The Polarizing Impact
Social Media Posts
College of William and Mary
AMERICANS HAVE GOOD REASON TO BE ALARMED at the extent to which social media sites have been harnessed for pernicious purposes. Media coverage has focused largely on the ways foreign countries, bots, and candidates have deployed social media to their advantage, fanning the flames of partisan polarization for their own benefit. Conversely, much of the recent scholarship on homophily, fake news, and social media has found that individuals frequently spread mis- and disinformation independent of foreign actors, that social media actually increases exposure to outside viewpoints, and that the majority of people report not participating in political social media exchanges. However, both of these views have provided insufficient attention to the intersection between social media and human psychology, particularly to the nonpolitical ways individuals signal their partisan affiliation, the complicated consequences of online exposure to alternative viewpoints, and the increasingly negative and personal associations social media users pin to those with opposing viewpoints.
In this regard, a shift in our conceptual framework—starting with the way people actually use social media instead of imposing our offline theories onto novel, online behavior—is important no matter what facet of social media political behavior we seek to explain.
In my book, Frenemies: How Social Media Polarizes America, I show that in the context of increasing partisan polarization among American political elites, the radical change brought by social media to the way people express their political identities, access information, and communicate with each other about politics has fostered people’s increasingly negative feelings toward those who identify with the opposing party.
How can it be the case that Americans exacerbate polarization online when most of them say they don’t post political content on social media? While a large proportion of people use Facebook for news, only 22 percent of users report that a lot or some of the content they post or share is about politics. Yet, huge majorities of Facebook users report they have learned others’ political views. The small proportion of users who report generating political content seems to have a disproportionate influence on everyone else. For example, Pew found that 37 percent of social media users report they are “worn out by how many political posts and discussions they see,” and 59 percent of social media users find disagreeable interactions on the site to be “stressful and frustrating.”
In short, only a small subset of Americans share political content, but almost everyone reports learning other people’s views, becoming fed up with political interaction in the process. The key to understanding this seeming contradiction is a deeper examination of the ways in which Americans use social media that are informative about their political identities and viewpoints.
To fully understand how social media contributes to the problem of psychological polarization, we must reorient ourselves to think about how Americans use Facebook more broadly, even when they think they are not being political, and what this can reveal about their political leanings.
Social Media Has Social Influences
Americans use social media primarily for social reasons, not political ones, and have integrated Facebook into their daily routines. ComScore reports that Americans spend one out of every five minutes online on a social networking site, with Facebook the most popular by far. The average American user spends 35 minutes a day on the site and, because most of the site’s users engage every day, the average user spends over 900 minutes a month on Facebook.
What does this imply about the effect of using social media on political behavior? The implication extends beyond the oft-noted fact that most Americans encounter political information inadvertently. Byproduct exposure has precedence: we have long known that Americans encounter politics when they aren’t looking for it. But before the popularity of social media, people typically encountered political news when they were searching for news of some kind. Now, they encounter political content when they are looking for information about people they personally know.
As a consequence, exposure to political information is filed away as part of the social information we seek about our family and friends. This exposure influences not just our attitudes about issues but our attitudes about each other, feeding into the processes of affective polarization. In other words, social media facilitates us linking our negative attitudes about the parties, politicians, and policies we disagree with to our attitudes about the people who support them.
Content Is Politically Informative, Even When It Is Not About Politics
The process of social inference described above is only impactful if there is a large enough quantity of information circulating on social media that communicates people’s political identities. And there is.
How do you recognize content about politics when you are scrolling through a social media feed? Have you ever learned someone’s political views based on their social media posts? Think a bit harder. Have you ever learned someone’s views based on posts that were not explicitly about politics?
Political scientists have never had a very precise definition for what we mean when we study “political discussion.” In surveys, we have always allowed our respondents to define it however they perceive it, and there is evidence that there is a wide variation in that definition. The definition is further muddled in the social media ecosystem.
Instead of focusing just on content that we as scholars think is about politics, we must broaden our scope to consider the effect of content that sends political signals. In my book, I advocate for the importance of “politically informative content.” Social sorting has led to increased alignment between our political and social identities, making it easier to infer someone’s political views based on their nonpolitical characteristics. Moreover, high levels of polarization in elite discourse have politicized many consumer choices. Is the person who unabashedly eats at Chick-fil-A a liberal or a conservative? Is the person who drives a hybrid car to the farmer’s market to purchase kale a Democrat or a Republican? As a result of this alignment and politicization, people can identify or surmise others’ political identities by interpreting social media content that the poster did not intend to be political.
This point is crucial when considering any potential solutions to how social media contributes to psychological polarization. Many of the ideas that have been tossed around—from increased civics education to moderated discussion platforms to a cessation of political argument online—are centered on fostering norms of civil and respectful political communication. But if even our nonpolitical communication has the potential to contribute to affective polarization, our solutions must also address that form of communication.
What Does “Exposure” Mean in a Social Media Context?
Once we have a better grasp of what social media content is politically relevant, we face the challenges of measuring the extent to which people are exposed to it. Scholars have long struggled to develop adequate measures of media exposure, and the difficulty is compounded on social media. If we want to capture the possibility that users encounter information that could influence their political attitudes or subsequent actions, what behavior should we measure?
Accurate measurement of exposure is a prerequisite for addressing a key question related to social media and polarization: how much cross-cutting content do people encounter? Cross-cutting information is broadly defined as content that endorses or espouses political viewpoints with which a person is inclined to disagree. We have typically conceptualized this as information coming from a news source that is perceived to be aligned with the political out-group, for example, what happens when a Democrat watches or reads stories from Fox News.
