Research On Information Interaction And Diffusion In Internet Communities

With the fast development of Internet technologies, Internet has become an important environment for people to share and disseminate information, to form and express personal opinions. The computer mediated interactions has become an important way for people to influence each other. However, basic principles of information diffusion and interaction among Internet users are still unkown. Some convincing models that would provide insights into user behaviors are needed. The dissertation is devoted to analyzing and modeling information interaction and diffusion in Internet communities. We analyze real statistics of user behaviors, propose agent based models that mimic real interactions, carry our computer simulations, study into how some complex phenomena come forth, and provide discussions on further regulation and promotion strateties. Results of the dissertation are useful to understand basic characteristics of information interaction and diffusion on the Internet, and to explain complex human behaviors in the real world. The work of the dissertation is supported by the National Natural Science Foundation of China (No.60972012) and Beijing Natural Science Foundation (No.4102047).Main contributions of the dissertation are as follows:1. After analyzing real distribution and time properties of users'reading and replying data in a BBS community, we propose a participating model, which would restore complex characteristics of real data. The model and its simulation result show that how users interact with the web server, how users organize their reading list, and how users'participating behavior distributes are important factors in formatting group behaviors of these Internet users.2. We propose an evolutionary game theory framework to model an opinion interaction, in which people are trying to persuade others to follow their own opinions, but unwilling to be persuaded. Considering the difficulity and loss in persuading others, people are allowed to quit some unpromising interactions. It is shown that strategy preference, noise in the strategy adoption, and ability in memorying historical interactions are influential for the opinion evolution. The model also proves that people might exhibit heterogeneous strategies after experiencing different interactions. We propose a cooperative game to model common people who tend to find an agreement in the opinion interaction, and a minority game to model contrarians who prefer to be different. The game theory model shows general evolving properties as most opinion models. Through game theory analysis and computer simulation, we study how contrarian behavior, network topology, updating rule in the strategy adoption, initial occupancy of different opinions, and uncertainy in the strategy adoption would influence the evolving process, and provide discussions about how to promote and hinder the evolution to the initially preferred opinion.3. We analyze some real properties of opinion interactions in the BBS community, and prove that agent based models could be used to model opinion interactions among Internet users. Considering real characteristics in the BBS opinion interaction, we propose a cooperative game theory model. Evolution results of the model show similar properties with real data, and offer explanations to some real phenomena.4. We propose to use agent based influence models to mimic the information diffusion process on the Internet. Through modeling and simulating mutual influence of two topics, we show how the acknowledgement way, the occurence time, the public participating interest, and the diffusion scale would influence the diffusion and competition of different topics. We propose an influence model to simulate some typical behaviors of Internet users in the information diffusion. The model and simulation show the influence of group size and distribution on the diffusion process.5. How to target promoting objects is an important and controversial problem. Considering the expecting proficts in diffusion and the promotion cost, we propose a decision theory model for the choice of promotion strategies. The model was applied into a basic influence diffusion process. Through simulation and analysis, we show that ignorance of heterogenous characteristics of users might be the best and economical strategy.