Understanding — and Countering — Hate Speech Is the Fight of Our Age

Eight years ago, a Nor­we­gian white suprema­cist named Anders Behring Brievik mur­dered 77 peo­ple, many of them chil­dren, in and around Oslo. The details of the car­nage he cre­at­ed are avail­able on Wikipedia and won’t be repeat­ed here. His attack shocked the con­sciences of the world, giv­en its bru­tal­i­ty and the tar­get­ed nature of the vio­lence against chil­dren. But what also shocked the world was why he decid­ed to go on a mass mur­der­ing ram­page in the first place: race.

Brievik was pulling from a grow­ing tra­di­tion in the Euro­pean far right: a fear of “Islamiza­tion” dri­ven by immi­gra­tion poli­cies. Espe­cial­ly in the years since, the creep­ing white suprema­cism in Europe, the U.S., and Aus­tralia and New Zealand have focused less on domes­tic minori­ties and more on immi­grants. When many white suprema­cist fig­ures make polit­i­cal state­ments, espe­cial­ly about race or class, behind those state­ments you will ulti­mate­ly find a loathing of dark-skinned peo­ple “out­siders” they sim­ply can­not accept as a part of “their” soci­ety. This is how even the pres­i­dent can tell non-white U.S. con­gress­women, includ­ing those who were born here, to “go back where you came from.” It is a state­ment built from the belief that white­ness is Amer­i­can­ness, so there­fore non-white­ness (even if you were born here) is non-Amer­i­can.

This sort of rhetoric is a mor­tal threat to the idea of a mul­ti­cul­tur­al law-based democ­ra­cy, one based on civic insti­tu­tions instead of race.

When Brievik mur­dered those dozens of chil­dren, how­ev­er, I down­played this threat such belief sys­tems post. In an arti­cle for The Atlantic, I argued that focus­ing on Brievik’s 1,500 page man­i­festo was hypocrisy due to a num­ber of weird polit­i­cal tri­an­gu­la­tion deci­sions that feel unfa­mil­iar to me now (in short: I was work­ing at a think tank that demand­ed stu­dious non-par­ti­san­ship in our pub­lic writ­ing, even when it didn’t make any sense to do so, and I regret let­ting myself be cap­tured by those pres­sures).

I was wrong to do this. While I do think we still don’t under­stand the pre­cise mech­a­nism by which some­one shifts from believ­ing abhor­rent ideas to act­ing on them, there is copi­ous research demon­strat­ing that abhor­rent beliefs do lead to increas­es in eth­nic vio­lence. If a belief sys­tem is encour­ag­ing of vio­lence and dehu­man­iza­tion then it has to be con­sid­ered along­side the vio­lent actors who say it inspires them. Both on first prin­ci­ples, and on the mer­its I mus­tered to make my case, I was wrong to say Brievik’s belief sys­tem was imma­te­r­i­al to his deci­sion to com­mit mass mur­der.

I repu­di­ate my 2011 arti­cle.

Instead, I think we need to take a few moments to under­stand how, as the debate over hate speech is manip­u­lat­ed in pro­found­ly bad faith by right wing pub­lic intel­lec­tu­als, the pro­lif­er­a­tion of hate speech is hav­ing a mea­sur­ably bad effect on us as a soci­ety. And, real­iz­ing that, I’ll also dis­cuss why plac­ing faith in inter­net com­pa­nies to fix the prob­lem absolves every­one else of the need to act.

Why Now

The recent mass shoot­ing in El Paso, TX was some­thing of a water­shed. It wasn’t the first time a white man inspired by a big­ot­ed belief sys­tem and egged on by a vio­lence-thirsty media ecosys­tem was dri­ven to com­mit mass vio­lence, but the scale on which it hap­pened — and the inten­tion with which he tar­get­ed the Lati­no com­mu­ni­ty — was hor­ri­fy­ing. In short, a man quite lit­er­al­ly dri­ven mad by the racist hate-speech direct­ed toward Lati­no Amer­i­cans in right wing media drove hours to a Wal­mart with the express pur­pose of mur­der­ing peo­ple of Lati­no descent. He even post­ed online a man­i­festo detail­ing as much.

I recent­ly explored why and how racist hate speech from the White House can have a pro­found effect on our soci­ety by alter­ing the terms and the red­lines through which peo­ple tend to view issues. In oth­er words, the White House can alter social atti­tudes for bet­ter or worse, and this White House is doing exact­ly that by alter­ing social atti­tudes for the worse.

