How to map your social network
Friday 2 November 2012, 12:05
One of the rules of thumb taught in many communications courses is ‘know your audience’. It can also be useful to know what your audience thinks of you.
For example, what interests a random sample of the followers of the @BBCCollege Twitter account?
Over the past couple of years I have been experimenting with the idea of 'social interest positioning' - a technique for mapping the social interests of audiences around a particular person, tag, search term or shared link.
First, we need to sample an appropriate audience. For a person this might be some or all of their followers; for a hashtag, search term or shared link, it may be people recently using that tag, or term, or sharing, liking or bookmarking that link.
Having sampled our audience, we can then start to map out their interests by seeing who they follow. So, for each member of the audience sample, we grab the list of people they follow and construct a network diagram:
We can then see who is followed by a significant proportion of the audience sample, removing those who aren't commonly followed by the audience:
This leaves us with a much smaller, and more easily managed, network. We now need to associate these commonly followed accounts with particular interests in order to build a picture of what the audience in general is particularly interested in.
The map is constructed as a network visualisation, with lines going from audience members to the people they follow.
One tool I often use is Gephi, a cross-platform desktop application. For Excel users on Windows,NodeXL does a similar (those less beautiful!) job.
The visualisation is laid out as a map using a technique called a force-directed layout. This imagines links between people exerting an attractive force between them, with the result that it tries to position people who are followed by the same audience members close to each other. To the extent that by following a person they reflect your interests, people who are positioned close to each other can be seen as reflecting the common interests of their followers.
Sometimes there may be tensions - for example, when a person is notable or of interest for two or more reasons. So a ministerial MP may be followed both by constituents who follow local businesses and people interested in the ministerial matters that the MP is concerned with. In Communities and Connections: Social Interest Mapping, I describe how the followers of a Milton Keynes community action group appear to split into interest groups relating to charitable concerns on the one hand and social enterprise in the Milton Keynes area on the other.
We can use the resulting social interest maps to both identify the perceived concerns of a particular user or hashtag based on the common interests of their audience and to segment that audience into groups with slightly different interests.
We can also look at the interests of a particular user account by grabbing the set of people followed by that user and seeing how they connect. If we try to position these friends so that they are close to each other if they follow each other, we can generate a map of the shared interests of those friends. It’s a bit like a cocktail party where birds of a feather flock together even if they are all known by the host or hostess.
An example of this is seen in Visualising How @skynews' Twitter Friends Connect. It shows how people followed by @skynews tend to follow each other, revealing a certain amount of structure in that network, including MPs, political journalists and Sky journalists:
Although Twitter is currently one of the easiest social networks to get access to social data from, it is also possible to extract this kind of friendship connections data from other social networks such as Google+. The image below shows a fragment of the Google+ network around Red Bull Racing:
If you would like to experiment with visualising some social network data, please visit Visualising Twitter Friend Connections Using Gephi: An Example Using the @WiredUK Friends Network which provides a linked-to data set and mini-tutorial on how to start visually analysing using Gephi.
