Sunday, November 23, 2014

Facebook Friend Clustering

My wife is currently reading Dataclysm: Who We Are (When We Think No One's Looking) and pointed me to this interesting Relationship Test tool she learned about in the book. It analyzes your Facebook friends and then shows both their connections to you and to each other. The graph assigns weight to relationships, so "cliques" of friends will cluster together.  Here is the result I get:

Ken Weiner Friend Graph
Ken Weiner's Friend Graph

The white circle in the center is myself.  The cluster in the upper left identifies my friends from my
high school, Agoura High School.  The cluster in the lower right is mostly coworkers from my previous job at The 2 less-defined clusters to the right of my high school cluster is a mix of family and college friends.

In addition to visualizing friend clusters, this tool also tries to identify your spouse or romantic partner using an algorithm from a paper called Romantic Partnerships and the Dispersion of Social Ties:
A Network Analysis of Relationship Status on Facebook. It tries to answer this question:
Given all the connections among a person's friends, can you recognize his or her romantic partner from the network structure alone?
It measures something called dispersion - the extent to which two people's mutual friends are not themselves well-connected.

On my network, it worked extremely well, identifying my wife as the person with whom I have the highest dispersion score (called assimilation score in this tool).

Portion of my Facebook Connections Sorted by Assimilation Score
Portion of my Facebook Connections Sorted by Assimilation Score

This means that it is through my wife that I am connected to the most people that aren't themselves connected.  This is natural, of course, because I have met my wife's high school friends, college friends, and coworkers, most of which don't know each other.

There are a few people that rank high on the list that aren't actually "central" to my life.  If I ignore all the people with less that 3 mutual friends, the list becomes much more meaningful and predictive of who really is central to my life.

The ability to analyze my social network like this motivated me to visit Facebook and Unfriend acquaintances that I really never developed much of a relationship with.  Now, both my Facebook news feed and Dispersion scores will be that much more meaningful!

No comments: