The Limits of Influence Maximisation in Online Social Networks

Influence maximisation over online social networks is a challenging ethical and technical problem. Among many applications, it is becoming relevant in public health and prevention: by choosing the right spokesperson for an advertising campaign, in fact, we can encourage people to adopt healthier lifestyles and increase organ donation rates. Traditional Influence Maximisation models assume that all members of a social network have the same propensity to spread a message, regardless of the topic, and that each user will forward a message to all of his or her contacts. These assumptions are clearly unrealistic: a user could be very sensitive to some messages (thus contributing to the information diffusion process) and completely insensitive to others (thus stopping the spread of information in a community). A user could also selectively decide to whom a message should be sent. To overcome the above limitations, we introduce a novel information diffusion model called IMBC (Influence Maximization with Budget Constraints), where the budget is an upper bound on the number of contacts to whom a message can be spread. We have experimentally verified our algorithm on six large real-world networks containing millions of nodes.