Adapting the Social Network to Affect Elections

S. Sina, N. Hazon, A. Hassidim, S. Kraus
AAMAS 2015
Abstract
We investigate the effect a social network could have on voting outcomes. We consider a group of self-interested agents where each agent has a strict preference order over a set of outcomes. Each agent votes strategically, taking into consideration both her preferences, and her (limited) information about the preferences of other voters. We assume that the information the agent has comes from her friends in the social network and from a public opinion poll. If agents were not strategic at all, the social network (and the poll) would not matter, since they would just vote according to their preferences. However, if the agents deviate and vote strategically the network plays a great effect. To measure this effect, we focus on iterative voting with Plurality voting rule. We show, both in theory and in simulations, that for many networks, adding a linear number of edges can make any outcome the winner. We view our results as yet another indication to the effect that a central organizer, such as a company who controls social media, could have on our lives - by introducing us to certain people it can affect our information and our decisions.

Experiments:

Election type Culture Candidates Voters Instances Parameters
Ordinal Impartial Culture [3-10] {25000, 50000, 75000, 100000} 1000 None
Ordinal PrefLib [3-10] {25000, 50000, 75000, 100000} 1000 https://www.preflib.org/dataset/00014
Ordinal Single-Peaked (Walsh/Uniform) [3-10] {25000, 50000, 75000, 100000} 1000 None
Ordinal Urn Model [3-10] {25000, 50000, 75000, 100000} 1000 “Urn model (with 2-urns and with 3-urns)”. Unclear what that means.