Chapter 1 How to Play Linguistic Whac-A-Mole Remember Whac-A-Mole? The slightly unhinged arcade game where new characters keep popping up no matter how many times you smack them with your mallet? It might seem ridiculous to use coin-operated violence as an analogy for the serious linguistic changes we''re experiencing online, but Whac-A-Mole truly is the perfect metaphor for explaining how humans react to censorship. As soon as a word is banned, we find a way around it; that is, until content moderators catch wind of the new word and ban that, too. Then another word pops up, and the cycle repeats, trapping users and platforms in a never-ending loop of new spellings and substitutions that disappear once the algorithm catches on and the mallet comes down again. The faster and better the moderation tools are, the more words will be created. This is because the underlying idea--and our desire to talk about it--remains. People have been playing this game since the early days of the internet. Frustrated with text filters on bulletin boards in the 1980s, netizens turned to "leetspeak," a hacker dialect characterized by creatively respelled words. Since terms like "suicide" were censored in some chat rooms, leetspeakers wrote out coded replacements like "5U1C1D3" in much the same way that people began using "unalive.
" If the chat-room moderators caught on, they could just change the spelling to another easily recognizable form, like "$U!C!De." The tools being used have since grown more sophisticated, but the underlying process remains the same. Rudimentary word identification scripts might have been upgraded to fancy, AI-powered algorithms, but censorship is still driving linguistic creativity. If anything, the algorithmic era has spurred more innovation, because the game is happening faster, with more guesswork involved. Due to intellectual property concerns, not much is known about how these apps actually recommend content. Influencers are often subject to the whims of unknown and unfair criteria, with only opaque community guidelines as reference points. This marks a major difference between leetspeak and algospeak: You could always immediately tell when leetspeak worked, because you would see your comment posted onto a message board. For algospeak, however, the goal is to make it onto a user''s "recommended page," and it''s much harder to tell when you''ve successfully done that.
Videos with sensitive keywords aren''t always removed outright. Sometimes, they''ll be "suppressed," or shown to fewer followers. Creators can also be "shadowbanned" without warning or notification. We--and I''m including myself here--receive very little communication from the platforms, so we''re not sure whether videos do poorly because they''re bad or because they''re being censored. Understandably, then, we''ll err on the side of caution when it comes to euphemization. Creators can sort of tell what does and doesn''t get onto the recommended page by looking at our video analytics. If I usually get most of my views from the recommended page, and then all of a sudden a video is getting views only from the "followers" feed, that''s a semi-reliable indication that some part of my video made the algorithm unhappy. Through this kind of trial and error, influencers can extrapolate a pretty good idea of what the algorithm rewards and penalizes, and it''s in this context that "unalive" was forged.
When the word "suicide" wasn''t getting views, people turned to the next best term to tell their story. While linguistic innovation like this is an exciting and normal thing, it''s reasonable to be alarmed at the way community guidelines are shaping important conversations, especially as people increasingly turn to short-form video for news or advice. Oftentimes, the mysterious rules governing social media are arbitrary or outright discriminatory. TikTok has historically been proven,1 for example, to artificially suppress videos by "ugly," old, and poor creators, because they''re not as appealing to new users. This means that it''s often difficult to include larger audiences in discussions about things like disability, age, and income inequality. Nevertheless, we have no choice but to play the game and tiptoe around community guidelines wherever we can. In 2022, the Charles Dickens Museum began a desperate social media campaign to get itself un-shadowbanned from TikTok. Whenever users would search for the museum''s account, nothing would show up.
