Page of

(p.289) Part Four Can Democracy Survive the Internet?

(p.289) Part Four Can Democracy Survive the Internet?

Network Propaganda
Yochai BenklerRobert FarisHal Roberts
Oxford University Press

“Can Democracy Survive the Internet?” was the provocative title of an April 2017 article by Stanford professor Nathaniel Persily in the journal Democracy. It perfectly captured the deep anxiety that the election of Donald Trump, and before it the success of the Leave campaign over whether Britain should leave the European Union, created for many observers of politics across the North Atlantic. The former featured an outsider who violated every rule of what would historically have destroyed a campaign within weeks of its launch; the latter, a campaign to effect a radical departure from decades of British commitment to the European Union. Both seemed to win against all that conventional wisdom could muster. Something had happened that was fundamentally different from everything that had gone before, and the question was how to understand this radical departure. The core thrust of the anxiety was that the Big New Thing was the internet: technology changed how citizens and voters engaged each other; how elites engaged masses; and how people were, or were not, able to distinguish fact from fiction. Persily’s characteristically nuanced analysis suggested three primary effects. The internet destabilized established institutions—most importantly, political parties and media—allowing marginalized voices and outsiders to reach out directly to audiences, but equally so allowing demagogues and nihilists to disseminate propaganda and “fake news.” Anonymity and lack of accountability allowed Russia to inject its influence. And the internet enabled not only hypertargeted advertising but also the development of filter bubbles and echo chambers that made users embrace their side’s partisan messaging without question.1 In the previous chapters we already outlined our skepticism about the actual incremental impact of several factors: Russia’s very real efforts, commercial bullshit sites, and hypertargeted advertising based on psychographics. The Russians tried but were unlikely to have been a critical factor. The commercial (p.290) bullshit artists made some money, but were peripheral. And while Facebook’s data team certainly did make it possible for a complete outsider running with little help from party institutions to identify millions of voters and reach out to them effectively, the Cambridge Analytica manipulative advertising and the dark ads part of the story was still, in 2016, more of a red herring than the game changer some made it out to be.

The echo chambers and polarization concerns are recurring themes in discussions of the internet and democracy, at least since Cass Sunstein first published in 2001.2 The filter-bubble concern has been salient at least since Eli Pariser published The Filter Bubble in 2011. Sunstein’s basic argument, which remained at the core of the two later editions in 2008 ( 2.0) and 2017 (#republic), is that the internet and social media’s capacity to allow people to curate what they read, see, and hear will interact with our baseline social-psychological tendency to seek out evidence that fits our preconceptions (confirmation bias), congregate with others who are like us (homophily), and avoid information that does not fit what we know and like to hear confirmed. Furthermore, relying primarily on experimental work in the behavioral sciences, Sunstein argued that when people talk only to like-minded people, and do not get exposed to opposing views and arguments except in the context of oppositional argument, each group gets more entrenched in its own views and perceptions. Fragmentation of the media ecosystem leads to polarization both in the sense of segregated spheres and in the sense of increasingly extreme versions of the opposing views coming to the fore. The innovation in Pariser’s intervention was to shift agency from the many diverse users choosing to self-segregate, to the companies that design the engagement algorithms. For Sunstein, what drove polarization were choices people made when offered the ability to consume only bias-confirming news and hanging out only with allies. Pariser argued that it was the algorithmic choices of social media and internet firms that pushed people into these media consumption and communication practices. What drives us to see content that reinforces our views were not our own choices, but algorithms that observed us, learned our responses, and fed us more of what increased our engagements. Although in both arguments the solution was to get internet and social media companies to serve us more varied information diets, in Sunstein’s case this suggested that companies nudge us to see things that we don’t naturally gravitate toward. Sunstein’s argument is based on a conception of deliberative democracy, or perhaps republican civic virtue, which requires our engagement with opposing views and an agenda set by general news reports even if, left to our autonomous (p.291) choices, we would rather not. Pluralist views of democracy, by contrast, could be quite comfortable with highly segregated information spaces, in which groups contesting for political power define competing positions crisply and resolve their differences not by agreeing, but by peacefully counting votes at the polls. By contrast, Pariser’s diagnosis required a less normatively-freighted framework to justify roughly similar interventions. If the polarization and segmentation we see were not the outcome of decisions that citizens make for themselves, but rather the product of choices companies make for their users, then regulating these companies to offer a more varied diet that was healthier for democracy did not require us to adopt any particular version of democratic theory. All that was needed was a commitment to the proposition that members of a democratic polity should be able to choose for themselves what they read, see, and hear, rather than have it thrust upon them by companies whose interests are fundamentally commercial, not political. The negative political communications patterns that resulted from these company choices were simply emissions from the commercial companies into the political media ecosystem; byproducts that needed to be regulated to keep from harming democracy.

