Remember that joke about college kids going out to that club, “The Library,” so they could say with a completely straight face they were “going to the library”? It was one of those great ways in which language was subtly subverted. Well, today’s youth is at it again, only this time online. They’ve developed “algospeak,” a way of tweaking language in real-time to route around the social media moderation seeking to block certain keywords associated with sensitive or harmful content.
In this post, we will explore why algospeak exists, how it works, and take a look at some current examples.
What Is Social Media Moderation?
Social media sites rely on UGC to help flood the feeds with blog, videos, and images that seemingly go on forever. Not all content, however, is either safe or appropriate. Moderation is how platforms enforce the rules and community guidelines to keep the area safe for users. This set of guidelines will spell out what’s allowed and what’s blocked.
While most platforms do some form of moderation, TikTok is an outlier due to its staggering popularity-150 million American users. It’s also well-documented that community guidelines for TikTok are some of the harshest out there. They contend, “Everyone who joins TikTok is free to express themselves, but we remove anything that violates our rules.” This has in part given birth to algospeak: users are working to communicate in ways that deceive detection and removal.
How does moderation work on TikTok?
A blend of AI-powered content moderation software and human moderators curate the TikTok content. The AI automatically scans the blog, videos, and comments for specific keywords, then flags any it believes violate its rules. Human moderators, in turn, review this flagged content for material that’s harmful, including blog that promote eating disorders, dangerous viral challenges, or misinformation.
The thing is, this system sometimes censorously flags and removes content that doesn’t break guidelines. For example, a video regarding the historical event of Martin Luther King Jr.’s assassination can get labeled “violent or graphic,” while mere mention of LGBTQ individuals can get flagged out of context.
While users can appeal content removals, the process can be frustrating, and appeals aren’t always successful.
Algospeak: A Modern Solution to a Modern Problem
TikTok took down a video, but people just learned to pivot. The workaround manifests itself in what users have come to call “algospeak”: an ever-changing set of alternative phrasing and symbols that get around the algorithmic filters. People who need to spread not-so-popular ideas through their videos are looking for help in using algospeak to navigate the restrictions.
But what, exactly is algospeak? Let’s examine the three key categories in detail.
1. Alternate Spellings
One of the easiest ways to bypass social media filters is to simply misspell words that are typically flagged. This can be as simple as substituting “sex” with “seggs.” More recently, users have begun incorporating emojis into these misspellings, such as typing “s🥚sy” where the egg emoji visually replaces part of the word. Similarly, “lesbian” became “le$bean,” before evolving to “le dollar bean” due to TikTok’s text-to-speech reading of it aloud.
2. Neologisms
Another way users get past the moderators is by creating entirely new words. When the COVID-19 pandemic was going on, TikTok started flagging most uses of the word “pandemic.” Users quickly figured out a way to get their words in when they began referring to it as the “panini” or the “pandemonium.”
3. Emojis
Perhaps the most iconic form of algospeak involves using emojis to convey a meaning. For example, the eggplant 🍆 and peach 🍑 emojis have long been code for explicit ideas. Other newer examples include the watermelon emoji , which has been used to represent solidarity with Palestine because of the color resemblance to the Palestinian flag. As new meanings are created overnight, emojis keep outsiders second-guessing.
This is a very clever use of languages and symbols here, serving effectively to highlight the adaptability in online communities that are always finding new ways around the content moderation systems.
Recent Examples of Algospeak
- 🌽, “corn,” or “corn star” — Refers to porn
- ⬛️ 🟧 — Another reference to porn, based on PornHub’s logo colors
- “Sewer slide” — A euphemism for suicide
- “Seggs” or “seggsy” — Refers to sex
- “Unalive” or “unalived” — Used to mean dead, kill, or killed
- “Accountant” — A term for a sex worker
- 🍉 — Symbolizes Palestine, as the colors resemble the Palestinian flag
- “Barcode” — Describes self-harm scars, which can appear like a series of lines
These terms allow users to discuss sensitive topics without triggering content moderation filters.