Crying 'Wolf' in a Network Structure: The Influence of Node-Generated Signals

摘要

Research into rumor spreading in a social network has largely assumed that information may only originate from an external content provider. In today's age individual nodes may also be content providers. The propagation of a signal generated by a node in the network may contribute or diminish the efforts of information diffusion, as signals become an imprecise indication of a node's knowledge. We present a model that allows for incorporating node-generated information into the well-studied area of modeling rumor spread in a network. We capture this by a stochastic information transmission mechanism at each node, with a positive probability to spread the rumor without holding its value. Simulations are performed using synthetic Watts-Strogatz networks, along with a real-world Facebook sample graph. Using decision trees as a descriptive tool, we examine the effects of the rate in which internal non-informed nodes generate information on the properties of the rumor spread process. As our main results we show that: increasing the rate of information generated by non-informed nodes may have monotonous or non-monotonous influence on the rumor spread time, in dependency with whether the network is sparse on not. We also identify that a strategy of increasing external communication in order to gain higher pureness level tends to be effective only for a medium level range of this generation rate and only in sparse networks.