Our society is becoming increasingly dependent on various technological networks, such as transportation systems, electrical power?distribution grids, computer networks, and so on. As those networks evolve and grow in complexity, their dynamical behavior is becoming difficult to understand and predict. This project will develop a general framework for modeling growth and evolution of such complex adaptive networks, based on the notion of interacting stochastic processes. The intuition behind this approach is that the interactions that form the network are informed by the collective state of the stochastic processes, while those processes themselves are affected by the forming network structure. The presence of such a feedback mechanism is vital for capturing realistic behavior of many real-world networks. The research will develop rigorous mathematical methods for analyzing structural and dynamical properties of adaptive networks, and define novel information?theoretic measures for quantifying their complexity.
Broader Impact: Our nation?s technological infrastructure of future will depend on our ability to control large?scale, dynamic networks of interconnected heterogeneous entities. This work will help to better understand, characterize, and predict the collective behavior of such networks. The project will train new professionals and scientists in an important interdisciplinary area, and develop a graduate course material on complex adaptive networks.
Our society is becoming increasingly dependent on various technological networks, ranging from infrastructure components such as transportation systems and electrical power–distribution grids, to computer networks and internet, and with the recent surge of the social web, to online networking sites such as Facebook, Twitter and so on. As those networks evolve, grow in complexity, and sometimes intertwine with each other, their dynamical behavior is becoming much more difficult to understand and predict. The main objective of our project was to develop novel methods for modeling and predicting dynamics of such complex adaptive networks. There is growing evidence that social networks can influence individuals' behavior, both positively and negatively. During the course of the project, we have developed approaches for modeling co-evolution of individual and collective behavioral traits. The insights gained from our models can be used for devising effective and targeted interventions programs that could help containing propagation of negative outcomes and spreading positive ones. More generally, they will help to understand how exactly social networks affect individual and group behavior, how does a behavior or a trait spread in groups, and how to identify the causal flow in individual and collective dynamics. For instance, several high profile studies have suggested that certain counterintuitive traits such as obesity or happiness can be socially contagious. However, it has been argued that the observed correlations might be due to hidden confounding factors rather than social influence. One of the specific outcomes of our project is a powerful statistical test that can rule out confounding factors if there is sufficiently strong social influence. In particular, we have shown any model producing correlations between actors through static confounding factors alone will obey certain constraints, and we have developed computationally efficient technique to detect violation of these constraints. The project has had a strong educational component and provided an excellent opportunity for students and a post-doctoral researcher to work on an interdisciplinary project of great importance. Some of the research findings from the project were included in the syllabus of a new interdisciplinary advanced graduate course., Computation and Physics, that was offered at the University of Southern California during the Spring 2012 semester, and which generated interest among both engineering and physics students.
