On cost-effective strategies for opinion diffusion in complex networks

Abstract

Better understanding of diffusion dynamics in complex networks is of notable scientific and social interest, since it allows predicting, controlling, and delaying information, innovation, or even epidemics. The most effective strategy for fast and impactful opinion dissemination is to use social agents to diffuse and indoctrinate neighboring peers, be it online or offline. Although these agents imply a real-world cost of operation, there is limited literature on the trade-offs between diffusion performance and incurred costs. Our study incorporates a cost awareness model into the classic linear threshold opinion diffusion model and studies the effect of variable network topology, number of spreaders, and spreader active time (i.e., injection strategies). By performing detailed discrete event simulations on several types of network topologies, we uncover a set of general rules for a cost-effective approach in targeting real-world scenarios of influence maximization. Specifically, we determine that a constant opinion injection incurs costs of up to 80% higher than intermittent injection, that increasing the number of spreaders results in a polynomial increase in the cost of operation, and that irregular (random, preferential attachment) topologies imply diffusion costs of up to 3 times higher than regular topologies.