Casade of Information in Networks
WG 6: Internet Emergent Properties
Here is a short summary of the session that took place during “The Wealth of Networks”
Our session was entitled “Cascade of Information in Networks” and the panelists were:
Paul Bourgine (CREA, France), Jeffrey Johnson (Open University, UK), Renaud Lambiotte (Imperial College, UK) and Jure Leskovec (Carnegie Mellon, USA)
The session has been organised as follows:
1) R. Lambiotte has first introduced the panelists and has briefly explained the notion of emergence in relation with the Internet: the emergence of collective decentralized phenomena. Decentralized in the sense that there is not one leading entity which determines the behaviour of the whole system, but that a collective behaviour emerge from the interaction between many alike individuals. The case of blogs vs. Mass media is given as an example. The concept of “cascade of information” is also introduced.
2) J. Leskovec has given a talk of about 20 minutes entitled: “How to detect disease outbreaks and influential bloggers at the same time?”. This talk focused on the empirical description of cascades of information among blogs, but also in a viral marketing setting of 16 million product recommendations among 4 million people. It is shown how one can not only detect cascade outbreaks at their early stage but also find influential bloggers whose posts lead to large cascades. These bloggers are not especially those with the largest number of readers or the largest degree, but those giving rise to the largest avalanches. A new blog ranking is proposed incorporating this idea
3) The discussions have then covered very different subjects:
– such large scale “online experiments” open exciting perspectives in the social science as they allow to validate models for social dynamics.
Such problems require a true multi-disciplinary approach.
– from a practical point of view, how can one find influential blogs which may contain interesting news before they arrive in the mass media (New York Times). There may be potential problems as these influential blogs may be very specific and written for specialized readers: one therefore needs intermediate blogs which select the information and rewrite it for a broader public. – applications to viral marketing and to the detection the birth of hype stories and rumours.
– what about the missing hyper-links? Is it possible to find them?
– this concept of cascades could also have applications to biological systems (e.g. neurons).
– problem of the adaptability of people: given some way to rank blogs, people will find ways to improve artificially their own ranking.