While hate speech from non-offi­cial sources is usu­al­ly dis­count­ed as a causal fac­tor in com­mit­ting vio­lence (as I argued for Brievik), there is, in fact, a lot of research that sug­gests it can be a pow­er­ful fac­tor in oper­a­tional­iz­ing the atti­tudes being dri­ven by lead­ers.

Mass media plays a pow­er­ful and obvi­ous role here. In a 2010 exper­i­ment, polit­i­cal com­mu­ni­ca­tions pro­fes­sor Nathan Kalmoe demon­strat­ed a strong link between “mild vio­lent metaphors” like the phrase “fight­ing for our future” and increased sup­port for real world polit­i­cal vio­lence. While this effect was lim­it­ed to peo­ple with aggres­sive per­son­al­i­ty traits — those who are not inclined to vio­lence are unaf­fect­ed by the mes­sages — there nev­er­the­less is a strong link­age between vio­lent speech and vio­lent acts. One leads to the oth­er in a con­sis­tent way. Fur­ther, there is strong evi­dence that Pres­i­dent Trump’s big­ot­ed state­ments about Mex­i­cans caus­es all peo­ple to say more offen­sive things about not just Mex­i­cans, but against all oth­er eth­nic groups.

Social sci­en­tists who study this have come to con­clu­sions every­one should find trou­bling. Research shows that when entire peo­ple groups are demo­nized by soci­ety elites, whether in the media or in gov­ern­ment, the process of “moral dis­en­gage­ment” makes it eas­i­er to com­mit acts of vio­lence with­out affect­ing your own self-image. In oth­er words, because the “oth­er” is bad, doing bad things to them is good. It is how peo­ple can get whipped into a vio­lent fren­zy by mali­cious speech: They believe they aren’t com­mit­ting vio­lence against a per­son who deserves pro­tec­tion, but rather against a per­son who deserves vio­lence.

Dur­ing the Rwan­dan geno­cide in 1994, a media out­let called Radio Télévi­sion des Mille Collines, or RTLM, spewed out a steady stream of dehu­man­iz­ing rhetoric direct­ed at the Tut­si eth­nic group. The onslaught of pro­pa­gan­da, which includ­ed encour­ag­ing vio­lence against Tut­sis, was so per­va­sive, and so effec­tive at spin­ning up peo­ple to want to exter­mi­nate their neigh­bors, that it was nick­named “Radio Machete.” RTLM did not spring up overnight. It took years of con­tin­ued dehu­man­iza­tion to prime Rwan­da for mass vio­lence.

At the sen­tenc­ing of RTLM founder Fer­di­nand Nahi­mana to life in prison at the UN’s Inter­na­tion­al Crim­i­nal Tri­bunal, pre­sid­ing judge Navanethem Pil­lay said, “You were ful­ly aware of the pow­er of words, and you used the radio – the medi­um of com­mu­ni­ca­tion with the widest pub­lic reach – to dis­sem­i­nate hatred and violence….Without a firearm, machete or any phys­i­cal weapon, you caused the death of thouds of inno­cent civil­ians.” It was a clear con­nec­tion between the spread of hate speech and the vio­lent actions under­tak­en by those who heard it.

The fact is that dehu­man­iz­ing lan­guage, repeat­ed to a large audi­ence, works. Refer­ring to humans in dehu­man­iz­ing ways like “ver­min,” and describ­ing a com­mu­ni­ty of peo­ple as an “inva­sion” are a well-trod path toward com­mit­ting mass mur­der. And in the case of the El Paso mass mur­der­er, the lan­guage he used in his man­i­festo is dis­turbing­ly sim­i­lar to the dehu­man­iz­ing rhetoric con­ser­v­a­tive media have deployed to talk about migrants of Lati­no ori­gin.

And as more vio­lence is com­mit­ted, the prob­lem of “con­ta­gion” rears its head. The Amer­i­can Medi­al Asso­ci­a­tion sug­gests treat­ing vio­lence as “an epi­dem­ic health prob­lem,” because it “exhibits the pop­u­la­tion and indi­vid­ual char­ac­ter­is­tics of con­ta­gious epidemics—clustering, geo-tem­po­ral spread­ing, and per­son-to-per­son trans­mis­sion.” Just since the El Paso shoot­ings alone, dozens more peo­ple have been arrest­ed try­ing to com­mit mass vio­lence in Amer­i­ca. The dis­ease is spread­ing.

