I had a very interesting conversation with David Huffaker last week.  David just finished his PhD in Media, Technology and Society at Northwestern University, which makes him an expert in social media analytics, an exciting relatively new field. 

David has analyzed behavior on social networks such as Google groups to get a better understanding of how influencers behave.  He defines influencers based on 3 criteria :

1. Does the content they create trigger replies?

2. Do these replies generate conversations?

3. Do the followers adopt the influencer’s language in these conversations.

The insights were not surprising but interesting nevertheless :

- Leaders are active – they are more likely to post messages, reply to others, and have a longer tenure in the community.

- Leaders are social – rely more on replying to others than merely on broadcasting.

- Leaders are passionate – they will demonstrate higher frequencies of talkativeness, affective language (use words such as good, bad, love, hate, …), assertiveness (use words such as always and never) and linguistic diversity.

He also analyzed the characteristics of the group.  He analyzed groups based on

- The size of the group

- The density of the group – how well connected are individuals in the group?

- The stability of the group – what is the group turnover?

- The participation equality – do group members participate equally or are there only a handful of members who account for the bulk of the activity?

The picture above is one of his group visualizations which show how dense the group is, where the influencers can be found (size of the bubble indicates degree of influence).  On the outside of the graph you can see how the group reaches out to new members in the outside world. 

His analysis showed that more nodes and connectivity results in more individual influence.  At the same time equality in participation and turnover don’t matter.

I asked David whether influencers were really that important.  One of my previous posts showed an interesting analysis on Facebook data that seemed to support Duncan Watts theory on the importance of group dynamics vs influencers.  David said that the jury was still out for him.  He said that much more empirical evidence was required to really validate or refute any of the theories on how social networks work.

That’s exactly what he will be doing next.  He is working on a post doc where he will analyze Second Life to get a better understanding of how groups behave (part of his research is funded by the military and the government).  According to David, Second Life is an interesting lab for studying group interactions because of the similarities with real life.  Second Life unlike other social networks has the notion of geography and “physical” spaces which could make the findings more applicable to real life.  I am looking forward to his findings.