Bodong Chen

Crisscross Landscapes

Notes: Dawson, S. (2010). “Seeing” the learning community



Citekey: @Dawson2010a

Dawson, S. (2010). “Seeing” the learning community: An exploration of the development of a resource for monitoring online student networking. British Journal of Educational Technology, 41(5), 736–752. doi:10.1111/j.1467-8535.2009.00970.x


Another great article by Shane Dawson, which sports a very well-written intro and discussion. The SNA analysis conducted in this study is different from those in SNAPP studies, emphasizing ego-centric networks of high- vs. low-performing students in a large lecture (N > 1000). The results are interesting (and mostly unsurprising), showing that high-performing students had larger ego-networks, connected with peers with higher performance and teachers.


It shows how data about student ‘movement’ within and across a learning community can be captured and analysed for the purposes of making strategic interventions in the learning of ‘at risk’ students in particular, through the application of social network analysis to the engagement data. (p. 1)

an unprecedented opportunity for educators to visualise changes in student behaviour and their learning network composition (p. 1)

There is increasing recognition within the education literature of the importance of student social or peer networks to learning, and more specifically to learning performance (Cho, Gay, Davidson & Ingraffea, 2007), a sense of community (Dawson, 2008), class cohesion (Reffay & Chanier, 2003), and information and resource exchange (Haythornthwaite, 2006). (p. 1)

shifts the focus of studies from developing an understanding of the whole network community (entire social group) to examinations of the patterns of behaviour and tie connections (p. 1)

for individual or ego-networks (Saltz, Hiltz & Turoff, 2004). (p. 2)

recognition of the importance of social networks in education milieu (Brown & Adler, 2008; Tinto, 1998) (p. 2)

this paper elaborates a study that examined the types of relationships (networks) students foster in an online education setting and their impact on learning performance. Specifically, the study explored differences in network composition for lowand high-performing students in order to identify patterns of network behaviour that might have influenced learning, and showed how these findings could be applied as a guide to assist educators in the evaluation of implemented online activities. (p. 2)

The ‘social’ in education (p. 2)

Terms such as social learning (Brown & Adler, 2008), social networks (Haythornthwaite, 2002, 2006), social presence (Garrison, 2007; Gunawardena, 1995) and social architecture (Bogenrieder, 2002) have gained increasing prominence within the literature, testifying as they do to the relational nature and purpose of learning. (p. 2)

Historically, the increased focus on the social in education can be traced back to the works of Dewey (19381963) and Vygotsky (1978), who maintained that the process of learning is facilitated through individual participation in social interactions. (p. 2)

More than a decade ago, Alexander Astin (1993) foregrounded the importance of peer learning interactions by arguing that an individual’s peer network is the ‘single most potent source of influence’ (p. 398) on student growth and development. (p. 2)

Almost a decade later, Richard J. Light (2001) noted that a student’s propensity to participate within small study groups was an accurate predictor of academic success. (p. 2)

Marc Prensky (2001) and others (eg, McWilliam, 2008; Oblinger, 2004) have argued that ‘digital natives’, ‘millennials’ or the ‘yuk/wow generation’ are first and foremost social (p. 2)

learners. (p. 3)

In short, the ‘social’ is not the context around learning—it is the learning process itself. (p. 3)

While implemented teaching practices can facilitate an increase in the size and make-up of student networks, there are few methods providing instructors with a process for monitoring, visualising and evaluating the evolution of individual student networks. (p. 3)

Social Network Analysis Social Network Analysis (SNA) provides a valuable methodology for examining the patterns of interaction that occur within a group of actors (a network). As such, SNA draws on various concepts from graph theory and structural theory to evaluate network properties such as density, centrality, connectivity, betweenness and degrees. (p. 3)

Wasserman & Faust [1994] provide an excellent comprehensive overview of SNA (p. 3)

The flexibility and value of this methodology is reflected in the quantity and diversity of studies adopting SNA techniques. (p. 3)

In the teaching and learning context, the usefulness of SNA as a methodology to assess student social networks has also been well demonstrated. For example, Haythornthwaite (1999) used SNA to investigate the frequency and types of interactions between distance learners. The author noted that the implementation of small group work activities promoted the development of stronger ties between individuals and a larger, more complex network. Dawson (2008) identified correlations between SNA centrality measures and student sense of community, noting that degree and centrality were noted to be positive predictors of student sense of community. Additionally, the size and complexity of social networks has also been demonstrated to influence student learning performance (Cho et al, 2007). (p. 4)

the overall extraction and collation of social network data has to date been problematic (Reffay & Chanier, 2002). (p. 4)

Online education—untapped data (p. 4)

In online learning, practitioners guided by socio-constructivist principles have tended to rely on the use of social communication tools such as discussion forums, and more recently blogs and wikis, to generate the peer interactions necessary to promote a sense of community. (p. 4)

The collation and analysis of online learning management system (LMS) data have been utilised by a number of researchers in recent times to illustrate relationships between student user-behaviour online and learning and teaching. For instance, Campbell and Oblinger (2007) combined prior academic results with student online effort to develop a predictive model of student attrition. Dawson (2006) extracted discussion forum data to demonstrate correlations between message frequency and student sense of community. Morris, Finnegan and Wu (2005) noted a positive relationship between time spent online in class discussion forums and overall academic performance. (p. 5)

The study (p. 5)

The study described below incorporates SNA within an online learning environment to identify differences between individual networks (ego-networks) of highand low-performing students. In so doing, the study addresses the following research questions: • Are there significant differences in the network composition between highand lowperforming students? • Do high-performing students have larger social networks than their low-performing peers? (p. 5)

• Is instructor presence more prevalent in the networks of highor low-performing students? • How can the visualisation of student social networks aid pedagogical practice? (p. 6)

Data were collected from a large prerequisite first level chemistry unit (n = 1026) during the September to December 2007 teaching period. (p. 6)

Extracting network data (p. 6)

While many commercial LMS have student tracking capabilities as generic software features, the depth of extraction and aggregation, reporting and visualisation functionality of these data has been disappointingly basic or non-existent. Generally speaking, the student tracking data can report on the time spent online, number of pages visited and the number of discussion forum posts contributed or read. (p. 6)

visualisation of these data is commonly restricted to poorly organised tabular formats, or statistical graphs (Mazza & Dimitrova, 2007). (p. 7)

What is lacking in current commercial LMS tracking functionality is the ability to report on the frequency, depth and quantity of peer-to-peer interactions manifesting within the online learning environment and the provision of visual resources to aid instructor interpretation and thus be of immediate pedagogical use. (p. 7)

The resource essentially automated the process of extraction, collation, evaluation and visualisation of student network data available as discussion forum postings and interactions. (p. 7)

The information and communication technology (ICT) resource was developed in javascript using Greasemonkey (, a Mozilla Firefox browser extension ( (p. 7)

These extracted forum data are then exported into Netdraw (Borgatti, 2002), a third party social network visualisation tool. (p. 7)

Performing SNA (p. 8)

In the study, SNA was applied to the extracted communication logs (discussion forum activity) to determine the network complexity, key central actors and the level of relationships developed. (p. 8)

Thus, the category ‘high performers’ included all students ranked in the 90th percentile. Low-performing students were derived from the 10th percentile. (p. 9)

Each ego-network was interrogated for the number and type (staff, student) of connections. The presence of teaching staff was recorded for each ego-network, providing an indication of teacher accessibility and activity in both highand low-performing subgroups. Additionally, the mean grade of all actors in a specific ego-network was calculated. (p. 9)

Statistical analysis (p. 9)

descriptive statistics and nonparametric tests (p. 9)

The nonparametric statistical analysis employed in this study was the Mann–Whitney U-test. This statistic was used to assess the level of significant difference between the network size and the academic scores of established social ties between highand low-performing students (90th and 10th percentiles respectively). (p. 10)

Results (p. 10)

Network size (p. 10)

A Mann–Whitney U-test revealed a significant difference exists between the observed the network size (degree) of lowand high-performing students (Table 1). The results indicate that high performers develop larger social networks (p. 10)

Grades (p. 11)

highperforming students primarily develop connections with students of a similar academic capacity. The mean grade score for high performers networks was calculated to be 77.32% (SD = 14.04, n = 376) (Table 2). Similarly, low-performing students were also more inclined to foster online relationships with peers of a comparable academic score (mean = 59.79, SD = 21.3, n = 80). (p. 11)

Teacher presence (p. 11)

Examination of the actors associated with the egonetworks indicated that the teaching staff members were positioned in 81.7% of the high-performing and 34.61% of the low-performing student networks. (p. 11)

Discussion (p. 12)

the clear differentiation in social ties fostered between the two student subgroups supports the view that ‘similarity breeds connections’ (McPherson, Smith-Lovin & Cook, 2001, p. 415). (p. 12)

Not all students are responsive to networking possibilities. As Cho et al (2007) found, there are often profound differences in students’ communication style and impact on network composition. They noted that students with a ‘low willingness to communicate’ adopted different network strategies from their more willingly conversant peers. (p. 12)

However, the problem for low-performing students is exacerbated because their ability to develop a greater understanding of the course content is also impeded by the particular character of the network associations they develop. (p. 13)

The central premise of a learning community is that members will share resources, information and collaborate in order to assist one another in a common endeavour, presumed to be learning the content of a specific discipline or, as Brown and Adler (2008) noted, ‘learning to be a full participant in the field’ (p. 19). Despite the rhetoric and best intentions of educators to develop a sense of belonging and shared purpose among a group of learners, conflicting demands and priorities placed on students may result in more self-centred reasoning processes. (p. 13)

It is understandable that, while an institution clings to more performanceoriented goals, its students will focus more on their individual performance than the shared project of community-wide learning. In this context, students develop network ties that are tightly connected to providing direct benefit to the individual. (p. 13)

Not all social ties, of course, are of the same character. Granovetter (1982) maintained that social ties can be classified as either strong or weak. Individuals with strong ties are mutually dependent and share multiple resources. (p. 13)

Richard Florida (2002), author of ‘The rise of the creative class’, suggests that individuals no longer require or seek to build ‘strong’ social ties as these commonly represent more long-term commitments and therefore, constrained mobility. (p. 14)

The capacity to visualise both class and individual networks provides educators with a resource to better identify potentially at-risk students and to also monitor the allocation and direction of teacher support. (p. 15)