HARVARD LAW professor, member of President Obama’s White House, and the most-cited constitutional law professor in the country, Cass Sunstein first addressed the internet’s role in the polarization of our culture 15 years ago. His argument has become an accepted part of any discussion about the net’s failings. Yet, he has now written the same book about this issue for the third time. To be sure, his persistence is laudable, but the problems with his assumptions are only becoming more obvious.
The first appearance of this book was as Republic.com in 2001, which Sunstein updated in 2007 as Republic.com 2.0. The latest version, #Republic, retains most of the text from these prior versions with a shift in evidence and examples from websites (2001) and blogs (2007) to social media (2017). The new edition includes elaborations inessential to his argument.
Why three times? Perhaps Sunstein is the prophet still unheard in his own land — although, given the prominence of his views, that’s not very plausible. Or perhaps the sins he decries have not become dire enough for us to act. Or perhaps we have failed to achieve the old ideals of civil discourse that he espouses because our new network is exposing a weakness in them.
If you give people a choice among an infinite supply of media, argues Sunstein, they will gravitate toward content that confirms their existing opinions. Let people connect with whomever they want, and they will connect with those who share their views. Their conversations will then reinforce their beliefs — and, worse, drive them to more extreme versions of those beliefs. They will, in short, form echo chambers.
The internet satisfies those conditions: it gives us access to a galactic selection of content and enables us to find others who share our beliefs, down to our micro-preferences. For Sunstein, this explains why our culture has become so much more fragmented and polarized.
According to him, nothing less than the fate of the republic hangs on recognizing and forestalling this danger. At its heart, he argues, the United States is an experiment in deliberative democracy: “[T]he framers’ greatest and most original contribution to political theory,” he writes, was the idea that “heterogeneity, far from being an obstacle, would be a creative force, improving deliberation and producing better outcomes.” Deliberative democracy thus requires that people who disagree be able to talk with one another, constructively and openly, in this way collectively discovering which beliefs are worth holding. Echo chambers, he worries, polarize us to the point that we are unable to have those conversations, and they thus pose a severe threat to democracy itself.
If this sounds old-fashioned, it is, in the same sense that the US Constitution is old-fashioned. Sunstein’s references are to John Stuart Mill, Justice Louis Brandeis and Judge Learned Hand, Thomas Jefferson, and other heroes of civil discourse.
Like much of Enlightenment thinking, his argument is framed by a sense of the ideal. “[M]y central claims are not empirical,” he writes at the beginning of the book, “they are about individual and social ideals. They are about the kind of culture that is best suited to a well-functioning democracy.”
He includes as well a sense of what ideal deliberations ought to be like. They don’t sound much like what happens on the internet. Instead, they sound like how we might imagine the great minds of the Enlightenment conversed, perhaps when gathered over sherry. Sunstein approvingly quotes Jürgen Habermas’s idea of an “ideal speech situation”:
Rational discourse is supposed to be public and inclusive, to grant equal communication rights for participants, to require sincerity and to diffuse any kind of force other than the forceless force of the better argument. This communicative structure is expected to create a deliberative space for the mobilization of the best available contributions for the most relevant topics.
For Sunstein, these are the real conversations we ought to be having. Such conversations force us to sublimate our personal interests and emotions, which do violence to our ability to rule ourselves as individuals and as a collective. A democracy needs to be deliberative not only to arrive at sound decisions, but also to show its citizens that even when they are being coerced into behavior they voted against, the coercion isn’t really coercive because it arose from the “forceless force” of a deliberative process.
The internet, according to Sunstein, has unleashed novel and insidiously coercive forces arising from its very structure and nature as a medium. Its openness and capaciousness trigger a cognitive form of optical illusion. Giving a human brain unlimited access to ideas will draw its owner to those sources and conversations that confirm what it already believes.
Because his argument is tied to essential properties of the internet’s architecture, it constitutes a potent form of media criticism, but I believe it springs from — and confirms — assumptions about ideal discourse that lead Sunstein to the wrong villain: echo chambers. 
The Problem with Echo Chambers
#Republic, like its two prior versions, is a carefully thought-out book that anticipates objections. Sunstein acknowledges, for instance, that extremism isn’t always bad; the Abolitionists held beliefs that were extreme for their time.
Likewise, he concedes that not everything that looks like an echo chamber is one, including “enclave deliberation,” which he defines as “that form of deliberation that occurs within more or less insulated groups, in which like-minded people speak mostly to one another.” He finds value in such enclaves: they enable the easily intimidated or disenfranchised to find their voices. This is important because “in deliberating bodies, high-status members tend to speak more than others,” disadvantaging those who have traditionally been excluded from power. “Properly understood,” he writes, “the case for deliberating enclaves is that they will improve social deliberation” by “incubating new ideas and perspectives that will add a great deal to public debate.” In short, deliberating enclaves have value because they foster diversity — underheard opinions emerge within them, enabling enclaves to present those ideas in public forums where divergent enclaves engage with one another.
