E-Learning Social Network Analysis for Social Awareness by Niki Lambropoulos
Presentation delivered at the Images of Virtuality Conference
Athens, 23-24 April, 2009
http://www.imagesofvirtuality.org/
Keynote @VocTEL: Vocational Dreams to Career Reality: Technology Enhanced Lea...
E-Learning Social Network Analysis for Social Awareness by Niki Lambropoulos
1. E-Learning Social Network Analysis for Social Awareness Dr Niki Lambropoulos London South Bank University HCI Education Research Fellow Images of Virtuality, Athens, 23-24 April, 2009 http://www.imagesofvirtuality.org/
13. Density (Quantiification) Table 6.5.7.1-1. Group Network Cohesion: Density & Reciprocity Fahy’s density does not consider interaction time and is highly sensitive to group size (2a/N(N-1)) GROUP NETWORK COHESION: DENSITY & RECIPROCITY GSN Research Pool All E-learners All E-learners Total nodes 698 122 81 73 Density (matrix average) 1.0872 0.0256 0.0470 0.0418 Standard deviation 2.0167 0.1819 0.2376 0.2226
It was the first time the e-learners saw such tools so they didn't understand their use; most common comments were: nice tools and are more useful to e-tutors. The Pedagogical Usability scores were just about the average.
Density was calculated with Fahy and colleagues’ density formula, 2a/N(N-1) (a=interactions, N=number of participants). Messages density was almost double in Moodle@GSN (0.19) compared with the research pool (0.1). However, Fahy’s formula does not consider interaction time, and is highly sensitive to the size of the group. Density is the proportion of possible links in network as it is the ratio of the number of links present in the network, to the maximum possible links. Density was evaluated by the adjacency connection reports in UCINET. E-learners’ density is rather low, 0.0256 in Moodle@GSN, however stable in the research pool (0.0418). This means that 2.6% in Moodle@GSN and 4.7% in the research pool of all possible links were present; however, there was an increase in density. The participants actually recognised their limited participation; they said they were not as active as they wanted to be. In addition, the 2 highest posters in Moodle@GSN influenced the groups’ density level; this means that the actual increase in participation was almost doubled (0.0418 - 0.0256= 0.0162). This was also evident in the collaborative e-learning episodes text richness, as it was doubled in the research pool (see Table 6.4.2-2).
Reciprocity is the proportion of all reciprocal ties to the total number of actual ties. There were 6 reciprocal ties in the first study as shown in the left hand diagram and 10 more distributed in the second case study as shown in the right hand diagram.
A clique is a subgroup, a set of actors with each being connected to each other as a maximal complete subgraph of three or more nodes (members) adjacent to each other and there are no other nodes in the network that are also adjacent to all of the members of the clique. Cliques may overlap, that is a forum member (node) can be a member of more than one clique. The results presented in the following table are cumulative and refer to cliques created by 3, 4, 5 and 6 participants. Most cliques were created by 3 participants in both environments. The e-tutors dominated the cliques gathering up to 6 participants. The cliques were developed without any intervention by any of the participants, e-tutors or myself. It is interesting to note that the top scorers had inter-clique connections. When the cliques increase, the social network remains active and thriving, especially if e-learners interact with other e-learners who did not appear in a clique before; these are the activated lurkers. In other words, the absence of cliques could have indicated a lack of clustering which means weak ties. Weak ties> weak relationships> weak communities..
Global SNA is defined as the actors’ similarity in patterns of relations to others by exhibiting similar communication behaviour. It presents a different clustering view within a human network; actors must not be thought as unique persons, but as examples of categories (sets of actors) who are in some way, "equivalent”. Two actors (nodes) are said to be structurally equivalent if they have identical ties with themselves, each other and all other peaks (vertices). The CONCOR technique (CONvergence of iterated CORrelations; White et al., 1976) uses dendrograms (tree-diagrams) for hierarchical clustering. The e-learners had 7 first level and 4 second equivalent communication behaviour in the first case study and 5 first and 2 second with 1 solo-actor position in the second.
Aneesha created a process the get the data in minutes.
In VIT Nodes the individuals are represented as circles (nodes), the direction of the messages is indicated by an arrow and the number represents the number of messages. P37 was the information broker in this CeLE. The reciprocal tie with O2 was an argument. She also responded to her own message a couple of hours later after the argument with O2. Most participants were replying to P37 and two of them talked to each other. It is interesting that this CeLE was developed by different individuals with only two interlocutors exchanging 2 messages. In other words, the discussion was a collaborative activity between 7 individuals. VIT Centrality provided a different viewpoint. In VIT centrality P37 is clearly located in the middle of the e-learning social network. VIT centrality also indicates the response time space related to geodesic distances between the participants. As a central connector and information broker she moved the knowledge around leading to a new proposition by taking into account her co-learners responses even though they appeared as low activity e-learners (i.e. only O2 was an e-tutor).
The learnability scores were low (2.7). The low learnability rating may explain why the tools were infrequently used (2.4).