Skip to main content

Information Networks in Voting and Public Opinion

This article and paper discuss an effect known as “Information Gerrymandering” in the context of elections. This term refers to the phenomenon in which voters with variable access to information can be induced to change their vote if they perceive most others to be voting differently from them. Researchers studied this effect in the context of an online game they created; players are split into two equal teams (i.e. “parties”) and are able to vote one of two ways. There are three possible outcomes of the game: party A wins, party B wins, or a deadlock in which no party receives a high enough majority. Players receive a high reward if their party wins, a low reward if the other party wins, and no reward in the case of a deadlock.

Players are incented to vote for their own party (to receive the high reward), but also to vote in accordance with the opinions of others in order to help minimize the chances of a deadlock. This is where the paper made its interesting discovery. By varying the information a given player saw (i.e. only showing a player from party A the opinions of players from party B), it’s possible to sway a player’s vote. At the macro scale, this effect swung a number of elections in the favor of a given party, even when both parties had equal participants with equal influence (same number of nodes viewing each node’s opinion).

This research has serious implications for the modern world in which the flow of information (along “information networks”) can be easily controlled by various media entities and social media networks. From my perspective, problems can arise in two primary forms.

Firstly, so-called “echo chambers” can form on social media sites which are designed to show users content they’ll like and content from their friends. Both types of content are likely to contain information the user will agree with (people often think similarly to their friends, and content people like often contains opinions/information they agree with), which in turn reinforces their opinion. (In the research experiment, this is the equivalent of showing a player from party A only the opinions of other players from party A.) Predictably, this leads people to cement their opinions and become more polarized.

The second issue is that elements such as propaganda, fake-news, mass-advertising, and even small numbers of vocal supporters (given a loud voice by the internet) can sway large numbers of voters if deployed intelligently. Strategic dissemination of information to the correct individuals can swing people to a different opinion and create ripple effects throughout the information network which swing even more people to that opinion. Such information has nothing to do with fact or which result would be better for a given person, but can influence their decisions nonetheless.

This problem is naturally framed through the tools we’ve developed thus far in class. We can view the information network as a graph in which each person/entity is a node and a directed edge between nodes means that the in-node can view the out-node’s information/opinion. This is a powerful framework with which to analyze and consider the problem. Viewing a real-world network’s structure could provide insights into which voting blocks are cemented in their opinions (i.e. high local clustering, few connections to other clusters of another party) and which are at risk of being swayed (i.e. lower local clustering or clustering of nodes from disparate parties, many connections to nodes/clusters from different parties). Additionally, as I alluded to earlier, interesting ripple effects of changes in opinion could be explained by network structure (i.e. if a person changes their opinion, others who see their information may change their opinions and so on). All this being said, I find it alarming that the information network and its connections can make such an impact on people’s opinions. We live in a time when this network can be easily studied (due to the internet) and exploited via propaganda, etc. At best, such exploitation leads to deadlock (no party is able to make meaningful progress) and at worst candidates who will poorly serve large segments of the population are elected. Even a handful of hostile actors can mobilize a huge voice through social media channels and bring about the advent of these issues, to the detriment of society. Intelligent solutions will have to be devised to counteract such hostile actors effectively. Ensuring that nodes remain connected to a good balance of other nodes will help to counteract “Information Gerrymandering,” for example.

In closing, I’d like to add that I think it could be interesting to further consider this situation through the lense of game theory. Depending on how we think about the motivations of individuals (better for the other party to win or to have a deadlock?), the parties’ choices to gerrymander information or not could be viewed as a variation of either the Prisoner’s Dilemma or the Hawk-Dove Game mentioned in the textbook. Both considerations have interesting implications for the real-world, although I agree with the article below that the situation is more of a Prisoner’s Dilemma as it stands currently.


MIT Article:

Nature Publication:



Leave a Reply

Blogging Calendar

September 2019