Facebook study proves catching

An article published in PNAS on 11 October 2010 by CABDyN's Jukka-Pekka Onnela and Felix Reed-Tsochas examines the spread of Facebook applications among approximately 50 million users between June 25th and August 14th 2007, shortly after Facebook introduced applications.  By looking at the number of users for each of 2,705 applications on an hourly basis, it is possible to infer whether the installation decisions of different users appear to be temporally correlated, or not.  Surprisingly, the method of fluctuation scaling reveals that there are two quite distinct regimes of behaviour for different applications.  Different users either decide to install applications without showing any sign of being influenced by the actions of other users, or above a certain threshold for the application, are strongly influenced by what other users are doing.  Strangely, there is no gradual change in user behaviour with increasing popularity, but rather a fairly sharp off-on transition.  Additional simulation results with synthetic time series show that this transition is not equivalent to a standard epidemic threshold.  At this point it is unclear whether the observed system properties reveal something special about online environments and social networks, or whether equivalent behaviour could in principle also occur in the offline world.  Certainly, media coverage of this work provides yet another interesting example of how ideas spread in networks.  A preliminary version of the findings published in PNAS was covered by The New Scientist, and following a news story released by Oxford University, recent media coverage includes Nature, the BBC, The Economist, and the ORF.

19 October 2010