Relationships are the connective tissue that binds communities of practice together. Social ties connect a new teacher with a mentor, a special education counselor with her counterpart at another school, and a group of school media specialists collaboratively developing lesson plans for their district. Though we instinctively realize the importance of relationships, they have historically been hard to see. Individuals know their own connections but often have a myopic view of the larger social space they inhabit. Even community administrators are often unaware of the many discussions and relationships among their community members.

As individuals and communities of practice increasingly adopt social media, data about who knows who can be captured, visualized, and analyzed to give us a better understanding of the larger social topology. Exploring this social graph can provide actionable insights to community administrators, marketers, and even community members. Until recently, the exploration of social network data was an important yet esoteric enterprise, limited primarily to Ph.D.s and those with extensive technical know-how. We have been working for several years with a group of colleagues across the nation (see SocialMediaResearchFoundation) to help make social network analysis more accessible to everyday users. Our efforts have been encapsulated in NodeXL (see image below), a free and open-source Excel plug-in that can be used by anyone who can work with a spreadsheet.

Example 1 scoial network graph

NodeXL helps novices acquire network datasets from social media channels like Twitter, Facebook, and YouTube, analyze them using powerful social network metrics, and visualize them using state-of-the-art techniques that help identify important individuals, cliques, and trends. Though some effort is required to internalize “network thinking” and get familiar with NodeXL, our experience teaching students and professionals of all experience levels has convinced us that even nontechnical novices can quickly gain actionable insights about their own networks and those that underlie the communities they help cultivate. (See this paper on the topic.)

Enough talk. Let’s see what a social network looks like and what insights can be gained from it.

Example 1: @openednews Followers

Example 2 social network graph

The image above is a snapshot of the network of Twitter accounts that followed the Twitter account @openednews. (See this article for a complete discussion.) It provides @openednews with a glimpse of the hidden connections between their followers, helping them to
  • Identify subgroups of people in the network who are well connected. This particular graph shows three main subgroups (i.e., clusters) identified by colors and grouped into separate boxes (blue, red, and green). These subgroups are automatically identified by the fact that they are more highly interconnected with each other than with the rest of the network. They represent cliques of people or organizations that know one another. In this case, the green subgroup includes prominent education news outlets, the blue subgroup includes research-oriented organizations and individuals, and the red subgroup includes many individual educators, researchers, and education-focused nonprofits.
  • Identify important individuals such as those labeled. In this graph, the size of the node is based on the number of Twitter followers, one measure of importance. Notice that some Twitter accounts (e.g., jolinarodriguez, creativecommons) have many Twitter followers but are not well connected to others following @openednews. In contrast, others like usatodaycollege and futurelabedu are popular among @openednews followers, as well as on Twitter globally. Other users have few Twitter followers globally (i.e., they’re small) but are very well connected to other @openednews followers (e.g., bon_education, oercommons, 4cinitiative, edinnovation). Network analysis provides many measures of importance (called “centrality”), for example, popularity (i.e., number of followers), betweenness centrality (people who span across subgroups such as eifdotorg and edinnovation), and eigenvector centrality (people who are well connected to “popular” people). @openednews could use this information to identify people whom they might personally contact or cultivate a relationship with.

Example 2: Classroom 2.0 Assistive Technology Forum Discussion

Node XL screenshot

The image above shows users (circles) who have communicated with each other (lines) via the Classroom 2.0’s forum on Technology in Special Education. (See this book chapter for a complete description.) Users who post a lot are larger and users not official members of the group are light blue. Thick lines indicate repeated interactions. Similar reply graphs can be created for forums, e-mail lists, wall post comments, blog comments, and other types of conversations. These graphs helps us to

  • Evaluate group cohesion. This forum shows a very tight-knit group of core contributors who often interact with each other. As is typical in these settings, a small group of people are most active, and most people are only peripherally involved. This shows a healthy community structure.
  • Identify social roles. People use the forum in different ways, as the user labels indicate. Some users predominantly answer questions, others start conversations that others reply to, and still others actively contribute in both ways. In this example, the core members are a nice mix of these different types, which is not always the case. In this example, we have highlighted two users who are already answering many questions yet are still not members (see potential users 1 and 2). A group administrator may want to invite these users to join the community.

As these examples show, network analysis can be applied to a variety of contexts and questions. Network ties may represent Twitter Follow relationships, Facebook Friend relationships, conversational acts (discussion forum and e-mail replies), and a host of others. Analysis may focus on identifying important individuals, subgroups, social roles, and group cohesion. The real power comes not from looking at one of these graphs, but from seeing how they unfold over time. Did our community event inspire more group cohesion? Did our formation of subgroups reduce across-group discussion? Has our mentorship program shown an increase in friendships and communication between new members and core members? The key to all these questions is that they focus on the connections between people, something we have only now been able to measure and systematically assess.

We hope these examples inspire you to downloadNodeXL, play around with it, and share your own stories and examples with us via Twitter (@NodeXL), Facebook, and in the comments of this blog. If you want to read up a bit more you may want to check out our book, Analyzing Social Media Networks With NodeXL: Insights From a Connected World, the ConnectedAction blog, the NodeXL graph gallery, or the other sources referenced in the blog.

Derek HansenDerek L. Hansen  is an assistant professor at Brigham Young University.




Marc SmithMarc Smith is chief social scientist at Connected Action Consulting Group.


This entry was posted in Innovations and tagged , , , , , , , , , , , , . Bookmark the permalink.

4 Responses to Exploring Online Community Social Relationships

  1. Azedine says:

    Wow!!! Solid Approach!!! My question is do you still that all this networks are “social” ?

    Thumb up 0 Thumb down 0

  2. Derek Hansen says:

    Good question. I would say these graphs are definitely social in the broad sense of the word “social” – i.e., they are based on relationships and interactions between people. These particular networks are based on following someone on Twitter or replying to them in a discussion forum, which doesn’t necessarily imply that people who are “connected” want to go to a movie together.

    Thumb up 0 Thumb down 0

  3. Charlotte Williams says:

    Thanks for explaining this great resource. Do you have any recommendations about how to balance immediate data results with the potential for long-term change? To take one of your examples, if you note that a community event failed to inspire more group cohesion, do you have any insights for determining whether you should abandon this tactic, or continue to host events since the trend might change over time? It would be a shame if immediate negative results deterred people from strategies that might take-off given enough time and effort, but then again no one wants to invest in a losing strategy (that the data may point out).

    Thumb up 0 Thumb down 0

  4. Derek Hansen says:

    Social network analysis, and really any analytics you run, really can’t answer those questions. However, having objective data that can inform decision-making puts you in a much better position as a community administrator to justify your choices, or perhaps question existing assumptions. Community interventions certainly take time. One strategy that I’ve seen work well is to consistently track various metrics (including SNA metrics) each month (or bi-weekly) and look for trends both in the short and long-term. Looking back at a 6-month period makes very clear whether existing approaches are leading to the desired results or if there were just bursts of activity that died down after some initial excitement.

    Thumb up 0 Thumb down 0