Photo: AP Photo/J. Scott Applewhite
Twitter has done an excellent job keeping ISIS off of Twitter. Amarnath Amarasigam, an extremism researcher, stated that he hasn’t “seen a legit ISIS supporter on Twitter who lasts longer than 15 seconds in two-and-a-half years.” They’ve been able to ban radicalizing posts full of hate speech and calls for violence from the group. So why haven’t they done the same for a far more common domestic threat in the United States, the terrorism and attacks from white supremacists and Neo Nazis, the so-called “Alt-Right?” Now we have a clue from inside Twitter. According to engineers familiar with Twitter’s AI based filtering algorithms, it would involve banning Republican politicians.
Now, I want you to take a step back and think about that one for a second. Slow your reading for just a moment.
Okay?
Twitter can’t ban Neo Nazis because they’d ban Republicans. The groups, to an AI algorithm, overlap. Basically, Twitter’s (unofficial) argument for not banning Nazis is that the Nazis are already too powerful, politically.
It’s the same coddling treatment of Nazis that lead to World War 2.
"We couldn't ban Nazis because then we'd have to ban Republicans"
is going to be on this country's gravestone
— L O L G O P (@LOLGOP) May 3, 2019
A report from a Twitter employee explains how they were able to ban ISIS, and why they won’t do the same for white nationalists and other far-right extremists.
In This Article:
Why Twitter Would Flag Republicans
U.S. Nationalists at the Unite the Right Rally. “Some very fine people,” according to Donald Trump. Photo: Edu Bayer/New York Times
According to J.M. Berger, author of Extremism and researcher into extremists, ISIS, and other far-right groups, many avowed white nationalists are avid fans of Donald Trump and other Republicans. AI, much like the human mind, works by finding patterns. It does this by taking assigned models and stretching the definition to find associations. You may already be able to see where this is going.
Flagging Content Without Looking at Content
Let’s look at a non-political example for a second. An AI algorithm aimed at finding the most attractive people on Instagram may not have to “look” at her photos at all. First, it would see who finds these people attractive. It would notice that some men are found attractive by predominantly women and gay men, and women are attractive to men and gay women. From there, a researcher could conclude that you don’t actually have to create an algorithm that can find attractiveness in a human. Instead, you just have to look at who finds them attractive. An Instagram star with a lot of male and lesbian followers is, most likely, an attractive woman.
Now back to politics. We can easily create a list of words and phrases used by white supremacists to create a list of outward nationalists. From there, we can look at the people that have the most white supremacist followers. Even if those people never say anything racist, homophobic, or sexist on Twitter, they may do so at rallies or Klan meetings. As such, their large following of white supremacists would be the only clue that the person is a white supremacist or other form of bigot. The algorithm could then flag them as a highly likely white supremacist.
Republican or White Nationalist? For AI, it’s Complicated
“White nationalists, in contrast, have inconsistent branding, diffuse social networks and a large body of sympathetic people in the population, so the risk of collateral damage might be perceived as being higher, but it really depends on where the company draws its lines around content.”
– J.M. Berger, Extremism Researcher
This is how AI would go about flagging white supremacists. Unfortunately, even if he never tweeted, Donald Trump would be flagged as a potential white supremacist. However, as we all know, Donald Trump tweets a lot. He has tweeted about banning all Muslims, enough for Twitter’s algorithm to ban him without a single human intervening.
This is, in the most simplistic of explanations, how machine learning algorithms would tag white supremacists for banning or tweet deletion. Of course, in practice, they’re far more complicated, and would use more natural language processing, location data, tweet history, and even information available outside of Twitter, elsewhere on the web. Unfortunately, if Twitter casts a wide enough net to be effective against white nationalist leaders, it would also capture Republicans in that net, because their ideas closely align with white supremacists and they have a large alt-right following.
Background: Powerful Hate
An example of a fringe and hateful conservative position. Hate speech on a Facebook post. Facebook allowed it to keep from upsetting conservatives, fueling anti-Muslim violence.
There’s going to be controversy here that we didn’t see with ISIS, because there are more white nationalists than there are ISIS supporters, and white nationalists are closer to the levers of political power in the US and Europe than ISIS ever was.
-J.M. Berger, Extremism Researcher
A political leader talked about banning all people of a particular religion from entering the country. No, this wasn’t a Nazi talking about Jewish people in the 1930’s. It was Donald Trump in 2016 on Facebook. Without context, it’s impossible to differentiate the language of Nazis in the 30’s from Trump’s Islamophobic proclamation. He tweeted the same opinion on Twitter.
To an AI algorithm without context, he’s a psycho talking about banning people of a particular religion. It would flag him as an alt-right extremist. It doesn’t have the context that he’s not just an extremist, he’s also the president of the United States and therefore, apparently, immune to such comparisons.
Where We Stand
It’s 2019. Republicans still largely control the government. In 2018, Democrats were able to take back control of the House, but only just. Years of gerrymandering under Republican control have made the party the default. To take over the House took the effort of the “Blue Wave,” a massive outpouring of support in response to the corruption, hatred, violence, bigotry, and disregard for the planet from the Republican party. It was a massive effort, and Democrats just barely took control of the House. They lost ground in the Senate.
The Senate is where bills go to die now. Bills like the Equality Act, which protects LGBTQ people from discrimination in the same way anti-discrimination laws have protected people of any race, religion, or gender, will die in the Senate, despite majority support from the American people. Most people do not believe it should be legal to fire LGBTQ people or kick them out of their homes for the way they were born, however, this is perfectly legal in most U.S. states. Shocking to my straight readers, I’m sure.
However, the Republican party does not rule by popular vote. They maintain control through their gerrymandered districts, corrupt ties to lobbying groups and campaign contributions, and the Electoral College, which disproportionately favors them. In fact, Donald Trump lost the election by over 3 million votes, but still won thanks to this outdated and corrupted system.
The Power They Wield
2018 Senate Election Results via New York Times
Because Republicans do not need to care about the will of their constituents, and because they have control of the Senate, Executive Branch (president), and now, thanks to a record number of judicial appointments, the courts, Republicans are almost as close to all powerful as the constitution allows. No bill passed through the House with mostly Democrat support has a hope of passing through the Senate. The will of the people is just about meaningless.
Well, almost. If they do anything so heinous that even Republicans are disgusted, they could lose control. For example, multiple women and even police reports pegged Roy Moore as a sexual abuser and child predator. He lost his election by an incredibly small margin, but he was ousted regardless.
Why Republicans Want to Control Twitter
Therefore, when Twitter lumps Republicans in with white supremacists, Neo Nazis, and the alt-right (for doing everything they want them to do like fighting LGBTQ rights, women’s rights, and restricting voting rights for minorities), Republicans get upset. Not all Republicans like the idea of voting for actual Klan members, supporters of domestic terrorism, bigots, and racists. However, if no one raises the issue to a prominent enough level, voters can continue to support the party in blissful ignorance.
Twitter banning a large number of Republicans for spewing nationalistic hate would be the breaking point for some Republican voters. Therefore, Republicans threaten Twitter and Facebook every time the issue comes up. They have the power to legislate Twitter and Facebook out of existence, so the companies play along.
Basically, the Nazis are too powerful in the United States to overthrow through free speech. Twitter cannot fight back against the Nazis, Klan members, alt-right, and white nationalists because they’re supported by Republicans.
Donald Trump himself has called himself a nationalist, though he dropped the “white” part that normally goes with it. The meaning, of course, is mostly unchanged.
Illegal Discrimination, But Twitter Permits It
The United States has freedom of religion. It’s a nation that has banned religious discrimination. However, Republican politicians have advocated Muslim bans, a direct and hateful violation of the U.S. Constitution.
Facebook thought about banning him. It’s the same language that lead to the second world war and the genocide of the Jewish people. Surely they shouldn’t allow it on their platform. It was this kind of hate that, most recently, lead to the genocide of the Rohingya Muslims in Myanmar.
But a conservative Facebook employee and former George W. Bush staffer said it would upset too many Republicans, who hold the same discriminatory, hateful views. Facebook decided to allow Donald Trump to stay.
Now, thanks to an employee at Twitter, we’ve learned the same discussion has happened within their walls, with the same potentially disastrous results.
Twitter Fears the Hateful
Photo: REUTERS/Jim Young – RTS97MX
Twitter’s efforts to ban ISIS were incredibly successful. However, posts from people who do not support ISIS are occasionally removed. If Twitter took such extreme action against white nationalists, even Republicans who aren’t extreme outward defenders of white nationalism would likely be banned. Of course, small measures like refusing to vote for the equality act, something every Republicans in the House did, or voting to require state issued IDs to vote specifically because it would affect black Americans are extremely racist, but pass off as “legitimate” political beliefs.
I guess what I’m saying is, if you’re “accidentally” banned by Twitter’s anti-Nazi AI, you’re probably supporting Nazis a little bit. Not every Republican is a white nationalist, but they sure do vote like them.
