Polarization as the technological goal – not the error
Wendy Hui Kyong Chun and Sonja Solomun
In this interview, Wendy Hui Kyong Chun, in conversation with Sonja Solomun, reflects on the state of polarization in Canada today, drawing on her work on homophily, mis- and dis-information, and the question of authenticity. She explains that we must “challenge the simple separation between on- and off-line” spaces, especially considering the increasing importance of technology to social and political life. Doing so will help us understand how social media platforms – and much of the Internet, for that matter – operate as “deep-fake versions of reality” that alter how we tell “truthful, nuanced, and complex stories about our past, present, and future.” In arguing that polarization can (and should) be seen as a technical goal rather than an error, Chun presents a thought-provoking approach to discourse on polarization and media that can help move policy response beyond technological fixes.
Sonja Solomun [SS]: There is often a public perception that polarization has both sharply increased and, at the same time, that Canada is somehow exempt from the patterns and the extent of what is happening in other countries around the world. What is your general take on the state of polarization in Canada today?
Wendy Hui Kyong Chun [WHKC]: As events that unfolded around and during the pandemic have shown, Canada is not immune from polarization, and comparing Canada to countries such as the U.S. should create a sense of urgency rather than complacency. I moved back to Canada from the U.S. and established the Digital Democracies Institute because the playbook used to amplify divides and discrimination is being tested and deployed globally. Canada is vulnerable to tactics popularized in the U.S. not only because of geographic proximity, but also because so many Canadians get their news and entertainment from the U.S., much of which is also produced in places like Vancouver (also known as Hollywood North)—the cultural border is porous at best. The Canadian imaginary is deeply infused by U.S. notions of freedom and its geopolitics. At the same time Canadians read/watch U.S. news and think “thank X that’s not me” (this is arguably the strategy that drives Canadian news outlets), they also ask: why isn’t that us? Or, they find themselves encompassed in the U.S. “us.”
[SS]: There are – unfortunately – any number of Canadian examples that reflect that, including the rising anti-Asian racism during the pandemic. That particular example also underscores what sometimes gets taken for granted in thinking through certain kinds of solutions to the problem, which is that polarization doesn't just stay online. Can you expand on that a little bit?
[WHKC]: Given how much of our life happens through and via technological media, I wonder who thinks that the online and offline are separate. From online government applications to Internet scams, from youth aspiring to become influencers to Zoom meetings, it seems impossible to function without being online. Your question, though, focuses us on the question of the consequences of polarization and hate speech, and it’s here where many insist on a difference. As Danielle Citron and others have shown, law enforcement has taken a long time to take online gender-based violence seriously because—against all evidence to the contrary—there’s a sense that what happens online isn’t real. So, death threats uttered online are inherently not real—more playful or ironic—when issued online. We need to challenge this simple separation between on- and offline so that we consider language more broadly. For example, in casual conversation, someone can say “I could kill X” without actually intending to kill X. This depends on context, intention, audience, etc.—so instead of thinking that speech on- and off-line are radically different, we need to understand and come up with policies that address the specificities of threats in ways that encompass both spheres.
[SS]: These questions really get us into the role of technology more broadly. What is the role of technology in enabling and maintaining some of these outcomes or to what extent is the technological infrastructure of platforms and online spaces designed to segregate?
[WHKC]: Current commercial social media platforms and recommendation systems are designed to segregate—polarization is a technical goal not an error. A lot of these sites seek to give you what you want—they want to give you advertising and recommendations that you’re interested in. Since most platforms don’t have enough data points about you, this “personalization” is not based on your actions but on actions of people determined to be “like you,” that is, part of your network neighborhood. The goal is to determine the proxies that are most efficient at creating clusters of interest—and these clusters often intersect with protected categories. For example, the fact that you like Wu-Tang Clan allegedly reveals your sexuality (straight man). What’s key about these neighborhoods is that they’re filled with people who are like you—it makes segregation the default.
[SS]: This gets us right into the subject of homophily, which has been central in your work. What is homophily, and why is it important to question our core assumptions?
[WHKC]: Homophily is the principle that similarity breeds connection or that birds of a feather flock together, and it's at the core of recommender systems. According to Miller McPherson, Lynn Smith-Lovin, and James Cook in their definitive review article on homophily, “the homophily principle structures network ties of every type, including marriage, friendship, work, advice, support, information transfer, exchange, co-membership, and other types of relationships.”[i] Because of homophily they thus argue, “people’s personal networks are homogeneous with regard to many sociodemographic, behavioral, and intrapersonal characteristics.” [ii] Homophily launders hate into love by treating racism as a “naturally” occurring preference within “human ecology.”[iii] It transforms individuals into “neighbors” who naturally want to live with people “like them”; it presumes that consensus stems from similarity; it makes segregation the default.
