Inferring and Visualizing Social Networks on IRC

By using an IRC bot to monitor the activity in an IRC channel, it is possible to infer a social network that connects the users in the channel. Visualizing these social networks is not only interesting, but has a variety of potential applications.

IRC has millions of users from all over the planet. It is arguably the most open chat system, making it great for general discussion and collaborating with other developers.

While you are happily chatting away, do you ever think of the channel as being more analogous to a real room than it first appears? Using simple heuristics that make use of this analogy, it is possible to infer and create visualizations of the social networks.

Looking at these diagrams rapidly reveals "who knows who" and how strong those relationships are. Some possible applications of the inferred social network are:

• Integrating the system with an existing IRC client. This would be a pretty neat feature. As the system is implemented using the PircBot Java IRC Bot API, it could even be turned into its own IRC client rather easily.

• Spam filtering on IRC. Did you ever think this was possible? Maybe it is when you have an inferred social network of a channel. Users who are far from you in the social network, or are not connected at all, are more likely to send messages that you would deem to be irrelevant.

• The social network can also be analysed to obtain approximate levels of trust between pairs of users.

• Similar principles can also be used to generate drawings of channel similiarity networks. These are useful for discovering new channels which you are likely to be interested in, by virtue of the fact that they share a common set of users.