How­ev­er, while the ways in which mass media rhetoric can con­tribute to vio­lence are gen­er­al­ly known, there is less research about how social media is affect­ing the oper­a­tional­iza­tion of hate speech. Some ear­ly research sug­gests there is a cor­re­la­tion between hate speech online and real world vio­lence, but to affir­ma­tive­ly estab­lish a causal link there needs to be more research.

Peo­ple seem to have a vague sense that some­thing is “wrong,” which has led the UN Sec­re­tary-Gen­er­al António Guter­res to con­demn the role social media has played in spread­ing hate speech ear­li­er this year. Thus, there is an oppor­tu­ni­ty to try to under­stand how and why the “churn” of dehu­man­iz­ing, racist, and oth­er big­ot­ed speech pro­lif­er­at­ing online has come to be such a press­ing con­cern.

Understanding Hate Speech on Social Media

Social media com­pa­nies are in a dif­fi­cult posi­tion. Their dom­i­nance of the online space gives them the pow­er to shape the dis­course by enabling or dis­abling cer­tain types of dis­cus­sion — pow­er nor­mal­ly wield­ed by gov­ern­ments. How­ev­er, none of them were designed as prod­ucts to be a replace­ment for the “pub­lic square” in civic life. (Most of them are real­ly plat­forms for research­ing con­sumer behav­ior to bet­ter sell adver­tise­ments.) As a result, they are prof­it-seek­ing com­pa­nies where a seg­ment of their cus­tomer base is demand­ing they pro­vide a plat­form for speech as a pub­lic good. It is an inher­ent con­tra­dic­tion: as com­pa­nies, they aren’t bound the same legal require­ments to allow all (or most) speech types as a gov­ern­ment would be, yet it is still some­thing their users are demand­ing.

The pres­sure to ban cer­tain types of con­tent is grow­ing, and while it is unre­al­is­tic to expect social media com­pa­nies to elim­i­nate all prob­lem­at­ic speech, it is worth not­ing where and how efforts to lim­it its reach and impact are work­ing. In some cas­es, like Twitter’s recent deci­sion to ban state-run media, it is pos­si­ble to nar­row­ly tar­get poten­tial­ly mali­cious accounts in a broad fash­ion; on an indi­vid­ual basis, how­ev­er, it is fiendish­ly dif­fi­cult, and has no clear answer for mov­ing for­ward.

Every main­stream social media ser­vice has a pol­i­cy that for­bids hate speech. But how that hate speech pol­i­cy is cod­i­fied and enforced is con­stant­ly up for nego­ti­a­tion — in oth­er words, hate speech on social media is a con­test­ed space with­out clear guardrails or end states. As a result, it is almost impos­si­ble for them to make a deci­sion that a broad con­sen­sus of users will regard as fair and appro­pri­ate.

No social media plat­form is inten­tion­al­ly host­ing hate speech, but the chal­lenges in address­ing it in a sys­temic way have led to frus­tra­tion and accu­sa­tions of com­plic­i­ty in vio­lence. These accu­sa­tions have some mer­it, like when Face­book famous­ly admit­ted it had a role to play in an attempt­ed geno­cide in Myan­mar because it did not remove gov­ern­ment-dis­sem­i­nat­ed hate speech from its plat­form. And that’s because of how hate speech behaves as a pathol­o­gy.

At a fun­da­men­tal lev­el, hate speech is just as con­ta­gious as social vio­lence. In a 2017 study pub­lished by the Nation­al Acad­e­my of Sci­ences, a team of researchers found that hate­ful behav­ior, includ­ing hate speech, by a major­i­ty eth­nic group at a minor­i­ty eth­nic group is more con­ta­gious than hate­ful behav­ior direct­ed at peo­ple of the same eth­nic­i­ty. Every time an epi­thet was spo­ken, or a slight phys­i­cal act to demean the minor­i­ty (like a shove) was under­tak­en, it became much more like­ly to hap­pen again — and far more like­ly that sim­i­lar behav­ior direct­ed between co-eth­nics. This, the researchers sug­gest, “may help to explain why eth­nic hos­til­i­ties can spread quick­ly (even in soci­eties with few vis­i­ble signs of intereth­nic hatred) and why many coun­tries have adopt­ed hate crime laws.” In oth­er words, hate speech begets more hate speech, which even­tu­al­ly begets vio­lence.

