Is Social Media Driving Political Polarization?
Battles rage on Facebook and Twitter—but their influence on real-world politics is subtler than you might think.
By Lee de-Wit, Cameron Brick, Sander van der Linden
Americans are more divided along party lines than ever before.
In the past two decades, the percentage of Americans who consistently hold liberal or conservative beliefs—rather than a mix of the two, which is the case for most people—has jumped from 10 percent to over 20. At the same time, beliefs about the other side are becoming more negative. Since 1994, the number of Americans who see the opposing political party as a threat to “the nation’s well-being” has doubled. This deepening polarization has predictable results: government shutdowns, violent protests, and scathing attacks on elected officials.
Why are we becoming more polarized?
There are probably many reasons. Could social media be driving polarization? Many people think so—and, indeed, Facebook, Reddit, and Twitter have all become sites of ferocious political argument. While polarization definitely plays out on social media, the evidence to date suggests that its impact is subtler than you might think. Social media, it seems, amp up moral and emotional messages while organizing people into digital communities based on tribal conflicts.
This makes consensus building more difficult—but, as we’ll discuss, it could also pave a more cooperative path forward.
Do we live in “filter bubbles”?
Many people argue that we increasingly live in online filter bubbles that only expose us to the ideas we already agree with. This is consistent with a broader psychological literature on confirmation bias, showing that we are more likely to seek out and agree with views that align with our pre-existing beliefs. Selecting our preferred news sites and curating our social media accounts potentially makes it easier to listen to groups or individuals who validate our own worldviews.
The filter bubble idea has recently been elegantly demonstrated in the lab by Cass Sunstein and Tali Sharot and colleagues. The authors tested who participants would turn to for advice in categorizing geometric shapes—an obviously non-political task. In fact, this study found, participants preferred to seek advice from people who held similar political views, deciding that they must be more competent—despite evidence to the contrary!
If following people on social media who are more aligned with your worldview exacerbates polarization, then it follows that listening to “the other side” would reduce polarization. However, a recent experiment found essentially the opposite.
Christopher Bail and colleagues from Duke University recruited hundreds of Democrats and Republicans who were active on Twitter, and paid them to follow a Twitter bot that would retweet content from the opposing side. After a month of exposure, the Democrats retained about the same attitudes—but the Republicans ended up more conservative than when they started the study! This result suggests that polarization in the U.S. could be driven by exposure to views people disagree with, rather than being separated from them by filter bubbles.
There are several ways of interpreting this result. For example, it could be that participants were reacting directly to the content of the messages they were exposed to on Twitter, but it could also be the case that they were simply responding to the messengers, not the message. In other words, the issues are not as important as group affiliation. Whatever the interpretation, this study suggests that more work is required to understand to what extent filter bubbles might drive political polarization.
A study by Levi Boxell and colleagues provided a simpler test of the role of the Internet: Is more social media use associated with more polarization? Boxell and colleagues assessed polarization in the U.S. for different age ranges—and they surprisingly found that polarization was highest for the age groups that use the Internet and social media the least, such as older adults (75+).
This suggests that if the Internet is fueling polarization, its influence might be more indirect. This indirect influence is plausible, however, because in many traditional newsrooms, activity on social media has itself become news. Indeed, Trump has proved particularly successful in dominating the traditional news media (TV and print) with his activity on Twitter.
Thus, it is possible that the climate of debate on social media influences the tone of debate on other media platforms. Could social media influence polarization even in Americans who rely on the traditional media?
How social media shape debate
William Brady and colleagues tested what types of political messages on Twitter are more likely to be shared. When the researchers looked at tweets from the Presidential, Senate, and House of Representatives candidates in the 2016 U.S. election, they found that tweets with more emotive and moral words were more likely to be retweeted. All voters responded more to words showing moral outrage, but effects were somewhat stronger for tweets from Republican candidates, and Republicans were more likely to respond to emotional words about patriotism or religion.
This work suggests that if politicians want to maximize their impact on Twitter, they need to resort to more moral and emotive vocabulary. This in turn might help explain why encouraging people to follow politicians from the opposing side appears to worsen polarization: Politicians tweet the policy positions that their political base wants to hear, of course—but they do so in moral and emotive language that may create negative reactions from the opposing side.
This language itself becomes news, as reporters turn tweets into headlines that can generate fear on the “other side.” In this way, the whole news cycle shifts towards more polarizing and emotionally laden content.
