IEEE WCNC (April 18, 2019)

Marrakech, Morocco

Diffusion of Information and Inference over Graphs

Information flow over graphs is a topic of significant relevance, especially in light of the proliferation of online platforms that facilitate communication and the exchange of opinions among members. In order to promote the diffusion of reliable information, it is important to understand which aspects of the graph topology favor the spread of misinformation, and which strategies can be used to enable belief control or sow confusion. This presentation provides an overview of research results on distributed information flow over weak graphs where the flow of information is asymmetric. This scenario is common over social networks. For example, it is not unusual for some influential agents (such as celebrities) to have a large number of followers, while the influential agent may not be following most of them. A similar effect arises when social networks operate in the presence of stubborn agents, which insist on their opinion regardless of the evidence provided by observations or by neighboring agents. It turns out that weak graphs influence the evolution of the agents’ beliefs in an interesting manner and facilitate the spread of false information over networks. While agents are able to learn the global truth from interactions over strong graphs, where there is a path between any two agents, anomalies arise over weak graphs where certain agents can control the statistical beliefs of other agents. This phenomenon permits the flow of misinformation and can be used to generate confusion. In particular, (a) agents in a graph can be made to arrive at incorrect inference decisions (a form of belief control); (b) they can be made to disagree among themselves (a form of social discord); and (c) and they can be made to continually change their beliefs about the truth (a form of confused learning). For example, some agents or sensors may be driven to believe erroneously that “it is raining” even though they may be observing “sunny conditions.”  This presentation examines these patterns of behavior over multi-agent networks and illustrates the results with examples and simulations.