Topline
Last weekend, actress and influencer Julia Fox apologized after she misread a TikToker’s reference to “mascara,” not knowing it was “algospeak” for sexual assault, the latest misunderstanding caused by code words used on social media devised to avoid algorithmic censors.
In this photo illustration, the download page for the TikTok app is displayed on an Apple iPhone.
Key Facts
Social media users are commonly using algospeak, code words used to avoid AI content moderation tools that flag user content for violating a social media apps rules, or which might be sensitive in nature.
Algospeak is mostly used on TikTok as their content moderation is more aggressive than other social media apps, and will restrict users from posting for a longer amount of time than other platforms for violating its community guidelines.
Not only are words altered but the use of emojis to infer different meanings has grown, and almost one-third of Americans who use social media have said they use emojis and altered phrases to communicate banned terms, according to data from Telus International, a Canadian company that provides AI content moderation services.
There are no laws in place that serve as a guideline for social media companies on how to navigate AI content moderation in a transparent manner, leaving them to their own devices on how to use AI for content moderation.
The automated content moderators often cast a wide net when looking at videos it thinks displays hateful, racist and sexually explicit content, though words are not so clearly defined.
Content creators who make money have to carefully navigate what words can be used as their content could be removed and their accounts banned, though TikTok does provide a way for content creators to appeal removed videos.
Common ‘algospeak’ Words
- Panini/Panorama/Panoramic = Pandemic
- Mascara = Boyfriend/Romantic partner or can refer to male genitals
- Unalive = Suicide/Kill
- Seggs/Shmex = Sex
- Corn or 🌽 = Porn/Adult Industry
- Cornucopia = Homophobia
- Leg Booty = Member of LGBTQ Community
- Le dollar bean = Lesbian
- Accountant = Sex worker
- S.A. = Sexual Assault
- Camping = Abortion
- Ninja or 🥷 = Derogatory terms and hate speech towards the Black community
Proposed Legislation
Back in 2019, U.S. Senator Ron Wyden (D-Ore.) introduced the Algorithmic Accountability Act, a bill meant to ensure that AI algorithms are fair and nondiscriminatory. “Transparency and accountability are essential to give consumers choice and provide policymakers with the information needed to set the rules of the road for critical decision systems,” Wyden said. The bill would rely on the Federal Trading Commission to make regulations and have a structured guideline for social media companies to assess and report how automating important decision making processes affects consumers.
Crucial Quote
“The reality is that tech companies have been using automated tools to moderate content for a really long time and while it’s touted as this sophisticated machine learning, it’s often just a list of words they think are problematic,” Ángel Díaz, a lecturer at the UCLA School of Law who studies technology and racial discrimination, told the Washington Post.
Further Reading
From Camping To Cheese Pizza, ‘Algospeak’ Is Taking Over Social Media (Forbes)
