superk

Personal

Full Name

Guy Kelman

Homepage

http://www.cs.huji.ac.il/~superk

Position

PhD student in Physics and Computer Science

Affiliation

Racah Institute for Physics in the Hebrew University of Jerusalem

Research topic

Ordering the disordered: Revealing the hidden structure of uncoordinated contributions to knowledge systems

Short CV

Hebrew University, Jerusalem B.Sc 2001 Mathematics and Life Sciences
Hebrew University, Jerusalem M.Sc 2006 Atmospheric sciences

Ph.D. project

Thesis title

Ordering the disordered: Revealing the hidden structure of uncoordinated contributions to knowledge systems

Defense year

2010

Faculty

Exact sciences, Givat Ram, HUJI

Ph.D. supervisors

Sorin Solomon Hebrew University of Jerusalem, Israel. ISI Foundation, Turin, Italy

Professor of Theoretical Physics
Complexity

Ph.D. Supervisors

Scott Kirkpatrick

General project description

We propose a way to extract the social networks implicit in the Web, and characterize them. Lately there is great effort to initiate and grow social networks and to use them for economic, marketing, social, citizenship ends. We claim that the social networks are already represented in the Web and one can map, predict and influence their structure and evolution using agent based and dynamical networks techniques. We propose to construct iteratively the geometry of those networks and to identify by periodic sampling the "hot spots" and "emerging phenomena" within them. Moreover we propose to identify the leading and most influential nodes as well as the ones crucial for the information flow between its sub-communities. By doing so we will be able to identify and characterize the various communities, their overlap and interactions, their social dynamics, their future potential and the ways to influence them

Motivation

The study of the Web became in the last years an established field.
The static structure stylized facts from power laws to clustering in cliques and from k-cores to logarithmically small diameter are now established.[Car07], [Sha05], [Sol04], [Alb02].
By contrast, the dynamics of the Web are still a matter of conjecture and the social forces and the social identity of its components are still largely a mystery. In recent times, more than ever, the Web has developed in such a way that, the action we fondly term “surfing the web” has reincarnated in the form of active “surfing”. That is, interacting with the matter that appears on the computer's screen by adding comments and links to elsewhere, or by revising the text, the images and other graphic material. Added the discussion about subjects with others, all recorded and appear together with the original material. In that respect, the social forces and identity of the web pages are no longer governed by their originator's policies, but by wide gamut of contributions [Muc07].

We claim that the 2 unknowns, social id and social forces, are related. Thus social dynamics underlying the Web can be understood only if a more precise knowledge of the rules by which new nodes and links appear is acquired.
Reciprocally, once the social structure is known, the future evolution of the Web can be inferred with some confidence.
Thus the study of the social structure of the Web participants can steer the Web towards better goals, more efficiently, more reliably and more safely.
In turn the knowledge of the Web network structure and dynamics may inform us of the real life social interactions and social dynamics. These in turn would allow the full understanding and use of the Web in an optimal yet free and safe way and allow us to influence and steer it towards a peaceful, non-confrontational, constructive and cooperative and co-evolving change. This ability can pave the way to forming an efficient distributed management scheme which may cater for a better distributed hierarchy governing the internet, Web and the real life [Sol04] [Gol05] [Mar03] [Muc06].

Approach and methods

Currently, most of the studies of the Web focus on a single static snapshot of sparsely sampled portions of the network. Our preliminary findings show that incomplete unguided sampling does not allow reliable retrieval of the properties of the actual Web. In fact, unless the sampling is quite complete, the network measures are strongly influenced by the sampling method. We first propose to use a more intensive way to sample the relevant Web regions in a way that uncovers more directly its social structure.
More precisely, the sampling proceeds by generating and monitoring a network consisting on nodes (representing people, institutions, concepts) and links (representing the co-appearance of these people, entities, concepts on the Web pages). Instead of a blind search, one would structure the search as a systematic walk through a network that follows the natural links between the relevant players.

Once the primary network is constructed, its monitoring along time periods will reveal the places where the action is at each particular time slice, the people and institutions leading and participating in it and the eventual interactions / interference / positive feedback between initially disjoint processes. After a few time slice analyses one can identify the various (overlapping) communities, their leaders, as well as the people and organizations that are irrelevant and should be discarded.

Contributions

The Time-Slice technique: to partition the network into time-components that have stronger internal connections and constitute identifiable sub-communities. This in turn defines the profile of the various Web "realms" and the expected content of the communication within those realms as well as the way it propagates

Curriculum Vitae

Scientific events organization

Complexity in Jerusalem 2008 - ECCS08 cluster of events http://jeruccs2008.org

Publications

Publications

Rosenfeld, D., W. L. Woodley, A. Lerner, G. Kelman, and D. T. Lindsey (2008), Satellite detection of severe convective storms by their retrieved vertical profiles of cloud particle effective radius and thermodynamic phase, J. Geophys. Res., 113, D04208