Control of Network Opinion Dynamics by a Selfish Agent with Limited Visibility

Manipulation of a linear network opinion dynamics by a single selfish agent is considered. The selfish agent is modeled as measuring a projection (statistic) of current network opinions, which may be either myopic or global. The agent then enacts a local feedback control using this measurement. The selfish agent's ability to manipulate network-wide opinions to a desired reference is assessed, both under nominal conditions and in the presence of a second, stubborn agent. Specifically, it is shown that low-gain proportional and proportional-integral controllers can be applied to manipulate network-wide opinions or opinion statistics. A nonlinear variant on the opinion-dynamics model, where the agents' opinions are capped, is also briefly studied. The control strategies are illustrated using a 62-node dolphin social network model. (C) 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.