While this sounds bad, the upside to hate speech exist­ing on social media plat­forms is that it can be mon­i­tored and coun­tered. Just recent­ly a man who post­ed to Face­book about his love of Hitler, Nazis, and Don­ald Trump (sigh) was arrest­ed when the FBI deter­mined that his threats to mur­der large num­bers of Lati­nos was cred­i­ble. He didn’t devel­op a tar­get­ed hatred of Lati­nos in a vac­u­um — he con­sumed media that encour­aged a tar­get­ed hatred of Lati­nos. Even so, if he hadn’t been post­ing those mes­sages to Face­book, then law enforce­ment might not have been able to catch him before he com­mit­ted vio­lence.

Counter speech is eas­i­er to con­duct on social media as well. And indeed, that is Facebook’s explic­it strat­e­gy for han­dling hate speech. As a crowd-sourced, seem­ing­ly spon­ta­neous response to hatred, it is appeal­ing on a num­ber of fronts: it doesn’t place the onus of respon­si­bil­i­ty (or the chal­lenge of walk­ing an extreme­ly nar­row path to sat­is­fy most users) on the com­pa­ny, it is the sort of free­wheel­ing com­mu­nal response many ide­al­ists pre­fer, and it doesn’t have to be direct­ed or con­strained. But coun­ter­ing speech is not a straight­for­ward process — as Demos, a think tank which con­duct­ed counter speech research on behalf of Face­book, dis­cov­ered, there is nuance to a counter speech con­tent strat­e­gy that a spon­ta­neous crowd source can­not deploy, espe­cial­ly when the sources of hate speech are engaged in a tar­get­ed and inten­tion­al cam­paign to spread their mes­sage. Because hate speech is often coor­di­nat­ed (for exam­ple, by a far right par­ty to encour­age dis­crim­i­na­tion against minori­ties), the counter to that hate speech also needs to be coor­di­nat­ed and tar­get­ed in order to be effec­tive.

Social media com­pa­nies don’t have to have their own proac­tive respons­es to hate speech on their plat­forms, but these line of research does sug­gest that those who wish to counter hate speech need to be bet­ter orga­nized. The asym­me­try in respond­ing to hate speech on social plat­forms is a chal­lenge that can’t fall on the social media com­pa­nies alone. Sim­ply put, the mali­cious actors are orga­nized, while those seek­ing to stop them are not.

Last­ly, the def­i­n­i­tion­al prob­lem of hate speech is rarely addressed by crit­ics of social media com­pa­nies. Where gov­ern­ments have defined the bound­aries of hate speech, social media com­pa­nies are pret­ty good at com­ply­ing with those bound­aries espe­cial­ly when they’re giv­en the force of law. How­ev­er, it isn’t Google’s or Facebook’s or Twitter’s job to define exact­ly what hate speech is, espe­cial­ly in Amer­i­ca where being a big­ot­ed jerk to peo­ple is pro­tect­ed by the con­sti­tu­tion. That being said, the com­pa­nies do make an effort to exclude hate speech and have offi­cial poli­cies say­ing as much (YouTube is an excep­tion where they seem to have carved out allow hate speech if some broad­er polit­i­cal argu­ment is made as well). The space cre­at­ed by the offi­cial poli­cies and the action tak­en by lead­er­ship at the com­pa­nies is a major fac­tor in why many peo­ple place the onus of enforce­ment on the com­pa­nies them­selves.

This is a hard prob­lem. You can­not exist as a main­stream inter­net com­pa­ny in 2019 and have no stance on hate speech, but actu­al­ly defin­ing the exact bound­aries of that speech and then enforc­ing it is incred­i­bly dif­fi­cult, one with­out a straight­for­ward tech­no­log­i­cal solu­tion.

What Happens Now?

So to come back around to Brievik: the net­works in which peo­ple are being rad­i­cal­ized by the far right are matur­ing into a glob­al machine. Their adher­ents are oper­a­tional­iz­ing their speech in a way they sim­ply weren’t eight years ago when Brievik com­mit­ted his mas­sacre.