According to Gordon Allport’s “contact hypothesis,” contact between groups lessens prejudice. However, decades of research testing this hypothesis has found some limitations. Although intergroup contact does tend to increase cooperation—and reduce prejudice—the positive effect may depend on important contextual factors, such as the nature of the conflict and whether the groups have equal status or a common goal. As Allport put it, sometimes more contact can lead to more trouble. That appears to be the case on Twitter.
Can we build social-media bridges?
So, perhaps we need to start thinking about how to structure interactions between groups on social media so that conflict becomes less likely and cooperation becomes more possible.
Social media companies need to do more themselves to counter online extremism and polarization by, for example, better regulating the political targeting of ads on their platforms. However, even in the absence of change in policy or on the sites themselves, there is a lot we can do as individuals to make social media less polarized.
A distinctive feature of social media is the importance of social consensus cues or online endorsement (e.g., likes, shares). Some research shows that the presence of these distinctive social cues can actually trigger decisions to select news in a way that reduces selective political exposure. In other words, if a story has been upvoted or shared a million times, it is likely to burst your bubble, even when the content is not ideologically congenial.
We can also try to cultivate a diverse network, extending beyond our immediate circles. It’s not necessarily a good thing that social media increase the volume of information that we receive from people whom we already know well. Rather, research suggests that even when we are not exposed to the “other side” directly, so-called “weak ties” (friends of friends, acquaintances) offer a degree of political diversity that might inspire more political moderation.
Unfortunately, as the research we’ve discussed suggests, this diversity won’t necessarily lead to harmony, since you’re more likely to encounter strong emotional and moral language that could trigger negative partisan reactions. You might find the antidote in a trait known as Actively Open-Minded Thinking (AOT). AOT is a cognitive style that allows people to be more thoughtful, flexible, and open-minded, even when information contradicts a strongly held prior view. For example, both conservatives and liberals who score higher in AOT are less likely to display strong polarization on hot-button issues, such as climate change. Similarly, a new study earlier this year found that high-AOT Twitter users were better at creating and responding to social media content in a thoughtful and reflective manner.
In short, perhaps the solution to our problem is not completely out of reach. When we treat online spaces like we would treat our own community, difficult conversations become more productive. No matter how confident in our arguments or aggrieved we feel, everyone benefits when we actively try to be more thoughtful and open-minded about what we say—and how we react to others with whom we disagree—both online and offline.
By Lee de-Wit, Cameron Brick, Sander van der Linden
Americans are more divided along party lines than ever before.
Why are we becoming more polarized?
There are probably many reasons. Could social media be driving polarization? Many people think so—and, indeed, Facebook, Reddit, and Twitter have all become sites of ferocious political argument. While polarization definitely plays out on social media, the evidence to date suggests that its impact is subtler than you might think. Social media, it seems, amp up moral and emotional messages while organizing people into digital communities based on tribal conflicts.
This makes consensus building more difficult—but, as we’ll discuss, it could also pave a more cooperative path forward.
Do we live in “filter bubbles”?
Many people argue that we increasingly live in online filter bubbles that only expose us to the ideas we already agree with. This is consistent with a broader psychological literature on confirmation bias, showing that we are more likely to seek out and agree with views that align with our pre-existing beliefs. Selecting our preferred news sites and curating our social media accounts potentially makes it easier to listen to groups or individuals who validate our own worldviews.
The filter bubble idea has recently been elegantly demonstrated in the lab by Cass Sunstein and Tali Sharot and colleagues. The authors tested who participants would turn to for advice in categorizing geometric shapes—an obviously non-political task. In fact, this study found, participants preferred to seek advice from people who held similar political views, deciding that they must be more competent—despite evidence to the contrary!
If following people on social media who are more aligned with your worldview exacerbates polarization, then it follows that listening to “the other side” would reduce polarization. However, a recent experiment found essentially the opposite.
Christopher Bail and colleagues from Duke University recruited hundreds of Democrats and Republicans who were active on Twitter, and paid them to follow a Twitter bot that would retweet content from the opposing side. After a month of exposure, the Democrats retained about the same attitudes—but the Republicans ended up more conservative than when they started the study! This result suggests that polarization in the U.S. could be driven by exposure to views people disagree with, rather than being separated from them by filter bubbles.
