algorithmic performativity

Chat, we all act differently when we’re being watched. There’s a pressure to avoid embarrassment, to present “authentically,” to put others at ease.

Sociologist Erving Goffman calls this performance—the idea that all public interactions are a kind of theatrical act. You’ll put on a different performance for your college friends and your grandmother; TV broadcasters will put on a different performance when on air and behind the scenes.

In the same way, social media algorithms are uniquely changing how we present ourselves online, since they come with a completely new type of spectator: the algorithm itself.

Let’s take a step back and identify the obvious: creators have an idea of who their target audience is, and accommodate their communication to match that.

This is why we have different “influencer accents” for different social media niches. The MrBeast-derived “entertainment influencer” accent is meant to grab the attention of easily distracted middle schoolers. The stereotypical “lifestyle influencer” accent is layered with the Valley Girl aesthetic expected by young women. My “educational influencer” accent is tailored for a nerdy yet brainrotted fanbase.

Essentially, we’re performing for different demographics through our self-presentation. You can also see this in our word choice: lifestyle influencers will all use the same coquette emojis (🧸🕊️🎀); I will use more academic language than MrBeast. These are aesthetic symbols of in-group belonging.

So far I’m just describing the natural human phenomenon of audience design—that we modify our speech to express connection with our perceived audience. But then there’s also the reality that we’re never exactly sure who is perceiving us on social media. Algorithms often distribute content in unexpected ways, leading to new or unintended audiences.

That might not be too different from the days of television (where you’re not directly aware of who’s watching), but even broadcasters have general ideas of their viewership. On social media, things are less predictable: you never know when a video is going to blow up or how it will spread.

I think this is partially why the word “chat” has emerged as a vocative noun. We know there is a chat watching us, but we’re not aware who is in the chat, so the term is a useful, shorthand catch-all for both our actual and unexpected audience. It reflects our own uncertainty of who’s watching us online.

While “chat” already indicates that algorithms are changing our presentation of self, we additionally perform for the algorithm itself. Influencers will deliberately use trending slang, because we want the algorithm to push our videos further. We’ll avoid sensitive keywords, because we don’t want our videos to be suppressed. We’ll use attention-grabbing language, because we know platforms rely on the commodification of user attention.

If anything, we may be performing more for the algorithm than for other people. I’ve interviewed dozens of creators who list “going viral” as their chief motivation behind content creation (rather than reaching a specific audience), and even well-meaning communicators can’t get around the fact that they have to first please The Algorithm for their message to be distributed.

In some cases, however, I’ve seen the algorithmic panopticon engender a fourth layer of performativity, one that acknowledges the others on a meta-linguistic level.

Consider the watermelon emoji (🍉), understood to represent the Palestinian flag, or the term “IOF,” an algospeak substitution for “IDF” that turns the word “defense” into “occupation.” These terms work on the literal and social signaling levels; they work for the algorithm, which suppresses discussion of the war in Gaza; and then they also signal hey, by the way, everything we say is being watched.

“Performance” isn’t some passive thing. It replicates previous meanings to create new meanings in our heads. An algorithmic substitution in a countercultural context also becomes a wink to the audience, its very existence a platform critique. By calling attention to what words can’t be used, you remind people of why they can’t be used—because these social media companies are surveilling and regulating our speech.

---

If you liked this essay, make sure to get my book Algospeak!! These next two days are particularly important because they can still count toward bestseller lists. The book is now available everywhere in the US, UK, and Canada.

And a few life updates since my launch post:

• Book tour has been completely insane. I went on CBS and NBC morning news, sold out the Harvard Bookstore and Kepler’s Books, and somebody made Wikipedia pages for me and the book!

• New coverage in AP News, LA Times, The Boston Globe, and NPR

• I’ll be talking at Wordhack in NYC this July 24 (tickets here) in addition to the Politics and Prose event in DC this July 25 (listing here)

Thank you all for your lovely support!! I appreciate you :)