Thursday, September 1, 2016

Within social networks, information spreads and ideas compete. People can be persuaded by strong arguments on social media to change their opinions on, say, who is the best presidential candidate in the 2016 race. However, the level of difficulty in changing an individual’s opinion through social media varies—while some people stand firm, others waver in their opinions on certain issues.

To better understand how competing opinions move through social networks, University of Maryland researchers created a new model based on those commonly used to study how contagious diseases spread, but with one major addition: a quality they call “stubbornness” and define as an increasing resistance to changing one’s opinion.

Similar models assume that opinions do not directly compete against each other. They also do not account for the idea that the longer people hold opinions, the less likely they are to alter their opinions. In the UMD model, if people hold on to their opinions for long enough, they lose their ability to change them.

“I was surprised that we could find a fit this good to our model,” said Michelle Girvan, associate professor in the UMD Department of Physics and the Institute for Physical Science and Technology. “Our model performs similarly to another model for opinion dynamics, but our model has the advantage of being more realistic than the other model, which fits the data but takes a bit of liberty about how opinions spread.”

The UMD model was published online March 3, 2016 in the journal Physical Review E. In addition to Girvan, the paper’s authors include UMD physics graduate student Keith Burghardt and William Rand, an assistant professor of business management at N.C. State University who conducted the research while he was an assistant professor of marketing at UMD.

In the team’s model, individuals are either “susceptible” or “infected.” Using the example of voter opinions, susceptible individuals include people who are eligible to vote but have not decided who they will vote for in a given election. Infected individuals are eligible to vote and know who they plan to vote for on Election Day.

Infected people try to persuade susceptible neighbors in their social networks to vote for a certain candidate. Infected people can also change the opinions of other infected people. For example, an individual who plans to vote for Hillary Clinton can persuade a Donald Trump supporter to instead vote for Clinton in the November election.

While other models assume that Trump and Clinton campaign for voters in social networks that are independent of one another, the UMD model assumes the opposite. The model accounts for the idea that people who are close, either geographically or through connectedness in an online social network, are more likely to share an opinion than those who are less connected to one another. The propensity for individuals to connect to others who are geographically close leads to spatial correlation. When votes are spatially correlated, it means that geographic closeness makes people more likely to vote for the same candidate in an election.

Recent studies have shown that for U.S. elections, as the distance between two voting districts increases, there is a decrease in the likelihood that people in the two districts will vote for the same candidate. In European elections, when two voting districts are geographically close, their turnout rates—fractions of eligible voters who actually vote in elections—are more similar than for two voting districts that are chosen at random. Research has also shown that the spatial correlation of vote shares in U.S. elections and turnout rates in European elections decrease as the log of the distance between two voting districts.

Using their model, UMD researchers reproduced the distributions of election results from several European countries when the results were rescaled by a factor of Q/N, where Q is the number of candidates and N is the number of voters in an election. The team’s model also showed that the closer people are in the space of a network, either in a physical sense or through connectedness on social media, the more likely they are to vote for the same candidate in an election. Recent U.S. presidential election data reflects the effect that physical proximity has on whether people will vote for the same candidate.

These results can be used to create more effective viral marketing campaigns and to improve the way a wide variety of information is spread, according to the researchers, who are now studying how stubbornness affects how jurors vote during trials. 


The research paper, “Competing opinions and stubborness: Connecting models to data,” Keith Burghardt, William Rand and Michelle Girvan, was published March 3, 2016 in Physical Review E. 

This work was supported by the Defense Advanced Research Projects Agency (Award Nos. N66001-12-1-4245 and D13PC00064) and the U.S. Army Research Office (Award No. W911NF-12-1-0101). The content of this article does not necessarily reflect the views of these organizations.

Media Relations Contact: Abby Robinson, 301-405-5845,

Writer: Rachel Crowell

University of Maryland
College of Computer, Mathematical, and Natural Sciences
2300 Symons Hall,
College Park, MD 20742

About the College of Computer, Mathematical, and Natural Sciences
The College of Computer, Mathematical, and Natural Sciences at the University of Maryland educates more than 7,000 future scientific leaders in its undergraduate and graduate programs each year. The college's 10 departments and more than a dozen interdisciplinary research centers foster scientific discovery with annual sponsored research funding exceeding $150 million.