Social Network Visualisation versus Social Network Analysis

[Graham Durant-Law]

I've spent the last few days as an invited speaker at the Australian Institute of Professional Intelligence Officers annual conference. My topic was "Using Social Network Analysis as an Intelligence Technique ", with a sub-title of "Sometimes a Picture is Only Worth a Few Words!" One of my concerns is that people have become enamoured with visualisation at the expense of analysis - they are doing social network visualisation, not social network analysis. Worse they are applying algorithms, such as closeness centrality without actually understanding what the algorithms produce, and what their limitations are. So building on these concerns I borrowed a concept from Dr Marc Smith, which he calls Network Nirvana, as the means to illustrate my concerns. Network Nirvana is achieved when every node is visible, every link is visible and direction can be discerned, degrees can be counted, and clusters are clearly visible.

The presentation began with a quick network lesson, where I discussed bounded and unbounded networks, and directed and undirected networks. Understanding what type of network you are working with is important because it changes the equations used to calculate various measures, and indeed whether certain measures can be used with confidence. For example, closeness centrality can only be used with confidence in a bounded network. (Some authors say it should only be used in a bounded network.) I then used an case study on Iranian Nuclear Physicists to illustrate the dangers of blindly doing social network visualisation and applying common algorithms that are readily available in the software tools, instead of doing a structured social network analysis. In particular I showed how social network visualisation that does not achieve Network Nirvana can lead to false conclusions.

read more



Get More from the original blog...