There’s much to be said for this fractal diversity. But Sunstein’s devotion to it misses what seems to be the most obvious “great benefit of deliberating enclaves”: almost all progress comes from within them. The preponderance of advances in evolutionary biology have been made by scientists who agreed about that theory and talked — in person and through published papers — with other believers. That progress did not occur in their conversations with creationists, nor did it alter creationists’ beliefs. Likewise, climate scientists did not make progress in using machine learning to model the climate by engaging in sincere conversations with climate change deniers. Shakespeare scholarship’s understanding of the role of race in Othello has been advanced by those who not only believe the Bard is worth studying, but who also broadly agree on the core texts and principles of literary scholarship; haters of the humanities and racists were not a part of those discussions. In short, disciplines are defined not just by their subject matter, but by their practitioners’ agreement about the discipline’s basic tenets, its rules of evidence, its norms of discourse, the role of credentials, and so on. Even when a discipline is disrupted by what we sometimes still like to call a paradigm shift, the disruption usually comes from within.
The main value of deliberating enclaves therefore is not, as Sunstein proposes, that they increase conversation across differences, but that they enable like-minded people to make progress in what they agree about.
So what differentiates such enclaves from echo chambers? Despite his frequent use of the term and careful step-by-step argumentation, Sunstein never defines echo chambers. Yet he uses the phrase in the very first line of the preface: “In a well-functioning democracy, people do not live in echo chambers or information cocoons.” If, as “cocoon” suggests, an echo chamber is defined by its self-imposed isolation from dissenting voices, we should be able to tell if a group is an echo chamber by counting its links to oppositional viewpoints. But, according to that analysis, a climate scientist mailing list or a Civil War historians’ Facebook page is likely to be as much of an echo chamber as climate denier and UFO conspiracy theory discussion boards.
Further, none of these groups look like echo chambers from within. The inhabitants of deliberating enclaves and echo chambers think they are disagreeing with one another about issues. The climate scientists are disagreeing about when the next arctic ice mass might detach. The Hillary haters are disagreeing about which is worse, her corrupt charitable foundation or her pantsuits. The ufologists are debating about whether last night’s bright light over Bayonne, New Jersey, was a probe from Arcturus or a Venusian water trawler.
These all look like echo chambers from the outside because the disagreements that advance our thinking necessarily occur on the surface of an enormous ocean of agreement about foundational beliefs, what’s worth talking about, norms of conversation, and so on. Conversation takes this form because it’s the form of understanding itself: we understand something new by assimilating it into our existing context. We put it into relation with what we already understand. Doing so not only makes comprehensible what was unfamiliar, but also reinforces our way of understanding. That is a feature, not a bug.
The real problem with echo chambers therefore isn’t that they consist of people who believe the same things and whose discussions strengthen their beliefs. The real problem is that some of them are wrong — in their beliefs, their methodology, or, often, in both.
Fifteen Years of Solutions
The seven solutions that Sunstein discusses in the penultimate chapter of the latest version repeat, with only minor tweaks, the proposals from the prior two. Some of the seven have been tried unsuccessfully and the reasons why no one has attempted the others have not changed. There’s little reason to think these solutions are going to work now any better than they did 15 years ago. The real question is why Sunstein persists in thinking that they will.
The proposals’ overall aims are twofold: to build new spaces where citizens can engage with one another, and to nudge existing media and websites into putting challenging ideas and information in front of their users’ eyes. His solutions propose more transparency from media and platform sites, incentives and possibly some gentle regulation to ensure media “cover substantive issues in a serious way,” and ideas for how sites can lead users toward diverse content.
As if recognizing the implausibility of his proposals, he inserts phrases such as “it might be pie in the sky, but …” and “we could easily imagine …” When suggesting that Facebook could present opposing viewpoints by default, he acknowledges that “some users would not love that,” but ends with an enthusiastic: “Count me in.”
Overall, these proposals aim at increasing serendipity. This is a running theme in the book, and it is odd. “My largest plea […] is for an architecture of serendipity,” he says on page five. Yet, by his own analysis, it is the oversupply of diverse content that got us into this position in the first place. If Facebook wants to give a Hillary supporter a load of Trump content, either by injecting it into her stream or by giving her buttons she can click to fetch it, the user is unlikely to read it for the very reasons Sunstein says echo chambers form in the first place — not to mention Facebook’s strong economic motives for feeding its users content that they will like and Like.
I will certainly regret going for the cheap irony here, but it’s as if Sunstein is in his own echo chamber. His proposals make sense in an idealized Enlightenment world in which we are rational agents momentarily distracted by sparkly objects. Perhaps with just a couple more editions of the book — Republic VR, Republic Embedded AI — we’ll see the light and opt for reading the oppositional posts that we have ignored.