However, as the party commands a large amount of power, Twitter can’t upset them. Even if they wanted to ban all the white nationalists on Twitter, they couldn’t do so through AI, and they couldn’t ban any politicians for fear of legislative revenge.
You don’t need a history book to learn how extremists and authoritarians take control of a government and squash free speech, it’s right there in front of us.
No representative from Twitter has officially stated this is why they haven’t banned the Nazis, as Twitter users often ask of the CEO, Jack Dorsey. However, between the anonymous reporting from inside Twitter, external machine learning experts, and my own analysis as a software engineer familiar with machine learning, it is the most likely and logical conclusion.
Twitter won’t ban Nazis and white nationalists because they’d end up banning conservative politicians for espousing the same views as avowed Neo Nazis.
How the Far-Right and Republicans Kill Free Speech
Right now, you might be wondering how banning certain speech actually increases free speech. It’s something I always get in the comments on Facebook when I share these posts. So, let me explain the two ways the Republican position here and those of the Alt-Right actually harm free speech.
Government Control of Private Businesses
First, the obvious. Twitter wants to do something with their private company and, because of threat from legislators, cannot. You can’t ban a person from your business because of who they are as a person, but you can ban them for their actions. So, if a person is drumming up violence in your store, you have the right to kick them out of your store. A person marching down the isles of your local supermarket could be kicked out for shouting about killing all the employees.
Obviously this is different than kicking someone out for their skin color. No one can change their skin color, gender, or sexuality. A person spewing hate speech or displaying hateful images, however, can stop.
On Twitter, they have the right to kick people off for spewing hate speech, which increases violence, leads to hate crimes, and can even lead to genocide. However, they know that if they take extreme action against these people, they’ll have to fight Republicans on it.
Basically put, the “small government” Republicans want to control a private business so they don’t make sweeping bans of hate speech. In doing so, they’re limiting the stances Twitter can take, effectively blocking the free speech of Twitter’s executives.
Limiting Speech out of Fear
Just as Twitter can’t do what they want out of fear of Republican reactions, marginalized groups can’t use Twitter the way they want to out of fear of the extremists Republicans force us to deal with. A person won’t want to speak out about the issues people of color, immigrants, LGBTQ people, or women face due to the hate they’ll receive from the alt-right. In some cases, like those during the height of Gamergate, women were forced out of their homes after their actual addresses were posted online and they were threatened.
Because of this very real threat of harassment and violence, women, people of color, and LGBTQ people simply won’t use the platforms. They’ll have no way to speak their minds. By allowing hate speech and harassment, you choose to instead censor non-hateful and non-violent voices.
What Twitter Should Do
“As a company, Twitter won’t say that it can’t treat white supremacy in the same way as it treated ISIS.”
– Joseph Cox and Jason Koebler, Motherboard/Vice
Here’s the part where I have a magical solution that makes everyone happy. Just kidding! Twitter has to pick a side. Right now, by not actively censoring white nationalists, misogynists, and Neo Nazis, they’ve chosen to instead censor the victims of these groups. They need to take greater action against them, using machine learning to place preliminary bans and having human curators for more questionable bans or tweet deletions. Machine learning algorithms can have a confidence level, and, therefore, items or people flagged with a lower confidence level, most likely politicians or non-racist views, would be curated by humans. Only the ones that are absolutely hate speech would be instantly removed by the AI moderation.
This will sometimes get legitimate posts. Twitter can rest easy on these. If someone complains about a deleted tweet, they can, if it’s a low enough confidence level, look into it manually. If it’s actually not hate speech or harassment, they can apologize, add it back, and even offer to show the post at the top of their followers’ feeds. I believe this would allow Twitter to appease conservatives as best as they can while still banning hate speech.
The problem, again, is the risk that 1) Republicans will still legislate Twitter for this, which is highly likely, and 2) the cost of manual review will be too high, resulting in many upset users.
Twitter has to decide between upsetting users and helping white nationalists gain more power. They’ve chosen Nazis and history shows that decision will lead to horrendous results.
Sources:
- Jane Coaston, Vox
- Joseph Cox and Jason Koebler, Motherboard/Vice
- HRC, HRC State Maps
- Vivian Kane, The Mary Sue
- Casey Newton, The Verge
P.S.
Republicans, if you’re really worried that you could be banned for white nationalism or alt-right extremism, maybe the filter for white nationalism isn’t the problem. Maybe your racist, sexism, and homophobic policies and statements are the real problems.