I men­tioned the role Face­book played in the eth­nic vio­lence in Myan­mar above. What­sApp, one of their sub­sidiary com­pa­nies, has also been impli­cat­ed in mass hys­te­ria and vio­lence in India. There is con­clu­sive evi­dence that when hate speech is left unchal­lenged on the inter­net it even­tu­al­ly leads to vio­lence. In 2011, it was unclear that hate speech online real­ly mat­tered all that much — did it real­ly mat­ter that Brievik was sub­sumed in a com­mu­ni­ty that engaged in big­ot­ed speech against Mus­lims and non-white peo­ple? If so, then why did he mur­der Nor­we­gian chil­dren, instead of the tar­gets of his hate speech? But now, in 2019, there is much more research estab­lish­ing a link between the engage­ment with hate speech and the deci­sion to com­mit vio­lence. It is a strong cor­re­la­tion, whether you’re in the U.S., or India, or Myan­mar, or Nige­ria, or Turkey, or Europe.

Unfor­tu­nate­ly, there isn’t an easy solu­tion to this chal­lenge — it is a con­test­ed space. Pro­tect­ed speech does not have clear bound­aries, and the def­i­n­i­tion of hate speech varies from coun­try to coun­try even as that lan­guage cross­es bor­ders. The oblig­a­tion of gov­ern­ments, com­pa­nies, and indi­vid­ual users to respond to it is also unclear. No one wants to see hate speech define the expe­ri­ence of being on the inter­net, but it remains unclear what the most effec­tive and rights-pro­tec­tive way is of achiev­ing that state.

One place we can begin to rethink how to address hate speech is by acknowl­edg­ing this com­plex­i­ty. “It’s com­pli­cat­ed” is not the most pop­u­lar response to a prob­lem on the inter­net, but ignor­ing how dif­fi­cult this chal­lenge is will lead to bad solu­tions that might make things worse. If this were easy, it would have been solved by now.

I float­ed an idea my piece about how and why the President’s rhetoric on immi­gra­tion has real world con­se­quences: the respon­si­bil­i­ty to speak up. Counter speech to iden­ti­fy and respond to hate speech, can be effec­tive in blunt­ing its impact. Of course, there will always be a con­test of that counter speech (con­sid­er Fox News host Tuck­er Carl­son assert­ing that white suprema­cy is a hoax the week after a white suprema­cist mur­dered 22 peo­ple for being non-white). But that is the nature of con­test­ed spaces — they are con­test­ed. It means that unless there is a response and push­back against the hate speech, that hate speech will dom­i­nate the plat­form. We have a respon­si­bil­i­ty to speak up against hate speech.

One thing in Amer­i­can soci­ety that can help on this front is that no one wants to be acknowl­edged as a racist. Peo­ple in fact hate it, even those who are plain­ly spo­ken racists who freely use hate speech. There is absolute­ly a dan­ger in using the term too loose­ly, but it remains a pow­er­ful one to deploy. A direct con­fronta­tion with some­one over their hate speech might not be con­struc­tive should they be called racist right out of the gate. But there is a con­struc­tive (though emo­tion­al­ly labo­ri­ous) path toward address­ing big­ot­ed speech through dia­logue that stead­fast­ly chal­lenges racist beliefs.

This is an approach that takes time and com­mit­ment. One of the rea­sons why hate speech has pro­lif­er­at­ed online is that the major­i­ty of peo­ple who do not like it have not con­front­ed it. While it’s easy to com­plain about how a plat­form is a “cesspool,” there hasn’t been an estab­lished norm that big­otry is unac­cept­able. The silence of assum­ing it isn’t your prob­lem is a form of com­plic­i­ty, but it’s one that can be reversed by sim­ply say­ing “no, this is not accept­able.”

Putting the onus on us, the users, to police the space we use online is unsat­is­fy­ing. Yet, the alter­na­tive is even worse. Com­pa­nies are by nature con­ser­v­a­tive — they are slow to act, and tend to do so in a min­i­mal way when rev­enue is on the line. It is unre­al­is­tic to expect them to solve this prob­lem on their own — we have to take respon­si­bil­i­ty for the sort of lan­guage we will tol­er­ate, whether online or in more tra­di­tion­al media. It won’t be easy, and there will be extreme­ly disin­gen­u­ous push­back against it. But unless we col­lec­tive­ly decide to act, the racists will win.5

Joshua Foust used to be a foreign policy maven. Now he helps organizations communicate strategically and build audiences.