Communications and Signal Processing Seminar
Decision-making tools for influence propagation in social systems
Add to Google Calendar
Social influence can be a force multiplier in a variety of public health, marketing, and political campaigns. However, utilizing social influence to pro-actively influence decisions is difficult for a variety of reasons: absent a model of human decision-making, commonly-used models of social influence are inexact; the data required for common targeting algorithms is usually unavailable or too costly to procure; and, even if the latter two hurdles are overcome, deciding who to target and when to act is computationally difficult.
I will present a synthesis of our work on decision-making for social influence, including work looking at the effect of limited network visibility on strategies that seek to choose a set of individuals to target ("seeds" ). I will then discuss our work on dynamic influence, which finds optimal strategies for budget allocation to multiple influence channels across time for a political campaign seeking to win an election.
In the latter, we show that for a general set of objective functions, the optimal influence strategy is to exert maximum effort in waves for every channel, and then to cease effort and to let the effects propagate. We also show that early on, the total cost-adjusted reach of a channel determines its relative value, while targeting matters more closer to election time. Through our analyses, we identify a new and adaptable temporally varying centrality metric, and show how it can effectively be used in the computation of these optimal allocations.
Soheil Eshghi received the PhD degree in Electrical and Systems Engineering from the University of Pennsylvania in 2015. He is currently a postdoctoral associate at Yale Institute for Network Science (YINS) and the Department of Electrical Engineering at Yale University. His research interests include mathematical models of epidemic control and influence propagation in social systems, as well as applications of optimal control and game theory to social, biological, and electrical networks.