Instead, they would be cautioned2 that "this phrase may be associated with behavior or content that violates our guidelines" and that "promoting a safe and positive experience is TikTok''s top priority." Of course, the Charles Dickens Museum wasn''t doing anything wrong. They were mostly posting house tours or excerpts of old letters. Instead, the problem was with the TikTok algorithm, which was flagging the museum''s videos as obscene because they included the keyword "dick." For all its fancy high-tech machine learning, the algorithm had fallen victim to a classic internet pitfall: the Scunthorpe problem, named for an English village where residents discovered they were unable to create AOL accounts because their hometown contained the word "cunt." Following an intense #FreeDickens campaign on Twitter, TikTok eventually agreed to unblock Charles Dickens-related search terms. However, it still remains difficult to curse--intentionally or accidentally--on any platform. While your videos won''t get removed outright, they''ll often be suppressed in search just like the museum''s.
Especially if you''re cursing too much or too severely, your content will be hampered by the algorithm. If a video is eligible for the recommended page but still contains mature language, TikTok and Instagram will prevent it from appearing in clusters of similar videos in a user''s feed, which means that it''ll be pushed to a smaller audience than its work-safe competitors.3 Likewise, on YouTube, creators posting videos with severe or repeated profanity have to contend with "demonetization," where they lose the ability to earn revenue from advertisements on their content. This is especially scary to influencers whose livelihoods depend on a steady income stream from their videos. All this means that we have a lot of reasons to respell our swear words creatively online. If you look up the keyword "bitch" through TikTok''s search function, for example, you''ll likely encounter variations such as "btch," "b!tch," and "b*tch" in the video captions. Same with "fuck," which will probably give you "fck," "fvck," and "f*ck"--none too different from the leetspeak letter substitutions of the past. The practice of respelling offensive words is a centuries-old tradition known as bowdlerization, named for the Englishman Thomas Bowdler, who is mainly remembered for publishing some egregiously family-safe edits of William Shakespeare''s plays.
Self-bowdlerizing to avoid media constraints is not new: People have been doing it since at least the days of early newspaper comics, where sequences of graphic characters called grawlixes are still used instead of swear words to circumvent stringent syndication standards. The earliest known example is from this 1901 Lady Bountiful comic by Gene Carr: Over time, some grawlixes got less thinly veiled as cultural norms against profanity loosened up. Cartoonists began drawing on symbols with a visual similarity to the letters they were replacing: "@##" and "$#*!" are now industry standards for the words "ass" and "shit." Fast-forward to today, and influencers are re-creating this process in social media captions. Words like "fvck," "b!tch," and "@ss" are born out of the same motivations, drawing on the same cultural tradition of bowdlerization. Until the twentieth century, the preferred method of bowdlerization across all media formats was the double em dash (----), typically replacing the entirety of a word save for a few identifiable letters. Grawlixes marked a turning point, and the asterisk (*) became more widespread for its ease and simplicity. Around this time, symbolic swearing also shifted to replace individual letters, specifically vowels.
The primary reason for this is intelligibility: There are simply fewer options to go through when guessing the meaning of "f*ck" than with "*uck," which could also mean something like "suck." It''s the same reason the vowel is omitted in truncations like "fck" and "btch" on social media. Another option is to change the word to something similar, but funnier. If you''re trying to sneak past robotic censors, why not make it a little silly? You can easily swap letters to spell words like "fukc" and "bicht," or drop them entirely to make "fuk" and "bich." You can also up the goofiness by replacing consonants entirely, like with "fucc" for "fuck" or "ahh" for "ass." One can also add humor or shock value by "swearifying" relatively innocent words, like "m*n" for "men" in some feminist circles of social media, or "Tr*mp" for "Trump" in political circles. However we choose to bowdlerize, there are so many profanity options that we don''t have to stop at evading content moderation filters--we can also parody that reality. Back when I was just getting started on TikTok, one of the first videos I made was on how the words "pen" and "pencil" are entirely unrelated.
In the video, I described how "pencil" actually traces to the Latin word penis, meaning "tail"--and, yes, that''s also the source of the English word "penis." Being uninitiated to the platform, I included the full word in my captions and received a content warning that TikTok wouldn''t let me post the video. I went back and edited the spelling to.