Except we now have 15 years of evidence that we won’t. Sunstein’s “Count me in” is a cry that will inspire only his fellow believers. Granted, his echo chamber is admirable. And his book is a model of this type of discourse. It’s a model that works for highly educated scholars and intellectuals. It is to be valued, prized, and imitated, which I say as a member of the same privileged class as Sunstein.
But it’s an ideal that likely overestimates diversity in the prior media regime, including the general interest media that Sunstein pines for. Given a lack of pre-internet data and the ambiguity of the terms — what counts as “encountering” or “paying attention”? — I don’t know how we can resolve this question. But I do know that the pre-internet household I grew up in subscribed to the newspaper that best reflected my parents’ political leanings, and when it carried stories about places and topics outside my accustomed interests, I generally just turned the page.
Sunstein is also more convinced than I am that if you live in an online echo chamber, then that is all you ever see of the internet. In fact, it’s more difficult than ever to shutter oneself from divergent ideas. You can’t do a Google search without being shown links that may poke holes in your echo chamber. You can’t watch a tweet stream or the front page of Reddit without seeing a stream of general interest items, at least some of which are not particularly at home in your echo chamber. You can’t be on the net without coming to realize that there are uncountably more ideas and values than your own. Perhaps media theorist Clay Shirky is right that at least some of our current polarization comes from seeing just how many people believe things that are crazily, dangerously different from our own beliefs.
It may simply be time to give up on the Enlightenment ideal of discourse as the sole model and measure of human conversation. Yes, there are certainly times and places where that style of discussion is essential. But holding it as the ideal can narrowly focus our gaze on the worst aspects of the new models of discourse emerging online. It need hardly be said that the worst is truly horrifying: echo chambers where anorexic young women egg each other on to further excesses, discussion boards where racists taunt each other into real-world violence, and the dismal like. Of course we should address such sites as serious problems, but not as emblematic of internet discourse overall, else we risk missing what may be new forms of political deliberation, and even a new implicit idea of what it means to be together with other humans in a shared world.
Although nothing is truly typical of the internet, just as nothing is typical of speech, there are some remarkable developments emerging in informal online discourse. Some of the popular political forums on Reddit, for example, provide rapid crowd-sourced fact-checking and sense-making. Tools like Genius.com’s web annotator and Hypothes.is enable open annotation of any page. More broadly, knowledge is now routinely developed in public as something to be shared. We are seeing an ever-growing explosion of links that embody collective meaning-making. We’re edging toward the recognition that we each talk from our own situatedness. We recognize differences without necessarily thinking they can or should be overcome. We read staccato-like exchanges that express a different musical style than calm, cool deliberation. We get to control how we present ourselves, allowing for a more multisided expression of who we are — as well as enabling mockery, fakery, and scams. We see truth disclosed through humor and even a rebirth of wit. We see new forms of friendship, new forms of knowledge, new forms of higher-order meaning.
Most of all, we see a persistently noisy self-organizing and self-complicating mess that refuses to resolve, resulting in a web of inconsistent and simultaneous meanings. But this is not noise. It only sounds like noise outside of our own echo chambers.
Never before have we lived amid such serendipity. But we have not yet learned how to appreciate ideas that surprise and even irritate us. Therefore, a big part of the solution to the problem Sunstein describes is education. Our educational systems need to do a better job of teaching us not just what we should know, but how to know in the age of the internet. How can we assess the worthiness of what we have just read? How can we use the affordances of the net — links, lookups, collaborative public sense-making, algorithms — to arrive at belief? What are the rules of evidence of the domain we are in? What are the forms of social engagement that are most likely to lead us toward truth in that particular domain? How can we learn to love difference? How can we learn to thrive on noise as well as signal?
Perhaps it will take a generation that accepts the internet as its natural habitat to rewrite knowledge and respeak discourse itself.
In the meantime, at this point in our history, after three editions, we and Sunstein need to be asking: What’s our Plan B?
David Weinberger writes about the effect of technology on ideas. His latest book is Too Big to Know: Rethinking Knowledge Now That the Facts Aren’t the Facts, Experts Are Everywhere, and the Smartest Person in the Room Is the Room.
 Recent research shows that the right wing’s discussions are more insular than the left’s, suggesting that the causes of our polarization may have more to do with the hyperpartisanship of the mainstream right-wing media than with the structure of the internet. Yochai Benkler, Robert Faris, Hal Roberts, and Ethan Zuckerman, “Study: Breitbart-led right-wing media ecosystem altered broader media agenda.” See also Michael D. Conover, Bruno Gonçalves, Alessandro Flammini, and Filippo Menczer, “Partisan asymmetries in online political activity.”