There are several ways of interpreting this result. For example, it could be that participants were reacting directly to the content of the messages they were exposed to on Twitter, but it could also be the case that they were simply responding to the messengers, not the message. In other words, the issues are not as important as group affiliation. Whatever the interpretation, this study suggests that more work is required to understand to what extent filter bubbles might drive political polarization.
A study by Levi Boxell and colleagues provided a simpler test of the role of the Internet: Is more social media use associated with more polarization? Boxell and colleagues assessed polarization in the U.S. for different age ranges—and they surprisingly found that polarization was highest for the age groups that use the Internet and social media the least, such as older adults (75+).
This suggests that if the Internet is fueling polarization, its influence might be more indirect. This indirect influence is plausible, however, because in many traditional newsrooms, activity on social media has itself become news. Indeed, Trump has proved particularly successful in dominating the traditional news media (TV and print) with his activity on Twitter.
Thus, it is possible that the climate of debate on social media influences the tone of debate on other media platforms. Could social media influence polarization even in Americans who rely on the traditional media?
How social media shape debate
William Brady and colleagues tested what types of political messages on Twitter are more likely to be shared. When the researchers looked at tweets from the Presidential, Senate, and House of Representatives candidates in the 2016 U.S. election, they found that tweets with more emotive and moral words were more likely to be retweeted. All voters responded more to words showing moral outrage, but effects were somewhat stronger for tweets from Republican candidates, and Republicans were more likely to respond to emotional words about patriotism or religion.
This work suggests that if politicians want to maximize their impact on Twitter, they need to resort to more moral and emotive vocabulary. This in turn might help explain why encouraging people to follow politicians from the opposing side appears to worsen polarization: Politicians tweet the policy positions that their political base wants to hear, of course—but they do so in moral and emotive language that may create negative reactions from the opposing side.
This language itself becomes news, as reporters turn tweets into headlines that can generate fear on the “other side.” In this way, the whole news cycle shifts towards more polarizing and emotionally laden content.
According to Gordon Allport’s “contact hypothesis,” contact between groups lessens prejudice. However, decades of research testing this hypothesis has found some limitations. Although intergroup contact does tend to increase cooperation—and reduce prejudice—the positive effect may depend on important contextual factors, such as the nature of the conflict and whether the groups have equal status or a common goal. As Allport put it, sometimes more contact can lead to more trouble. That appears to be the case on Twitter.
Can we build social-media bridges?
So, perhaps we need to start thinking about how to structure interactions between groups on social media so that conflict becomes less likely and cooperation becomes more possible.
Social media companies need to do more themselves to counter online extremism and polarization by, for example, better regulating the political targeting of ads on their platforms. However, even in the absence of change in policy or on the sites themselves, there is a lot we can do as individuals to make social media less polarized.
A distinctive feature of social media is the importance of social consensus cues or online endorsement (e.g., likes, shares). Some research shows that the presence of these distinctive social cues can actually trigger decisions to select news in a way that reduces selective political exposure. In other words, if a story has been upvoted or shared a million times, it is likely to burst your bubble, even when the content is not ideologically congenial.
We can also try to cultivate a diverse network, extending beyond our immediate circles. It’s not necessarily a good thing that social media increase the volume of information that we receive from people whom we already know well. Rather, research suggests that even when we are not exposed to the “other side” directly, so-called “weak ties” (friends of friends, acquaintances) offer a degree of political diversity that might inspire more political moderation.
Unfortunately, as the research we’ve discussed suggests, this diversity won’t necessarily lead to harmony, since you’re more likely to encounter strong emotional and moral language that could trigger negative partisan reactions. You might find the antidote in a trait known as Actively Open-Minded Thinking (AOT). AOT is a cognitive style that allows people to be more thoughtful, flexible, and open-minded, even when information contradicts a strongly held prior view. For example, both conservatives and liberals who score higher in AOT are less likely to display strong polarization on hot-button issues, such as climate change. Similarly, a new study earlier this year found that high-AOT Twitter users were better at creating and responding to social media content in a thoughtful and reflective manner.
In short, perhaps the solution to our problem is not completely out of reach. When we treat online spaces like we would treat our own community, difficult conversations become more productive. No matter how confident in our arguments or aggrieved we feel, everyone benefits when we actively try to be more thoughtful and open-minded about what we say—and how we react to others with whom we disagree—both online and offline.