New maps of influence – 10 visualisations of the social graph

Driven by waves of change in society, technology and business, new publics are emerging and forming new spheres of influence in the world of online relationships. Corporate communicators have to map this space out, so as to identify potential issues and influencers. Since the traditional means of media resonance analysis doesn’t reach into this world of direct communications, new maps of influence have to be created.

Social cartographers are now trying to visualise the relationships between people as they manifest themselves online. They are writing a new chapter in the analysis of social networks. While this discipline has been around since the 1930ies, I think there are three aspects that make today’s social network analysis unique:

  1. Wealth of data: There is more data available than ever, since hundreds of millions of people are organising their social networks online.
  2. Networks analysing themselves: Social networks are evolving as self-referential systems with participants indicating what they are interested in, so that peers can connect with them. A growing number of easy-to-use technologies also makes it easier for participants to become their own network analysts. The Rolodex is evolving into a cockpit of network intelligence.
  3. Mashups: People and content can be mashed up in myriads of ways offering a multitude of insights into people’s social life.

Let’s take a look at a couple of examples of how relationships in social networks are being analysed and visualised today. I believe that these social graphs ultimately will change the way we work, organise our knowledge and measure influence, simply because they are an effective way to navigate through the exploding amounts of content we produce and because they help us to connect with the people we share an interest with.

Flickr tags - June 2008 An easy way to visualise social data that has become pretty popular among social media users is the tag cloud. Tags are keywords users assign to content on web sites. Usually they are listed alphabetically with the font size representing the frequency of their use. E.g. on the left you can see the “tag cloud” of Flickr, the photo sharing and community site. It gives an idea of the most popular targets millions of people had when shooting photos, but it also connects to the people who took them. Each tag is a hyperlink that leads to the content tagged and the people who used the tags and created the content.

Gustavo G, a power user of Flickr, created the “Flickrverse”, a depiction of the relationships between the photostreams and the photographers on Flickr. It illustrates that there are clusters of influence in this community. Some users not only create large amounts of content, they also influence others by stimulating discussion groups, blogging, linking and commenting. Imagine you had this kind of atlas for your global communications team! With this knowledge on content people produce and the network they have, it should be much easier for them to to collaborate. That’s what IBM had in mind when they created their Atlas application. It enables employees to map out their professionel network within their organisation (story via Technology Review). E.g. there is a component shown in the screenshot on the left that visualises how closely someone works with others, both in terms of content and geography. The closer you are to the person at the centre of the circle, the more you communicate with her. Atlas will help to find colleagues and content relevant to your work. It seems like a great tool to collaborate within a business, but it does have the limitation that it only works with data from IBM’s own social software platform Lotus Connections. While social networks within a business will always need some degree of exclusivity, most public networks thrive on openness between separate layers of data that users can combine freely to create something new. Google Maps is one of the most prolific applications users mash up with other data sources in creative ways. For instance, I played with a tool called Wohnungskarte when moving to Düsseldorf, one of the many applications you can find combining geography and housing information. One of the most intriguing Google mashups I’m aware of is outside.in, a site tracking news, views and conversations in your neighbourhood. It allows you to zoom into the social life around you. Local news and interactions on social media are being bundled into one stream of social information that users can filter from a city level downwards. E.g. you can look at the river of news for New York City in total and then break it down by the five Burroughs Manhattan, Bronx, Brooklyn, Staten Island and Queens. Finally, you can look into specific neighbourhoods within the Burroughs, say the Upper West Side of Manhattan. Of course, the information will only be as good as the participation in the particular neighbourhood, but in many cases it provides an interesting micro-map of the social life in a particular area. The information is cut in different ways, so that you can follow your interest. You can check on categories such as arts and culture, education, real estate, public spaces or shopping, complete with links to pictures on Flickr. Or you can look up specific places and see how often they have been mentioned in recent stories or conversations. Consider the value of this information for new residents, local politicians or businesses trying to track the interests of residents.

Another example for the power of mashups is Facebook. Since this social network opened its Application Programming Interface (API) to external software developers, the number of free applications for Facebook literally exploded, some of them empowering its users to do their own network analysis. For instance, the image on the left shows my social graph on Facebook. I created it with an application called Nexus. It not only illustrates the connections between the friends I have on this network, it also provides information on the interests they share. When clicking on one of the nodes of my network, I can see the name of the respective person, her photo, the names of her friends in my network and commonalities she has with these people, e.g. membership in groups or common interests as stated in their personal profiles.