But, because of the fusion of social and political information on Facebook and the plethora of informative signals on News Feed content, “cross-cutting” information could be conceived of in a wider variety of ways. Users can see the media source of content on the feed, but they also recognize the political views of the person posting the content. Thus, cross-cutting information might be content coming from a person with which we typically disagree, even if the content itself is not disagreeable. Or, it might come in the form of a content thread, where both agreeable and disagreeable viewpoints are expressed by a mix of a user’s friends and the friends of their friends.
Therefore, accurately measuring exposure to politically relevant information first requires understanding the structure of social media networks that produce content. In theory, our online networks should be more heterogenous than the circle of people with whom we could interact about politics face-to-face. It is also easier to encounter political contention and divergent viewpoints on social media than it would be in a face-to-face context. Facebook data scientists have shown that the composition of our networks matters more than the site’s algorithm in structuring people’s information environments. This relationship between heterogeneity in network ties and heterogeneity in information exposure merits more future attention.
Once we have ascertained the potential volume of politically informative and cross-cutting information to which people might be exposed, we also need to understand how users process the information that does circulate. Political scientists have focused thus far on the effects of fairly sustained engagement with concordant versus cross-cutting content, such as measuring what content people click on the News Feed. But the threshold for influence from social media content is likely much lower. Social media users can be influenced by content that they don’t read or click.
“What kind of content do people remember seeing? Must people consciously process what they see to be influenced by it?”
We do not yet know which measures of exposure are appropriate for which questions. But we should move beyond just thinking about “clicking” as a measure of interest. We do not know much at all about gaze time, clicking to reveal comment threads, or hovering over the reaction buttons. What kind of content do people remember seeing? Must people consciously process what they see to be influenced by it?
Why Does Precision in Measurement of Exposure Matter?
There is an unspoken—and perhaps faulty—assumption that exposure to heterogeneity is preferable. Concerns about echo chambers or filter bubbles are generally premised on the idea that exposure only to information with which we agree has pernicious consequences. Many argue that democracy hinges on our ability to deliberate and engage with oppositional viewpoints and that exposure to cross-cutting information will moderate our political opinions and temper our attitudes toward each other. Valid measurement is an absolute requirement to test these assumptions.
First, there is some evidence that exposure to information from the other side can actually make people even more entrenched in their own viewpoints. Counterattitudinal messages have to be delivered in just the right way—perhaps as a narrative, perhaps with primes to be empathetic, etc.—to be persuasive, and it is not clear that information is delivered that way in the real world. Moreover, users who hear both their side and the other side tend to be even more convinced of the validity of their own views.
Second, we know that the source of counterattitudinal information matters for its effects. On social media, not all friends are created equal. Our strong ties likely matter more for influence, and there is evidence that personal recommendations can matter more than media source when selecting an information source. But we have many, many more weak ties and these weak ties are likely to hold more diverse political opinions than our strong ties. Unfortunately, cross-cutting information introduced by these weak ties might do more harm than good. Because users now have the ability to observe interactions among disagreeable others—to be flies on the wall, so to speak—they’ve become more judgmental about the political competence of these weakly connected social ties and impute those judgments to the out-party writ large.
There are a multitude of ways in which social media facilitates adverse consequences for the American political system, and many of these threats must be addressed by technology companies or government action. But Americans should not abdicate their responsibility as citizens, as we all play a role in shaping the public sphere. The choices we make on social media—what to post, what to click, and what to infer—should be made deliberately and in a manner informed by what social scientists have learned about the consequences of our actions.
1. Affective polarization refers to partisans’ increasingly negative feelings and negative trait attribution toward identifiers of the opposing party. See Shanto Iyengar, Gaurav Sood, and Yphtach Lelkes, “Affect, Not Ideology: A Social Identity Perspective on Polarization,” Public Opinion Quarterly 76, no. 3 (2012): 405–431.
2. Jennifer Fitzgerald, “What Does ‘Political’ Mean to You?” Political Behavior 35, no. 3 (2013): 453–479.
3. Lillian Mason, Uncivil Agreement: How Politics Became Our Identity (Chicago: The University of Chicago Press, 2018).
4. Markus Prior, “The Immensely Inflated News Audience: Assessing Bias in Self-Reported News Exposure,” Public Opinion Quarterly 73, no. 1 (2009): 130–143.
5. Eytan Bakshy, Solomon Messing, and Lada A. Adamic, “Exposure to Ideologically Diverse News and Opinion on Facebook,” Science 348, no. 6239 (2015): 1130–1132.
6. For more on echo chambers, see Cass R. Sunstein, #Republic: Divided Democracy in the Age of Social Media (Princeton, NJ: Princeton University Press, 2018). For more on filter bubbles, see Eli Pariser, The Filter Bubble: How the New Personalized Web Is Changing What We Read and How We Think (New York: Penguin Books, 2012).
7. R. Kelly Garrett et al., “Implications of Pro- and Counterattitudinal Information Exposure for Affective Polarization,” Human Communication Research 40, no. 3 (2014): 309–332.
8. Robert M. Bond et al., “A 61-Million-Person Experiment in Social Influence and Political Mobilization,” Nature 489 (September 13, 2012): 295–298.
9. Solomon Messing and Sean J. Westwood, “Selective Exposure in the Age of Social Media: Endorsements Trump Partisan Source Affiliation When Selecting News Online,” Communication Research 41, no, 8 (2012): 1042–1063.
Jaime Settle will appear at "Internet Speech and Democratic Politics," December 10, 5:00-6:30 pm.
Originally published in items: Insights from the Social Sciences, 4 September 2018.