Cha, M., A. Mislove, and K. P Gummadi. 2009. A measurement-driven analysis of information propagation in the Flickr social network
Contents
Citation
Cha, M., A. Mislove, and K. P Gummadi. 2009. A measurement-driven analysis of information propagation in the Flickr social network. In Proceedings of the 18th international conference on World wide web, 721–730.
Online version
Summary
This paper addresses the problem of measuring how information actually propagates in real world social network, in particular Flickr social network.
Information in this paper refers to the favorite markings of pictures; and the spread of information is therefore defined as how pictures are being favorite marked throughout the network.
Given the structure of the Flickr social network that obeys power law distribution and small world network structure, the findings in this paper were surprising. Although social links appear to be the dominant force in propagating information in this network (accounts for about 50% of information propagation), the information do not travel widely nor quickly in the network.
Description of the method
To measure how widely information travels in the network, the paper measures:
- whether popularity of pictures is a local or a global phenomenon in the network
- how far the distance between a fan and the uploader of the picture he likes in the network
To measure how quickly information travels in the network, the paper measures:
- the pattern of growth of pictures' popularities
- the dominant pattern of growth in the long run
Datasets used
subset of Flickr network crawled over a period of 104 consecutive days in 2006-2007.
Experimental Results
This paper aims to empirically analyze Information Cascade in real-world social network.
The findings of the paper are surprising, they find that despite the structure of the network (which indicates that information should be able to travel widely and quickly in the network), information does not travel widely or quickly in the network.
They find that popularity or the spread of favorite markings of pictures are limited to 1-2 hops neighborhood of the picture.
They find that the pattern of growth of popularity does not follow the theoretical pattern of "exponential growth followed by saturation/maturity" assumed in small world network. Instead they follow a period of steady linear growth sustained over an extended period of time.
They find that social links contributed about 50% of information propagation in the network, but find that the speed of propagation is slow (in average it takes 140 days for a picture's favorite marking to propagate from one user to another).
Discussion
In the discussion section of the paper, the authors try to make sense of the surprising findings by hypothesizing that it is the burnout aspect of information diffusion that limits the propagation of information in the network.
Another factor that could be at play is the information overload that overwhelms users in social networks. The inability to keep track of all of user's friends lowers user's exposure to picture updates from friends, causing favorite markings to propagate slowly through the social links.
During the in-class discussion, several interesting insights surface:
- The use of pictures' attributes could explain the different rates of growth of pictures' popularity. For example, funny pictures may grow more widely and quickly in the network than depressing pictures.
- The paper does not take into account the features that are unique to Flickr network, such as the frequency of user's login, their patterns of usage and profiles which may contribute to the surprising findings. Users may not login frequently, hence their exposure to other user's pictures or favorite markings may be limited, causing the slow propagation of information in the network. Users may also use Flickr only for storing personal pictures which are unlikely to gather large following, causing the "narrowness" of information diffusion in the network.
Related Papers
R. M. May and A. L. Lloyd. Infection Dynamics on Scale-Free Networks. Physical Review E, 2001
D. Liben-Nowell and J. Kleinberg. Tracing Information Flow on a Global Scale using Internet Chain-Letter Data. Proc. Natl. Acad. Sci. USA, 2008 Discussed on the slides during the Class Meeting for 10-802 03/24/2011