While it is manageable to analyse a small network like mine on Facebook, stronger tools are needed to crunch a huge network like the blogosphere in its entirety which also doesn’t live on only one platform like Facebook. Matthew Hurst, a scientist at Microsoft’s Life Labs and co-creator of the blog search engine Blogpulse, used data from Blogpulse to visualise hyperlinks between blogs (story via Technology Review). His images show that there are a few thousand blogs clustered at the centre of the blogosphere linking to each other and to many other sites at the edge. Further analysing this central cluster he found two sub-clusters, one focused on politics, the other focused on technology. They can be seen on the image here, with political blogs sitting on its left half, technology blogs on its right. The colour pink around influential political blogs indicates that they are connected to their surroundings by links going both ways, a habit that is obviously not as common in the sphere of technology blogs. Counting links as Matthew Hurst did for blogs is certainly a good way to create a helicopter view of influence clusters in the blogosphere. Linking to someone else’s blog is an act of interest saying “I want to be in the loop whenever this person has new content”. However, that doesn’t say much about the quality of the relationship when the number of links is very high. Some blogs have thousands of inbound links, some people have even hundreds of thousands of “friends” on social networks. To further qualify the relationships in a social network, MIT Media Lab researchers Dietmar Offenhuber and Judith Donath started to monitor the flow of comments between members of a social network (story via Technology Review). Rather than simply depicting links between sites, they visualise where people leave comments and how often, i.e. activity in the network that goes beyond reading updates. I believe this kind of analysis is very valuable. It reveals another layer of influence in social media: While social media is more democratic than traditional media, since potentially everyone with Internet access can raise her voice, the share of those who not only watch but actually create content is lower than you might expect. E.g. only about 1 % of Wikipedia users are writing articles. Given the huge popularity of Wikipedia this number is still in the tens of thousands, many more contributors than any traditional encyclopaedia ever had, but many less than the millions of Wikipedia readers. Similarly, stories are making careers on rating portals like Digg. Web developer Brian Shaler has created a graph that sheds some light on this. There are people who are much more active on Digg than others. The bright orange spots on Brian’s “heat map” represent those Diggers who have a lot of fans or friends. So, whenever they are rating a story it’s likely many other will, too. The oldest Digg accounts are located in the centre of the map, the youngest at the edge. Interestingly, many of the hottest spots are sitting in the centre, too, so it took some time for them to build their community. The interactive map also allows to search for specific Digg users and displays their spot on the map.

Visualising the social graph is based on data mining, i.e. filtering out relevant information by putting data points into a context that creates relevance. In the examples above we have seen how relevance is created by clustering and matching content and people. However, all of these approaches rely on pretty simple and isolated data points such as keywords and links. As we have seen, these data can create a lot of insight, but what you don’t get is the opinion of participants. You might know that they are interested in a particular topic, but not how they think about it. In other words, statements like “I love company A” and “I hate company A” were being picked up the same way. If you wanted to find out about participants’ opinion you needed to read their content. As a result, the next frontier of mapping influence within the social graph is opinion mining rather than data mining. Interone has made a big leap in this direction (full disclosure: Interone is a BBDO company, as is my employer Pleon). While their opinion mining tool still involves a degree of human analytics, their software can crunch through vast amounts of data, mining opinions in a far more sophisticated way than the “positive – neutral – negative” ratings we know from media monitoring. They can map people’s opinions against dozens of attributes and compare brands within that context. They even can do this in different languages! It’s a big tool that is made for big tasks, so you might not use it to analyse smaller networks, but it is powerful stuff for sure.

Finally, a software framework called Commetrix is also worth mentioning. It is being developed by a group of German researchers lead by Matthias Trier in Berlin. Other than all the other network analysis tools I know Commetrix allows for “dynamic network mapping” which means that you can observe network changes over time. Commetrix visualises the career of issues and influencers within a network by animated social graphs which is pretty impressive to watch. I believe the potential applications for a tool like this are mind-blowing. It could become a new way to write history. In fact, they kind of do that already. As you will see in the video below, they analysed e-mails of Enron, so that you can see how people within this organisation picked up specific topics and spread them around. It’s like an archaeology of influence in the Internet age! Imagine you could do something like this with the social networks relevant to your organisation. You could visualise and measure influence in real time. It certainly all depends on the accessibility and quality of data, though.

Get the Flash Player to see this content.

I’m well aware that this post is neither scientific nor exhaustive, but I hope it has at least provided an educated anecdotal view on the new maps of influence we need to be aware of as communicators. While all of this is pretty exciting, some of the pragmatists amongst you will rightfully ask: what can we do with it now? Unfortunately, there is no one-size-fits-all approach. In some cases free tools will be sufficient to shed some light on an area that is of particular interest to your business. In other cases you might have to use commercial software, or the ideal tool for your challenge simply doesn’t exist. In any case strong analytics and judgment will be needed to create an approach that makes business sense. But I do believe that we have to watch this space and experiment in it, so as to allow for informed decisions on a rapidly growing new sphere of influence for corporate communications.

Since I am still a German, I will end on a note of caution 😉 . Social network analysis is based on tracking human behaviour. As a result, it often includes privacy issues. While people are exposing themselves publicly like never before, they certainly still want their personal information only being used in a manner they agree with. Some power users have already crafted a bill of rights for users of the social web. It’s an early sign for big discussions lying ahead of us. Just think of the controversy on Facebook’s Beacon. But that’s an issue for another post. This one has been long enough.

Some useful links:

Georg Kolb