Incentive Compatible Influence Maximization in Social Networks

Information diffusion and influence maximization in social networks are well studied problems and various models and algorithms have been proposed. The main assumption in these studies is that the influence probabilities are known to the social planner. The influence probabilities, however can vary significantly with the type of the information and the time at which the information is propagating. The most accurate sources to obtain influence probabilities are the agents in the social network. In this work, we formulate game theoretic models of the information diffusion process so as to elicit influence probabilities truthfully from the agents. For these models we design several mechanisms to truthfully extract the influence probabilities from the users. We consider two different game theoretic models namely (1) influencer model and (2) influencer-influencee model. In the context of the influencer model, we design a Vickrey-Clarke-Groves based mechanism. In the influencerinfluencee model, we design a scoring rule based direct mechanism. We analyze the incentive compatibility of all these mechanisms.