Structural hole-based approach to control public opinion in a social network

Abstract

Structural hole spanners play an important role in information diffusion. Compared with opinion leaders, structural hole spanners have better locations in social networks to expand the scope of information diffusion. In the past, researchers focused on evolution rules and opinion dynamics environments to monitor and even manage public opinion. In this study, we propose a novel structural-hole-based approach to control public opinion in social networks, hereinafter referred to as the SHCPO approach. We discuss the influence of both ordinary agents and structural hole spanners on opinion evolution using our improved Friedkin–Johnsen (FJ) model. Further, we analyze the evolution tendency of public opinion, which leads to the final consensus of public opinion, via the FJ model with ordinary agents in a community and structural hole spanners in joint communities. We reveal three kinds of connections between structural hole spanners and ordinary agents in joint communities. These comprise structural hole spanners connecting (1) two opinion leaders; (2) two ordinary agents; (3) one opinion leader and one ordinary agent. The three connections will lead to different opinion evolution conditions. According to the structural balance theory, we reconstruct the social network by changing the connections between structural hole spanners and agents in different communities. This guides the public opinion tendencies of joint communities towards the positive. Experimental results demonstrate beneficial effects of the SHCPO approach. We use three evaluation indicators to compare the SHCPO approach to five alternative methods. The percentage of positive opinions is used as an evaluation indicator. The SHCPO approach, compared with adding informed agents, add edges, the method from WWW and varying susceptibility to persuasion method, which guide the agent with a negative opinion towards positive opinion, has improved about 17%, 10%, 9%, 1%, respectively.