We have completed maintenance on DiscoverMagazine.com and action may be required on your account. Learn More

Researchers Use Math Model to Map Echo Chambers on Twitter

A mathematical model that visualizes echo chambers on Twitter shows how they coevolve with polarization on controversial topics.

By Yuen Yiu
Mar 23, 2020 6:00 PMMar 23, 2020 8:34 PM
(Credit: Baumann et al., Physical Review Letters)


Sign up for our email newsletter for the latest science news

(Inside Science) — The picture at the top of this story is not a blurry image of two colliding galaxies or a splitting cell viewed through a microscope. It is a snapshot of today’s divided America.

On the left is a blob representing Twitter’s liberal echo chamber, and on the right, a blob for the conservative one. The brightness of the blob represents the concentration of tweets that echo other tweets with similar opinions. The researchers who generated the plot described their method in a paper published in Physical Review Letters in January.

Echoes and Filters

You may have heard of the terms “echo chamber” and “filter bubble.” While they are sometimes used interchangeably, they mainly refer to two different features in online discussions. The term echo chamber describes a phenomenon where people tend to interact with those whose opinions are similar to their own, and filter bubble describes the phenomenon of people not interacting with opinions different from their own.

“In our model, we looked at homophily, which refers to how people with the same opinion or similar opinions are more likely to interact with each other,” said Fabian Baumann, a physicist from Humboldt University of Berlin in Germany and an author of the paper.

Using the visualization protocol they developed, Baumann and his colleagues mapped thousands of tweets over several days around topics such as Obamacare, gun control and abortion. After weeding out the suspected bot accounts, the tweets were placed according to their political lean along the horizontal axis. Next, tweets from accounts that the tweeters frequently interacted with were placed according to their political lean along the vertical axis.

“If people interact constructively, I would give you my opinion, and you would give me yours, then we would meet in the middle,” said Baumann.

Instead, their data shows the extent to which people choose to echo only the tweets of those who share similar beliefs with them. In other words, liberals (toward the left side of the graph) tend to interact mostly with liberal neighbors (toward the lower end of the graph), and conservatives with conservatives.

The researchers found that while an overall consensus is possible, echo chambers could cause the discussion to destabilize and become polarized if the topic meets a certain level of controversy. Once a discussion is polarized, subsequent exchange of information would strengthen the echo chambers and drive the two groups further apart.

The visualization of echo chambers in the liberal and conservative Twitter spheres surrounding Obamacare, abortion and gun control. Readers can refer to Baumann et al. for more detailed definitions of the axes and parameters. (Credit: Baumann et al., Physical Review Letters)

Their data also showed that because more active Twitter users tend to have more extreme opinions, they play a significant role in the polarization of online discussions, especially for politically controversial issues. This may help amplify extremist ideas otherwise limited to a smaller population.

In addition to analyzing real-life data, Baumann’s paper also provides a mathematical model for predicting if a discussion online would become polarized based on the strength of the echo and what they dubbed as the “controversialness” of a topic, and proposes ways for quantifying these parameters. It may provide a starting point for studying the polarizing effect of echo chambers for different social and political issues.

Baumann acknowledges the need to also consider the filter bubble side of the equation explicitly, which they did not include in their model, and how online users may be repulsed from opinions different from their own due to personal preferences or algorithmic filters -- or a coevolution of the two.

“There have been studies that tried modeling constructive interaction as well as repulsive interaction, but most of the time it has been difficult to compare them to real data,” said Baumann. “Ultimately, we should combine all these effects, but right now, to quantify which one is more dominant, it's really hard to say.”

Limited by Platform

Among all the popular social media platforms, Twitter is perhaps the favorite choice for academic researchers to study because it is easy to access and because their users interact with each other in ways that are easy to quantify in mathematical models.

“How Twitter works naturally makes it really easy to be studied. There are basically just tweets and retweets, and these interactions are very easy to quantify,” said Baumann. “As soon as you have an account, you can pretty much read everything on Twitter. For Facebook, it’s a lot more difficult.”

Facebook, by far the most popular social media platform in terms of number of active users, has privacy options that make the platform more difficult to study compared to Twitter, which in essence is an open forum where every user can interact with each other in view of everyone else.

A series of scandals, most notably the Facebook-Cambridge Analytica data scandal, where Cambridge Analytica used Facebook data to influence the Brexit campaign and the 2016 U.S. presidential election, has prompted Facebook to restrict access to its data.

“Academics have, in a sense, been our own worst enemy as far as access to Facebook’s data goes,” said Joe Walther, a social scientist from University of California, Santa Barbara, referring to the Cambridge Analytica scandal that originated as a research project by a data scientist at Cambridge University in the U.K. “It's hard to blame Facebook for shutting off outside access to their data.”

The tendency of researchers to focus on Twitter may lead to results with underlying biases and blind spots. Still, the results provide at least one lens on the polarization that many perceive as widespread in American society today.

Good for Business. Bad for Society?

“Echo chambers may be harmful from a societal perspective, because they drive polarization,” said Walther. “But as long as they also drive user engagement, they allow tech companies to sell more advertising or charge more for their advertising, which is attractive from a corporate perspective.”

Since user engagement is a crucial part of the business model of companies like Twitter or Facebook, there is little financial incentive for them to rein in echo chambers as long as they help generate revenue.

But before we reach for our virtual pitchforks and demand action from Twitter and Facebook that may limit online speech, Walther urged that we need to better understand echo chambers first. While the internet has provided a platform for niche groups with insular perspectives, such as people opposed to vaccination or people who think the Earth is flat, Walther said the accumulated research is unclear how much of an effect echo chambers have on broader social and political issues.

“I don't want to advocate censorship, but sometimes having absolutely no limits can lead to some not-so-good things,” said Walther. “And in order to think about policies and how we should adjust these platforms, we need an intellectual foundation before we can take that next step.”

[This story originally appeared on Inside Science.]

1 free article left
Want More? Get unlimited access for as low as $1.99/month

Already a subscriber?

Register or Log In

1 free articleSubscribe
Discover Magazine Logo
Want more?

Keep reading for as low as $1.99!


Already a subscriber?

Register or Log In

More From Discover
Recommendations From Our Store
Shop Now
Stay Curious
Our List

Sign up for our weekly science updates.

To The Magazine

Save up to 40% off the cover price when you subscribe to Discover magazine.

Copyright © 2024 Kalmbach Media Co.