In an earlier post, we discussed a live Twitter chat led by UNDP Director Helen Clark that was sponsored by Guardian Development. Based on a global view of the use of the #askUNDP hashtag, we found that the conversation was a dialogue (as almost 2/3 of the posts were replies or retweets) and that the conversation was global (as twitter users in Africa and Asia outnumbered those in North America).
While these findings are interesting in their own right, a map of the network of tweets and retweets surrounding the event can tell us more. Extracting tweets using the #AskUNDP hashtag and taking note of their interconnectedness produces a network map that looks like this.
While this is visually pretty cool, it tells us little. Network graphs use two mathematical concepts to describe the nature of relationships in a network. The notion of betweenness centrality refers to the distance between two people. Put simply, it’s a number that tells us how often in a network a point (here a twitter account) connects two other people. High numbers indicate high betweenness centrality. The notion of eigenvector centrality refers to the connectedness of a person’s connections. If one’s connections are themselves highly connected, a person has a high eigenvector centrality. In a sense, then, nodes with a high betweenness centrality are transmitters of information, while nodes with a high eigenvector centrality are centers of attention.
We can use these two concepts to better understand the twitter network using the #askUNDP hashtag. In the figure above, those nodes that have betweenness centrality scores of greater than 1 are represented as squares, and those with scores less than one are represented by filled circles. Larger squares, then, represent high betweenness centrality.
Three of the top four nodes with the highest betweenness centrality are UNDP twitter accounts: Clark’s (with the highest score) UNDP’s Algeria Office, and the main UNDP twitter account. The node with the second highest betweenness was the Guardian Development twitter account. Many of the remaining nodes with high betweenness were private individuals. As their tweets were RTd (or replied to by Clark) they were central to this network.
Turning to the eigenvector centrality, it’s not accidental that the centers of attention were the four accounts mentioned above (in order, @HelenClarkUNDP, @GdnDevelopment, @UNDP, @PNUDAlgerie). These are denoted as the four large blue and purple boxes at the center of the above network chart.
It is worth noting that while these four accounts were at the core of the #askUNDP project, the actual roles that they played were distinct. The @HelenClarkUNDP account initiated discussions by making replies to questions that came in. In the sample of tweets that we have, the @UNDP and @GdnDevelopment accounts did not initiate tweets, so their centrality to the discussion came passively in the way of mentions. Finally, the @PNUDAlgerie account played the role of a transmitter, retweeting much of the earlier discussion.
So what can we learn from this? First, for IOs looking to reach the public, multiple accounts make a difference. Each of the three UNDP accounts reached different audiences. It’s not too difficult from the above to see that removing any one of the UNDP accounts would have shrunk the size of the audience. Importantly, this is not just about the issue of differences in language. Of the 40 tweets from @PNUDAlgerie, only 9 of them were in French. This then ties into a second point. One of the things that made this event go was the collaboration with the Guardian. The Guardian worked for days preceding the event to build an audience, relying in turn on its own network in order to do so. In an important sense, collaboration with the the Guardian served as an important multiplier for the main message. While we cannot estimate the hypothetical reach of an #askUNDP event that was solely conducted from Clark’s office, the network analysis tells us that @GdnDevelopment was mentioned so frequently that it was a central part of the discussion. For other IO heads looking to get the message out about their operations, then, this collaborative twiplomacy strategy has much to commend it.