Restricting the spread of Firestorms in Social Networks

We investigate the social-media phenomenon defined as “online firestorms”: sudden discharges of large quantities of negative word-of-mouth that spreads rapidly through online social networks. Firestorms can start due to various reasons, such as online marketing campaigns that backfired or dissatisfaction of customers, and are a consequence of opening social media channels to the crowds. Firestorms have affective and viral nature and therefore posing severe threats, such as harming brand reputation and causing customer losses. Our motivation in this paper is the development of optimized forms of counteraction to firestorms, which engage individuals to act as supporters and initiate the spread of positive word-of-mouth, helping to constrain the firestorm as much as possible. We describe the required optimization as a seed-selection problem in the context of firestorms by explaining the differences it has compared to its examination in other existing contexts. We propose a new seed-selection method that is based on the concept of local centrality, which unlike existing social network analytics, selects supporters based not on the global structure of the social network, but locally, i.e., by taking into account the areas of a social network that have been affected by the negative word-of-mouth. Experimental evaluation with data from a real social network demonstrates that the proposed method presents several advantages compared to seed-selection based on commonly used global centrality scores.