Influence Maximization in the Voter Model

Influence maximization --the study of strategically influencing agents on social networks with an aim to align their opinions or choices with certain targets-- has major applications ranging from advertising and marketing to the political campaign problem, preventing the spread of extreme opinions and radicalization, or limiting the spread of fake news. In this talk I will summarize some of my recent results on influence maximization, or optimal opinion control, for the two state voter model. In the first part of the talk I will explore the effects of noise in the voting dynamics on optimal control, showing that generally two well-separated regimes of low-noise and high-noise control exist. In second part of the talk I will focus on aspects of non-stationary control, exploring the role of time horizons of strategic influencers on optimal influence allocation strategies. Results in both parts of the talk will explain that hub control is not always optimal and demonstrate that it might be best to control low- or intermediate-degree nodes under some conditions.

[1] D. Kempe et al., "Maximizing the spread of influence through a social network.'', Proceedings of the Nineth International Conference on Knowledge discovery and Data Mining (KDD), Washington, DC, USA (2003).

[2] N. Masuda, "Opinion control in complex networks'', New Journal of Physics 17, 033031 (2015).