Modeling Information Networks

The remarkable development of electronic communication networks in the past two decades is also the sign of a deeper socio-economic evolution. The explosive growth of cell phone usage, Internet-based communication methods, the World Wide Web, are technical advances which have far-reaching social consequences. Such techniques correspond closely to the rise of a "network society" where social and economic relations are being constantly renegotiated and restructured. These new technologies have promoted the emergence of virtual communities through e-mail communication, chat, forums, wikis, blogs and virtual collaboration spaces. Human beings are now more than ever social beings and we can notice that knowledge and information exchanges mostly take place within social networks. We believe that it is very important to study the structure of the technical and social networks determining the information flow in our society. These networks may be seen as dynamic, self-organizing structures that both facilitate and constrain the information flow between their nodes. There are fortunately models already available in several disciplines. The branch of sociology known as "network analysis" has elaborated sophisticated formal models of relational structures among social actors. Mathematicians have explored random networks, and lately "small-world" structures. More recently, researchers at the interface between mathematics and computer science have developed models of various network types, biological, social or technical. Precise relational data is often hard to obtain, however, and mathematical models are still under-determined. Also most structural models are static, and there has not been enough work on the evolution of networks through time. So for the study of dynamic phenomena, computer modeling and simulation seem to offer a promising avenue of research. Starting from different dynamic network models to be found in the literature, we first designed a modeling language that could be used to define different kinds of models. We wanted this language to be as flexible as possible while remaining easy to use. We then built a generic software tool with graphical display capabilities in order to generate various types of network given basic generative rules written in our modeling language. For instance, transitive closure (in social networks), Hebbian reinforcement (as in neural networks) or preferential attachment (typical of Web sites) give rise to very different structures, associated with different patterns of behavior or use. This generic tool can be used to study the structures most likely to arise in specific situations: e-mail communication, social collaboration networks, growth of Web links… Simulation results should of course be compared with real data whenever possible. We intend such modeling to help obtain useful insights about how to design better communication software, information search systems, Web browsers and so on. For example Google’s famous search algorithm was designed after careful study of the Web’s structure. We have thus designed and built a software that can simulate models of relational structures and information flow between social actors. This type of research lies at the interface between technological development and social usage, because the very nature of communication networks requires such a global approach for effective treatment. We will first present the syntax of the modeling language that can be interpreted by the system, we will then briefly explain the general software architecture and finally, we will comment a detailed example of a typical social network generated by simulation with this tool.