Suppressing the Endogenous Negative Influence Through Node Intervention in Social Networks

Viral marketing, a marketing strategy that utilizes word-of-mouth (WOM), is effective in increasing brand awareness and acquiring new customers, as WOM allows information to reach a large audience in social networks. In the past two decades, for efficient viral marketing, many studies on maximizing advertising reach, known as influence maximization, have been conducted in the field of data mining. However, most of them ignore the possibility of the emergence of negative opinions in the information diffusion process. In general, negative opinions are more contagious than positive ones, and ignoring them may even lead undesirable outcomes, such as a decline in brand image and a decrease in purchases. To address this issue, we consider the problem of suppressing the negative influence that emerges endogenously on social networks through preemptive node interventions, such as persuasion, nudging, or warnings. Namely, given a limited budget for interventions, who should be targeted to efficiently suppress the spread of negative opinions in the social network? We formulate this problem as a combinatorial optimization problem on graphs. We prove that this problem is NP-hard and propose approximation algorithms to identify optimal intervention nodes that minimize the negative influence. Through numerical experiments, we demonstrate that our algorithms effectively suppress the negative influence regardless of the type of social network or experimental setting.