Journal on Policy and Complex Systems
Figure 6 . Initial configuration of opinion network .
and the messy complexities of real societies , democratic , or otherwise . Any two nodes in the modeled networks have either a fully open communication link or no communication link at all , for example , with no attempt to model the topic-specific censorship that often characterizes authoritarian regimes . Within the networks considered , moreover , distinctly layered hierarchical structures do not have the clear prominence that they often do in social organization . Here , as elsewhere in modeling , the goal is not a detailed representation of reality but an intentionally schematic diagram that helps us understand it . Abstract as this model is , the hope is that attention to the dynamics of abstractly modeled opinion on this selective landscape of artificial networks can start to give us a grasp on the far messier volatility of opinions within far more complex social structures .
3 There is a range of related work on voter model variations on complex networks , though without the focus on political implications explored here . See , for example , Castellano , Villone , and Vespignani ( 2003 ); Suchecki , Equíluz , and San Miguel ( 2005 ); Sood and Redner ( 2005 ); Schneider-Mizell and Sander ( 2009 ).
94
Opinion Volatility
Consider the 50-node network
shown in Figure 6 . We begin with a percentage of agents in the network holding opinion p , coded in green , with another percentage holding opinion not-p , coded in red . At each step of the simulation , we have a certain percentage of our agents update their opinions using a simple majority version of the threshold model ( Andjel , Liggett , & Mountford , 1992 ; Durrett & Steiff , 1993 ; Liggett , 1999 ). Agents adopt the opinion held by the majority of those with whom they are in contact in the network , with a random choice in the case of a tie . 3 In what follows we have a random 10 % of our agents update on the majority view of their network contacts .
In order to measure volatility within different networks we introduce