‘Algospeak’ is changing our language in real time

As the pandemic pushes more people to communicate and express themselves online, algorithmic content moderation systems have had an unprecedented impact on the words we choose, especially on TikTok, and given rise to a new form of internet-based Aesop.

Unlike other major social platforms, the main way content is distributed on TikTok is through an algorithmically curated “For You” page; having followers does not guarantee that people will see your content. This change has led ordinary users to adapt their videos mainly to the algorithm rather than watching them, which means that it is more important than ever to follow the rules of content moderation.

When the pandemic started, people on TikTok and other apps called it “Backstreet Boys reunion tour” or calling it “panini” or “panda express,” platforms like low-level videos tagged in the name of the pandemic to combat misinformation. When young people began to discuss coping with mental health, they spoke of “being lifeless” to have open conversations about suicide with algorithmic impunity. Sex workers, long censored by moderation systems, call themselves “accountants” on TikTok and use corn emojis instead of the word “porno.”

As discussions of major events filter through algorithmic content delivery systems, more users are bending their tongues. Recently, when discussing the invasion of Ukraine, people on YouTube and TikTok used the sunflower emoji to represent the country. Users will say “blink in lio” for “link in bio” while encouraging fans to follow elsewhere.

Euphemisms are especially common in radicalized or harmful communities. The pro-anorexic eating disorder community has long adopted variations on moderation wording to avoid restrictions. A paper from the Georgia Institute of Technology’s School of Interactive Computing found that the complexity of such options has increased over time. Last year, anti-vaccination groups on Facebook began changing their names to “dance party” or “dinner party,” and anti-vaccination influencers on Instagram used similar code words, calling vaccinated people “swimmers.”

Adapting language to avoid surveillance predates the Internet. While people living in repressive regimes developed code words to discuss taboo subjects, many religions refrained from invoking the devil’s name.

Early Internet users used alternative spellings, or “leetspeak,” to bypass word filters in chat rooms, picture boards, online games, and forums. But algorithmic content moderation systems are more common on today’s Internet, often silencing marginalized communities and important debates.

During YouTube’s “adpocalypse” in 2017, when advertisers pulled their money from the platform for fear of dangerous content, LGBTQ creators talked about the demonetization of videos for saying the word “gay.” Some have started using the word less often or replacing others to monetize their content. More recently, TikTok users have started saying “cornucopia” rather than “homophobia” or saying they are members of the “leg booty” community to indicate they are LGBTQ.

Sean Szolek-VanValkenburgh, creator of TikTok, which has more than 1.2 million followers, said: “There’s a line we have to walk, it’s a never-ending battle to say something and try to get the message across without saying it directly.” “It disproportionately affects the LGBTQIA community and the BIPOC community because we’re the people who are creating this conversation and creating the colloquiums.”

Chats on TikTok about women’s health, pregnancy and menstrual cycles are also consistently low, said Kathryn Cross, a 23-year-old content creator and founder of Anja Health, a startup that offers cord blood banking. He replaces the words “sex”, “period” and “vagina” with other words or writes them with symbols in the titles. Many users say “nip nops” rather than “nipples”.

“I feel like I need a disclaimer because I feel like having these weirdly spelled words in your headlines makes you look unprofessional,” he said, “especially for content that’s supposed to be serious and medically oriented.”

Because online algorithms will often flag content that mentions certain words out of context, some users avoid pronouncing them at all because they have alternative meanings. Lodane Erisian, community manager for Twitch creators (Twitch considers the word “cracker” an insult) “Now, when you’re literally talking about crackers, you have to say ‘salties,'” said Lodane Erisian, community manager for Twitch creators. Twitch and other platforms have even gone so far as to remove certain emotes because people used them to convey certain words.

Black and trans users and those from other marginalized communities often use algospeak to discuss the oppression they face, substituting the words “white” or “racist”. Some are too nervous to say the word “white” at all and hold their palms up to the camera to represent white people.

