Influence maximization considering fairness: A multi-objective optimization approach with prior knowledge

Abstract

The influence maximization problem (IMP) has been one of the most attractive topics in the field of social networks. However, sometimes fairness in IMP should be considered, especially when information access is critical to people’s lives, such as information about job opportunities and medical resources. In this paper, we study IMP considering fairness (FIMP). We introduce a new definition of fairness, which can correctly evaluate the fairness of a seed set, and then, the FIMP is modeled as a multi-objective optimization problem and addressed by the FIMMOGA that we proposed. The results of experiments on several empirical and artificial networks indicate that our method can provide better and more diverse solutions compared with previous methods. In addition, the introduction of prior knowledge improves the quality of solutions and gives our framework good compatibility with the existing methods and significant extendibility. Our research can easily be extended to other contexts, such as gender equality, religious equality, and racial equality.