created in 2009 (Breuer et al., 2009)
Apart from the project collaboration network, other types of network can be identified in eTwinning. A number of them could be elaborated as the following:
- Contact-list network where the ties between nodes are contacts that eTwinners have added to their contact-list for potential project partnerships;
- Journal-posting network where the ties between nodes are the comments that eTwinners have left on each other’s online Journals; Blog network where the ties between nodes are the comments that eTwinners have left on each other’s blogs;
- Guest book commenting network where the ties between nodes are the comments that eTwinners have left on each other’s Guest book.
New understanding gained from Social Network Analysis
The very first data sets from the eTwinning platform were generated in June and in November 2010 by the Tellnet project partners. The data model used for these data extractions contain, among other things, information about teachers who register on the eTwinning platform and some of their activities, such as participation in projects, number of contact that teachers have collected, their communication patterns, etc.
In order to understand better the current state of the eTwinning network, a set of relevant questions was constructed for the use of SNA to study eTwinning. Based on these questions, the process of conducting the appropriate analyses will provide new insights. In the following part, some of the first questions are elaborated.
Questions selected for the first round of Social Network Analysis to test the analysis tools.
Question 1: When looking at the project collaboration network, is it possible to divide the network into sub-communities and if so, what is their relation to the rest of the project collaboration network?
Even if the project collaboration does not constitute the most important part of eTwinning since 2008, studying the Project collaboration network, its structure and core using the SNA measures gives interesting insights into how possible new mechanisms could be created to help other networks to grow in the future.
Through the analysis, we were able to identify 2776 separate clusters (see Table 1). These clusters are formed through eTwinners collaborating in projects. First observations show that there are four (4) gigantic clusters that create the main core of the eTwinning project collaboration network. The biggest one contains 8807 eTwinners, two separate clusters with about 3000 eTwinners and one of size of 1172 eTwinners.
Table 1: eTwinning network clusters
Apart from the gigantic clusters, there are small clusters (n=2772). As seen in Table 2, 2’627 of them consist between 2 to 9 eTwinners. It seems that the small clusters are those of people who collaborate only on one project during the time they have been part of eTwinning, most likely the cluster size corresponding to the size of the project partners.
What we can understand from the clustering formation is that, for example, in the biggest component there is a group of eTwinners who have collaborated with each other in a high number of projects where partnerships create complex ties between them. Moreover, we see that there four sub-communities in the core of eTwinning.
Lastly, we can calculate the modularity of the clustering. The modularity indicates the quality of the cluster, a fraction of any node's connections within its cluster (internal edges) and its connections to other clusters (Pham et al., 2011). Empirical observations indicate that a modularity greater than 0.3 correspond to significant community structures. In our analysis, we observe modularity of 0.4, therefore corresponding to significant community structures.
Question 2: When looking at the project collaboration network, how dependent is the eTwinning project network structure on a small core group of eTwinners?
The analysis was done based on the projects eTwinners participated in at the time of the snapshot, in mid 2010. eTwinners who did not participate in project collaboration were excluded from the analysis. Figure 2 shows a typical degree distribution that follows Power Law (http://en.wikipedia.org/wiki/Power_lawa), therefore indicating that the project network is scale-free. In a scale-free network one can usually observe a few big hubs followed by many small clusters.
Figure 2 : Project Clustering vs. Degrees
This means that the project collaboration network is dependent on core eTwinners that can be seen as bridges (hubs) between different clusters. Nodes with a higher degree tend to have a lower clustering coefficient (clustering decreases when degree increases). That means lower degree nodes are placed in dense groups (clusters) and these clusters are connected via hubs (nodes with high degree). However, as the betweenness is quite low (less than 0.1) there are apparently no super-hubs who exclusively connect the clusters. Clusters are typically connected via several hubs. In conclusion, although eTwinning is dependent on a core group, this is a large and well connected group.
Question 3: Over the years, how many eTwinners have gone inactive and were these eTwinners individuals who were connected through the project collaboration network?
The eTwinning platform uses different indicators to calculate “inactive” teacher, for example, if they have not logged in onto the eTwinning platform during a predefined period of time. At the time of the snapshot, in mid 2010, out of the 114’020 teachers, there are 2750 individuals who have been flagged as “inactive”, resulting to 2,4% of all participants. We can observe that during their active time in eTwinning, 1137 (41.4%) of now “inactive” teachers have been part of projects and therefore can be found in the project collaboration network, 955 (34.7%) are in the messaging network and 123 (4.4%) are in the blog network. The rest of the inactive teachers, 1123 (40.8%), did not have any recorded activity in these networks.
The degree and clustering coefficient was calculated for these teachers on those three networks. From the degree distributions, we can see that they follow Power law. Actually, inactive teachers seem just a sample of the same distribution of the whole network. This distribution also holds up when we constructed a network based on the blogs or the emails. The fraction of teachers who have clustering coefficient equal to NaN (means that they have only a connection - degree = 1), is 17,5% (project collaboration network), 49,01% (blog network) and 63,41% (email network). 41.4% of the inactive teachers do not have any activity in these (project, blog or email). Even for those who took part in various networks (projects, blog or email), they are quite isolated (as they have low degree and are placed in small, possibly disconnected, groups).
Question 4: eTwinners can create lists of MyContacts on their Desktop adding interesting people to the list for possible future collaboration. Is there any evidence that teachers have added people from different countries in their contact lists?
As eTwinning by nature promotes cross-border collaboration, we also find that in “MyContacts”, eTwinners overwhelmingly have added people from other countries than that of their own. If the creator of the list has a value of 0, it means that all contacts are from other countries, and 1 means that all contacts are from the same country. The mean for all eTwinners who had “MyContacts” is 0.16, indicating a strong preference to having eTwinners from other countries in the list.
Download the full papers here:
RWTH Aachen University, Data management of large-scale lifelong learning data, Tellnet Deliverable 2.1
Fetter, S., Berlanga, A. J., Cao, Y., Sloep, P., & Vuorikari, R. (2011), Major trends arising from the network, Tellnet Deliverable 3.1
M. C. Pham and R. Klamma. The Structure of the Computer Science Knowledge Network. Proceeding of the 2010 IEEE International Conference on Advances in Social Networks Analysis and Mining (ASONAM 2010), 9-11 August, 2010, Odense, Denmark.
Breuer, R., Klamma, R., Cao, Y., & Vuorikari, R. (2009). Social Network Analysis of 45,000 Schools: A Case Study of Technology Enhanced Learning in Europe. Learning in the Synergy of Multiple Disciplines (pp. 166-180). Retrieved from: http://dx.doi.org/10.1007/978-3-642-04636-0_18
Web Editor: Valentina Garoia
Last changed: Tuesday, 21 February 2012
Last changed: Tuesday, 21 February 2012