“The reality is that tech companies have been using automated tools to manage content for a long time, and even though it’s presented as this sophisticated machine learning, it’s often a list of words that they find problematic,” said Angel Díaz, a teacher. UCLA School of Law studying technology and racial discrimination.

In January, Kendra Calhoun, a postdoctoral researcher in linguistic anthropology at UCLA, and Alexia Fawcett, a doctoral student in linguistics at UC Santa Barbara, gave a presentation on language on TikTok. They explained how new algospeak code words emerged by self-censoring the words in TikTok’s captions.

TikTok users are now using the phrase “le dollar bean” instead of “lesbian” because TikTok’s text-to-speech function pronounces it as “Le$bian”, which users believe will circumvent content moderation by censoring typing “lesbian”.

Evan Greer, director of digital rights nonprofit Fight for the Future, said it’s foolish to try to remove specific words from platforms.

“One, it doesn’t really work,” he said. “People who use platforms to do real damage are pretty good at figuring out how to get around these systems. Second, it leads to collateral damage of literal speech. Greer argues that trying to regulate human speech on a scale of billions of people in dozens of different languages ​​and trying to combat things like humor, sarcasm, local context and slang can’t be done simply by downgrading certain words.

“I think this is a good example of why aggressive moderation will never be a realistic solution to the harms we see from the business practices of big tech companies,” he said. “You can see how slippery that slope is. Over the years, we have increasingly witnessed a misguided demand by the general public to quickly remove more content, regardless of the value of the platforms.”

Major TikTok creators have created shared Google Docs with a list of hundreds of words that the app’s moderation systems find problematic. Other users try to reverse the system, keeping a table of terms they believe suppress certain videos.

The site, Zuck Got Me For, created by a meme account administrator employed by Ana, is a place where creators can upload raunchy content banned by Instagram’s moderation algorithms. In a manifesto about his project, he wrote: “Creative freedom is one of the only silver linings in this fiery online hell we’re all in… It’s the independent creators who suffer as the algorithms tighten.”

It also describes talking online to avoid filters. “If you have violated the terms of service, you will not be able to use profanity or negative words such as ‘hate’, ‘kill’, ‘ugly’, ‘stupid’, etc.,” he said. “I often write: “I the opposite of love Choose xyz’ instead of ‘I hate xyz’.

The Online Creators Association, a labor advocacy group, also published a list of demands demanding more transparency from TikTok about how it manages content. “People need to keep their mouths shut so they don’t offend these all-seeing, all-knowing TikTok gods,” said Cecelia Gray, creator of TikTok and co-founder of the organization.

TikTok offers an online resource center for creators who want to learn more about its recommendation systems, and has opened multiple transparency and accountability centers where guests can learn how the app’s algorithm works.

Vince Lynch, CEO of IV.AI, an artificial intelligence platform for language understanding, said that in some countries where moderation is more severe, people are creating new dialects to communicate. “It translates into actual sub-languages,” he said.

But as algospeak becomes popular and replacement words become common slang, users find themselves forced to get more and more creative in order to evade the filters. “It becomes a mole game,” said Gretchen McCulloch, author of Because the Internet and author of Because the Internet, a book about how the Internet shapes language. When the platforms started seeing people saying “seggs” instead of “sex,” for example, some users even reported that they believed the substitute words were flagged.

“We create new ways of speaking to avoid this kind of bias, and then we adopt some of these words and they become common vernacular,” says UCLA Law School’s Diaz. It all came from an attempt to resist moderation.”

This does not mean that all efforts to prevent misconduct, harassment, abuse and misinformation are futile. But Greer argues that there are key issues that need to be prioritized. “Aggressive moderation will never be a real solution to the harms we see from the business practices of big tech companies,” he said. “It’s a task for policymakers and to create better things, better tools, better protocols and better platforms.”

Finally, he added, “you can never clean the Internet.”