Homophily’s relationship to segregation is not accidental but fundamental. Indeed the term homophily stemmed from a study of U.S. residential segregation, specifically a study of a biracial housing project during the mid-20th century. Following World War II, there was a housing crisis in the United States and at that time 70% of all Americans supported public housing; but private developers were pushing for complexes like Levittown, supported by mortgages guaranteed by U.S. government that openly discriminated against black Americans and mixed neighborhoods. We know what happened; we know who won and the resulting wealth disparity. This study tried to determine the impact of public housing on democratic engagement through four-hour long interviews with a member of every family living in two housing projects: a worker cooperative code-named Craftown based in New Jersey, and a bi-racial housing project in Pittsburgh code named Hilltown. The researchers focused on morale, which they viewed as central to democratic engagement, and they also sought to measure the effectiveness or not of housing projects as forms of social engineering by measuring friendship formation in these projects. Now, they did not presume that homophily is absolute, in fact they coined the terms homophily and heterophily. And they did not view homophily as an explanation but rather as a phenomenon that needed to be explained. They showed that in Addison Terrace homophily wasn’t necessarily true, except in two cases, gender and race. And in order to explain why this was happening, they argued that status homophily, choosing friends according to a similar status, was actually driven by value homophily, choosing friends according to shared values.
To prove this, the researchers examined and modelled the racial attitudes of Hilltown’s white residents. They focused on the answers to the following three questions from the extensive interviews: “Do you think colored and white people should live together in housing projects?”; “On the whole, do you think that colored and white residents in the Village get along pretty well, or not so well?,” and “Could you tell me who your closest friends are (regardless of whether or not they live in the Village)”?[iv] Based on the answers, they divided the white residents into three camps: liberals, who “believe that ‘colored’ and white people should live together in housing projects and who support this belief by saying that the two racial groups ‘get along pretty well’ in Hilltown”; illiberals, who “maintain that the races should be residentially segregated and who justify this view by claiming that, in Hilltown, where the two races do live in the same project, they fail to get along”; and ambivalents, who “believe that the races should not be allowed to live in the same project, even though it must be admitted that they have managed to get along in Hilltown.”[v] They then analyzed friendship patterns among white residents with white friends who resided in Hilltown and determined that liberals overselect other liberals by 43 percent; illiberals overselect other illiberals by 30 percent; liberals underselect illiberals as close friends by 53 percent; illiberals underselect liberals by 39 percent; and ambivalents do not overselect or underselect.[vi]
To make this argument, they ignored the responses of Black residents. They removed their answers from their analysis of value homophily because there were “too few illiberal or ambivalent Negroes with friends in Hilltown.”[vii] Thus, at the core of value homophily lies racial segregation—an implicit assumption that values do not cross racial borders, or if they do, that this crossing is less significant than consensus or conflict within a race—and racial exclusion.
What’s key is that they presented these percentages without absolute numbers, numbers that they promised—in a footnote—would be forthcoming in a longer report, “Patterns of
Social Life,” written by BASR researchers Merton, Patricia J. West, and Marie Jahoda.[viii] In another footnote they write, “It must be emphasized that such extreme concentration of personal ties within each racial group obtains only for the most intimate friendships. (It will be remembered that these data refer to the three closest friends of residents.) Short of these most intimate attachments, however, there have developed numerous personal relations across race lines in Hilltown, as will be seen in the complete report, Patterns of Social Life.”[ix] This forever forthcoming report—it was never published—actually discounts the statistical significance of the overselection of illiberals for other illiberals, since the total number of friendships was under fifteen.
What’s amazing is that this highly qualified discussion of homophily—one that points out its limitations and the diversity of relations around us—has become the justification for the naturalness of homophily. But, even within ecology, birds do flock together. The world of homophily makes things like heterosexuality and electricity anomalies.
[SS]: In your work, you maintain that polarization is the goal, not the error. Given this history, it feels urgent to unpack that.
[WHKC]: Yes, the process of polarization—and the transformation of mass media into new media—recalls the classic physics experiment in which a solid mass of inert iron pilings is magnetized and pulled into clustered networks. The similarly charged filings gathered at either pole also repel each other, but they are stuck together by their overwhelming attraction to their opposite. Sustaining this magnetic charge in usually non-magnetic materials requires a previously-charged magnet or a constant current. The “neighborhoods” discussed in these media are like these clusters of charged filings, in which similarly-charged shards are both repulsed and welded together through their overwhelming attachment to their opposite.
