Do You Really Know How Social Networks Operate?

Back in 2007, the Wall Street Journal did an analysis of more than 25,000 user postings from several of the largest sharing and collaboration Web sites of the time.  On AOL's Netscape site, which then had roughly a million users, the paper’s reporters found that 13% of the postings rated “most popular” came from a single user—a 27-year-old Dayton, Ohio, computer programmer with the screen name STONERS.  And at Digg, a full third of the postings popular enough to make it to the home page came from just 30 of that site’s 900,000 members.

But perhaps most interesting of all was what the reporters found at Reddit.  In 2007 Reddit had 400,000 members, but one of the most influential and widely read was a man named Adam Fuhrer, who specialized in newsworthy items about criminal justice and software releases.  Fuhrer had garnered particularly favorable ratings from other Reddit members for his appraisals of the security flaws and price tag of Microsoft’s new Vista operating system when it was released, for instance. So when the Wall Street Journal’s reporters looked him up to better understand how he had come to have such influence, imagine their surprise at learning that Adam Fuhrer was 12 years old.  He lived with his parents in Toronto, where he attended elementary school.

We are all affected more and more by online social networks – from the social sentiments and trending topics that sweep like weather patterns across Facebook or Twitter, to the connections we make or the news and information we get right here on LinkedIn.  Unfortunately, however, most of us don’t fully realize how social networks actually operate, and the important implications this has for how we use (or abuse) them.    

Start with the fact that networks like Facebook, Twitter and LinkedIn don’t just pop into existence, fully formed. An online social network grows from day to day, as new members join and current ones build new connections.  Academic studies indicate a network evolves over time by “preferential attachment” – meaning that whenever a new member joins or a new connection is made, it is highly likely to involve those current members who already have the most connections.   

Preferential attachment provides a good explanation for many phenomena involving networks and connections, including Adam Fuhrer's rise to prominence on Reddit.  Because of preferential attachment:

1. The structure of an online social network is path-dependent.  That is, how a network looks today is based on how it looked yesterday, and what new connections were built since yesterday.  Tomorrow will look just like today, along with any changes that happen today.  But new connections and new members result from individual decisions by independent people so, like the weather, a social network’s structure is impossible to predict more than a few days in advance.
2. Influence in a social network tends to have a power-law distribution, not dissimilar to the 80-20 rule sometimes used as a short-cut for understanding customer values.  This means that a few key influencers will have a great deal of sway, and their links to other networks may impact thousands, or even tens of millions, of others.  But again, the ultimate influence of any single member is impossible to predict very far in advance, because it will have as much to do with the randomness of how the network evolves from day to day as it does with the expertise or knowledge of the member.

So how did Adam Fuhrer acquire such remarkable influence on Reddit?  Most likely because he authored something very smart that someone else who already had a great deal of influence voted up (giving Adam "karma" points, in Reddit lingo).  And because this influential member shared his own favorable opinion, other members began to pay attention, and more members tracked the next item that Adam posted – which, if it was also reasonably smart, would attract even more attention, and more karma points for Adam.  This is a positive feedback loop -- a self-reinforcing mechamism that is common in many network phenomena.

But here’s the thing: Suppose you were Microsoft and you wanted to predict who the next Adam Fuhrer-type influencer was going to be, either on Reddit or on any other online network.  This is an impossible task. You won’t be able to predict the next Adam Fuhrer any more than you could predict whether it will rain in Dubuque, Iowa on May 12, 2023. 

On the other hand, you can definitely say that there will be another Adam Fuhrer, just like there was that one Netscape user with 13% of the home-page postings there, or those 30 Digg members who generated a third of the most popular postings there. 

You could say that the structure and flow of online social influence is locally random, but collectively predictable.  The local, individual outcomes (i.e., the identities of key influencers or the particular links among members) are inherently random, because they are based on path-dependent preferential attachment.  But the power-law distribution of influence is an emergent characteristic of every online social network, so it is quite predictable, even inevitable. 

In a future post I'll take up the issue of how a business can deal with the inherent unpredictability of social sentiment and key influencers.

(Illustration credited to Bruce Kasanoff)