Our data, particularly the highly asymmetric structure of the networked public sphere, question both diagnoses. If “the internet” is what leads to polarization and increasing extremism we observe, then one would expect to see the media ecosystem develop symmetric patterns of polarization. We should have expected to see (as early studies relying on less data did see) a symmetric pattern of segmentation, assuming that users on the left and the right operate under similar social psychological dynamics and algorithmic decisions. What we observe instead is that everyone outside of the roughly 30 percent of the population that pays attention primarily to right-wing media exists in a mixed-media ecosystem that is not fragmented and is more or less normally distributed in its attention around a core of traditional professional media outlets. People on the left certainly read left-oriented materials, but they also read and engage with publications and media outlets anchored in the professional journalistic norms. These are precisely the kinds of generalized information intermediaries that played the role of shared agenda setting from the end of World War II to the 1980s. . Moreover, polarization is a process that took a sharp upturn among political elites in the mid-1970s, well before TCP-IP had even been defined as the internet protocol or Mark Zuckerberg was born. Rush Limbaugh had already been declared honorary member of the class of Republican House Freshman when Newt Gingrich led the Republican “Contract with America” 1994 capture of the House. (p.292) Both the extensive political science literature on polarization and the rich literature on American media history argue against “the internet did it” narratives. Instead, we suggest that technological, institutional, and political dynamics have been interacting for over 40 years to lead the Republican Party and Republican voters to gradually become more extreme versions of themselves, without operating symmetrically on the Democratic Party and its supporters or on most Independents. The 2016 presidential election was the moment at which these long-term dynamics reached an inflection point in the Republican Party, but not among Democrats. These dynamics explain how Jeb Bush, the son and brother of the two most recent Republican presidents, can be castigated for his “close Nazi ties” by the eighth-most tweeted media site on the right; or how Sean Hannity and his guests on the most highly watched cable television program on the most highly watched cable news channel can vilify lifelong Republican law enforcement agents as agents of “the deep state” who are personally corrupt and conflicted out of a major national security investigation. As we have seen repeatedly throughout the preceding chapters, the prominent outlets on the left and center simply do not exhibit a parallel structure, content, or vehement outrage that we observe on the right. These facts are as inconvenient to academics seeking a nonpartisan, neutral diagnosis of what is happening to us as they are to professional journalists who are institutionally committed to describe the game in a nonpartisan way. Both communities have tended to focus on technology, we believe, because if technology is something that happens to all of us, no partisan finger pointing is required. But the facts we observe do not lend themselves to a neutral, “both sides at fault” analysis. Our lived experience is one in which a highly partisan House committee “absolves” the president of allegations that are under active investigation by an independent counsel and that president then tweets: “As the House Intelligence Committee has concluded, there was no collusion between Russia and the Trump Campaign. As many are now finding out, however, there was tremendous leaking, lying and corruption at the highest levels of the FBI, Justice & State. #DrainTheSwamp.”3 Only a conscious and frank assessment of why the two sides of the divide developed such different dynamics will help us avoid the vortex.

In Chapter 10, we review the literature on polarization and see that polarization long precedes the internet and is rooted in asymmetric political-elite-driven dynamics. In Chapter 12, we turn to media history and recount the rise of second-wave right-wing media, beginning with Rush Limbaugh and the shift of televangelism into political coverage. Again, asymmetric (p.293) polarization precedes the emergence of the internet. We then turn to evaluate various hypotheses about why the differences between the left and the right emerged as they did and cover both psychological explanations and historical-institutional factors. We cannot exclude psychological explanations that claim to identify systematic cognitive and affective differences between conservatives and liberals, but we offer reasons to see the asymmetry more as a function of historical contingency and institutional-political factors interacting with universally applicable models of motivated reasoning than a result of any intergroup psychological differences. As such, we think that while path-dependency will make addressing our present crisis difficult, it will not require going against “human nature” in any deep sense, but “merely” formidable political will on the part of both Republicans who still reject their party’s takeover by its radical wing, and Democrats who will have to work with those Republicans to re-establish a more symmetric political and media ecosystem. (p.294)


(1.) Nathaniel Persily, “Can Democracy Survive the Internet?,” Journal of Democracy 28, no. 2 (April 2017): 63–76.

(2.) Cass R. Sunstein, (Princeton, NJ: Princeton University Press, 2001).

(3.) Donald J. Trump, Twitter, March 17, 2018,

Copyright © 2021. All rights reserved.
Privacy Policy and Legal Notice