Similarly, networking algorithms can create agitated clusters of comforting rage, in which people who angrily repulse each other are held together by their overwhelming attraction/hatred of something else. Intriguingly, homophily is usually justified in terms of comfort—you’re allegedly more comfortable when you’re with people “like you.” However, as Frances Haugen’s whistle blower testimony has shown, when Facebook changed the feed so that it's just your friends and family, it became even more toxic.
[SS]: There are deeply political implications here, especially related to the kind of power asymmetry that Haugen and others have highlighted. Yet, there is a tendency to depoliticize disinformation and polarization, which are very intertwined by focusing exclusively on the information ecosystem. In doing so, disinformation gets treated as a purely technological or media problem rather than the historically constituted political and racial problem that it is. What does history afford us in understanding the state of polarization today?
[WHKC]: Amongst other things, history tells us that we can't fix a political problem by fixing technology—and that seeking technical fixes to political problems doesn’t work. For example, the Nazis used radio to spread disinformation—but the solution to the atrocities committed during WWII was not: let’s fix radio. We need to acknowledge that technology matters—and it can be part of a solution and problem—but that it's part of this larger system. I think the example of global climate change models makes this clear. When a model we trust says that global mean temperature will rise by 3 degrees Celsius, we seek to fix the world not the model.
[SS]: Exactly. History also allows us to recognize how successful disinformation campaigns mobilize long standing myths about inequality that drive polarization today. One area where this comes up is through climate obstruction online, which is now being termed “climate disinformation” to categorize the long-standing problem of climate denial and the well-worn strategies used to delay response. Does online polarization present a new challenge for climate change or does it replicate established concerns we've had about climate denial for decades?
[WHKC]: In terms of homophilic clustering technologies, what’s key is that these techniques can be used to string together different clusters, based on their shared hatred/love of another group. Once these clusters are in place, the goal is to string them all together into a strand of pearls to form a majority. Many studies have shown that COVID-19 deniers were drawn from existing clusters of conspiracy theorists, climate change deniers, folks distrustful of government, etc.
In terms of climate change denial, it’s interesting that it’s being framed as a problem of disinformation, which implies that if the correct information was provided, everything would be OK. Intriguingly, global climate change scientists such as Jim Hansen have moved away from making this claim to arguing that a main difficulty—which makes disinformation so compelling—is that we experience weather not climate.
[SS]: Exactly, which also brings to bear the political implications of certain framings to this debate. Climate denial is not about truth versus falsehoods—which propagate an endless debate and we empirically know the facts—it is about ordering and “staging” information in a particular way for a particular goal, often distrust. We’ve also seen that fact-checking is a very limited, partial solution—so are we framing the problem in the right way?
[WHKC]: Fact-checking is important but not enough: fact-checking sites lag behind the deluge of rumors produced by disinformation sources and spread via private interactions. Also, verification alone does not dispel misinformation (inadvertent) and disinformation (intentional): corrections and 'fake news' stories often reach very different audiences; corrections can create new interest in debunked stories; users spread stories they find compelling or funny, regardless of their accuracy. Tellingly, the 2016 U.S. presidential election was both described as "the authenticity election" and as normalizing 'fake news.'
Rather than presuming this means that we now live in post-truth era, our team at the DDI have been exploring the centrality of authenticity to the viral spread of mis/information. Our goal is to provide insight into how and under what circumstances—social, cultural, historical, and technical—information is deemed 'truthful.'
[SS]: That duality is so central to the concept of authenticity and Sarah Banet-Weiser has done crucial work on this, where the mandate to be authentic is always to be both more and less, to present as authentic becomes more urgent even as we simultaneously no longer believe in the concept of the authentic. I know you have been grappling with some of these questions at the Digital Democracies Institute, under what circumstances do people find information to be “authentic”?
[WHKC]: DDI researchers Prem Sylvester, Sage Hughes and Matthew Canute just did a recent study of stories that covered anti-Asian racism during COVID-19, or otherwise intervened in the discourses around such violence, and we found that these news stories don’t just present a collection of facts. They are written, or scripted, to find an audience through certain claims that appear factual and deploy rhetoric of inclusion and exclusion that mark the boundaries of who is, for example, ‘authentically’ Canadian or Asian-Canadian. Mis- and dis-information around these matters that invite social and political passions, then, rests not simply on ‘facts’ but as much about subjective, cultural, and social relations.
This also means that to understand what’s happening, we need to think beyond automatic detection to thinking more broadly about questions of transparency, accountability, and empowerment. This is something that the Commission for the Canadian Commission on Democratic Expression—which I had the honor to serve on —and the Citizen's Assembly – have been thinking about.
[SS]: The empowerment piece is so crucial here. One way you are currently working towards that is through the data fluencies project in particular. Can you tell us a little more about why that feels urgent?
