How does active participation effect consensus: adaptive network model of opinion dynamics and influence maximizing rewiring

In this paper we study the impact of active participation -- or deliberately seeking out other agents with an aim to convince them -- on the dynamics of consensus formation. For this purpose, we propose an adaptive network model in which two processes shape opinion dynamics at interwoven time-scales: (i) agents adapt their opinions subject to influence from social network neighbours who hold opinions within a tolerance interval $\delta$ and (ii) agents rewire network connections with an aim to maximize their own influence on overall system opinion. We study this system in both an endogenous setting in which all agents are subject to influence and also attempt to maximize influence, and in a setting of exogenous control, in which external agents not subject to influence adaptively attempt to maximize their influence. In both settings we find three regimes of stationary opinion configurations: (i) for low $\delta$ a regime of two evenly balanced radicalized opinion clusters at the extremes of the opinion space, (ii) for intermediate $\delta$ a 'winner-takes-most' regime of two unevenly sized radicalized opinion clusters, and (iii) for large $\delta$ a regime in which very low spread compromise consensus states can be reached. Comparing to adaptive processes of random and deliberately spread-reducing rewiring, we demonstrate that in regime (iii) competitive influence maximization can achieve near-minimal opinion spread within near-optimal times. Further, we also show that competitive influence maximizing rewiring can reduce the impact of small influential minorities on consensus states