I’ve been looking at the veloGraf twitter data the last couple of nights to speed up some of our recommendation algorithms and started thinking at what point does the “follows” relationship becomes a worthless measure of how valued the ideas or personal relationship is to the account holder.
For instance, a user who follows 20 or 30 people probably has a personal connection or interest in those they are following. An account like @BarackObama, however, is following about 715,000 people at this time. An army of social media PR people would have a hard time sifting any relevant information from 715,000 people. From a recommendation algorithm’s point of view I believe these relationships become mostly random noise. @BarackObama is an example of a broadcaster: someone who sends out information, but rarely engages in personal conversations.
In discussing this with Gary tonight he mentioned Dunbar’s Number which is a postulated limit of one’s social network cognitive limit, which is around 150. This number is derived from physical social networks of primates. What is buried in Twitter, however, are virtual relationships with some of whom we converse, some accounts we vicariously watch and sometimes engage, and the some who are broadcasters we look to for an information feed. Andreas Kluth explored this notion with the help of Facebook in his piece Primates on Facebook. So 150 seems low in the digital age, but multiple thousands of relationships seems qualitatively unmanageable.
In the future I think it would be beneficial for us to consider tuning algorithms to these concepts of cognitive social limitations. In the social media age there is little keeping us from “following” and “liking” our way to thousands of shallow relationships in the social graph but we are still cognitively constrained to a much smaller communities and a meaningful algorithm should understand our social behavior including our social limitations.