[WHKC]: The Data Fluency Project explores, analyzes, and seeks to counter the impacts of mis- and disinformation on cultural diversity by developing expansive and interdisciplinary data fluencies. Moving beyond literacy, data fluencies combine the interpretative traditions of the arts and humanities with critical work in the data sciences to express, imagine, and create innovative engagements with (and resistances to) our data-filled world. This involves bringing together groups that have been working largely in parallel, non-intersecting tracks: social-justice-oriented research and pedagogy in the arts and humanities and similarly focused computational work; Indigenous media studies researchers and intersectional technology developers; arts-based data literacy efforts and data science curricular development. We’re focused on:
creating mixed qualitative-quantitative research to i) help diagnose how discriminatory and misleading information spreads and is authenticated on major social media platforms and ii) test methods and tools to counter these dynamics;
creating alternative language-based systems that meet the needs of communities usually considered to be at the “fringe” and forensically analyzing existing language systems to understand how and why they discriminate;
building new data fluencies by expanding whose voices matter and how they do through community-led data centers, public night schools, exhibitions, performances, intersectional technology development workshops, and media projects to deter white nationalist radicalization;
mainstreaming these approaches through media production, courses, small research grants, and workshops.
We think this is key because discriminatory mis- and dis-information make many social media feeds and much of the Internet “deep fake” versions of reality, therefore threatening our ability to tell truthful, nuanced, and complex stories about our past, present, and future. Left unchecked, mis- and disinformation and the current algorithmic solutions to them—which reduce the future to the past and thus often automate, rather than learn from, past mistakes—make it impossible for us to form complete and accurate narratives about our shared humanity. They also undermine the dialogue necessary for a robust democracy. To help address these fundamental challenges to truth, history, and knowledge, we are building models and prototypes of data fluencies—an approach that supplements efforts by digital story-tellers, archivists, critical ethnic and race scholars, and digital humanists, amongst others, to bring rich humanistic thinking about learning, history, narrative, and critique to bear on our data-filled world.
[SS]: Polarization is sometimes talked about in policy and public debates as something that needs to be eradicated for democracy to function or flourish or both. Democracy is itself an unfulfilled promise—as Astra Taylor puts it in her book, democracy may not exist, but it still manages to disappoint, right? Do we need to eradicate polarization?
[WHKC]: Astra Taylor addressed the hatred of equality that is at the center of so many threats to/revisionist theories of democracy. Many techniques to create polarization seek to undermine commonalities and structures that foster equality. Importantly, there’s a difference between homophily and community, even though the former is used to explain and undermine the latter. Homophily makes everything about individual choice and thus glosses over the effort and infrastructures to create community—and also sustain discrimination. What’s also dangerous about polarization and network neighborhoods as they exist now is that they are segregated. What is missing are these kinds of engagements or what Danielle Allen would call talking to strangers, which is the fundamental negotiation of democracy.
Endnotes[i] McPherson, M., Smith-Lovin, L., and Cook, J.M. (2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology 27, 415.
[ii] McPherson, M., Smith-Lovin, L., and Cook, J.M. (2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology 27, 415.
[iii] Chun, W. (2021). The space between us: Network gaps, racism, and the possibilities of living in/difference. Catalyst: Feminism, Theory, Technoscience.
[iv] Lazarsfeld, P. and Merton, R.K. (1954). Friendship as a social process: A substantive and methodological analysis. In Berger, M., Abel, T., and Charles, H., Eds., Freedom and Control in Modern Society, 26.
[v] Lazarsfeld, P. and Merton, R.K. (1954). Friendship as a social process: A substantive and methodological analysis. In Berger, M., Abel, T., and Charles, H., Eds., Freedom and Control in Modern Society, 26.
[vi] Lazarsfeld, P. and Merton, R.K. (1954). Friendship as a social process: A substantive and methodological analysis. In Berger, M., Abel, T., and Charles, H., Eds., Freedom and Control in Modern Society, 27-28.
[vii] Lazarsfeld, P. and Merton, R.K. (1954). Friendship as a social process: A substantive and methodological analysis. In Berger, M., Abel, T., and Charles, H., Eds., Freedom and Control in Modern Society, 27.
[viii] Lazarsfeld, P. and Merton, R.K. (1954). Friendship as a social process: A substantive and methodological analysis. In Berger, M., Abel, T., and Charles, H., Eds., Freedom and Control in Modern Society, 27.
[ix] Lazarsfeld, P. and Merton, R.K. (1954). Friendship as a social process: A substantive and methodological analysis. In Berger, M., Abel, T., and Charles, H., Eds., Freedom and Control in Modern Society, 22.
