A good social network analysis tool for advocacy & journalism

Some of the most interesting transparency and journalism-related datasets are those in which the relations of various actors and institutions are portrayed over time: corporate ownership, lobbying or organized crime networks all present similar challenges.

And yet there seems to be no good off-the-shelf package to display such data. Sure, there are things like UCINet for classic SNA and Gephi for graph data visualization. But I think they’re inadequate. The gold standard for graph presentation in 2011 is not graphviz, its Facebook. I don’t mean to imply that networks need to be stored as massively centralized, ad-powered dotcoms, but there are a couple of lessons to learn:

  • It needs to be fully web-based and not overtly about graph management. Rich attributes for both nodes and edges need to be stored and presented in a task-specific form.
  • Some level of access control enables more interesting use cases.
  • Network effects apply to networks. Ha.

The problem in building this is simple: its a trap. For any coder, our greatest dream is to transfer some new domain of knowledge into a perfect graph (I’m guilty myself, having worked on liquid democracy). In this case, the temptation is almost irresistible: the data is already a network.

Yet as we work to translate things into this graph, we tend to lose sight of the goal. A nice example of this is OWNI’s InfluenceNetworks: while it is supposed to be the the tool this whole post is about, they ended up just building a snazzy graph management app. Nice, but useless.

Even more dangerous: doing this internally. The linked data community seems to be crippled by graph fever. More than once have I sat through an hour-long presentation on the use of linked data in which no actual use for an end-user facing application was mentioned. While I’m not saying its impossible to build good applications on RDF, it seems like a huge distraction from actually tackling the domain.

But the problem remains: we need some easy-to-use solutions for advocates, journalists and other investigators to map out networks. The best effort towards this I’ve seen so far seems to be andoc, but even this does not feel like it is hitting the right trade-off between abstract graphness and concrete modelling yet. But: how do we find that trade-off? Seems some conrecte use cases and a nice agile process is needed…