This is the first in a series of posts exploring the use of social media platforms as tools for language acquisition.
Addictive platforms are good, actually
If we want to develop a lasting relationship with a language and a culture, perhaps we should integrate it into our daily lives. To really do that, our interactions with the language can't feel too much like a chore.1 We shouldn't need to be guilted or gamed into acquiring competence in the language.
In my first post, I argued that Toucan and Fluent offer an alternative to the disenchanting content worlds imposed on us by traditional classrooms and apps like Duolingo. By harnessing our every day web browsing context, these apps start from quotidian L1 internet content (the learner’s native language) and transform it, via algorithmic code switching, into a mix of L1 and L2 (the target language).
The result is a clever, context-first way to acquire vocabulary. But it’s also a little dubious, and still riddled with errors. Regardless of whether we Toucan, it’s clear that it shouldn’t be the sole basis for our daily language acquisition habit. We should also consume organic, real-life L2 inputs. Luckily, the internet is a vast trove of L2 content that is much more compelling than the stuff of Duolingoland. We just have to find it.
One important class of such content is media that is professionally developed to captivate lots of people: books, films, and television shows. Especially the television shows. In recent years we’ve seen foreign language Netflix series like Casa de Papel and Lupin rise to unprecedented global popularity. These shows are so compelling and watchable that millions binge them; they have almost addictive properties. I will turn to this case and language learning apps that attempt to harness these properties in a few weeks.
Another class of compelling content is the stuff produced and consumed by people like us, every day, on social media. Twitter, TikTok, and Instagram are content recommendation engines designed to grab, maintain, and monetize our attention. They keep us glued to our screens for more hours than we like to admit. This addictive property is, according social media's detractors, a bad thing.
In the language acquisition setting, this negative property is flipped on its head. The ability of these apps to surface compelling, relevant, and addictive content becomes a good thing. It presents another potential escape route from Duolingoland, another way of turning Being Online into serendipitous moments of educational value.
For the past few months, I have been reading French every day by following a few French language Twitter accounts. I am an active Twitter user; these accounts are just integrated into my main Twitter feed. In the course of this experiment, I have begun to cook up a set of arguments regarding what makes Twitter a potentially powerful platform for context-first language acquisition.
Before I can make that case, though, I think it’s important to get a handle on what Twitter is. And that’s not so straightforward a thing.
An introduction to Twitter and its concepts
We all have a sense of what the bird app is, but on the odd chance that you’re not on it, or you’re one of my friends who is an electrician with a voracious appetite for history and writerly aspirations whom I always try to convince to get on Twitter, I’ll attempt a description of it and its concepts. Later, I’ll be using some of these concepts and I’ll assume readers know what they mean. If you are already on the bird app, you might fast forward to the next section. Or read this, and tell me what I’m missing.
Twitter is a social media application for microblogging. These microblogs, or tweets, are short chunks of text, sometimes containing images, video, audio, and linked internet content. Their most common form is just a short chunk of text. Accounts publish microblogs to their profile - they tweet or post tweets. Go to an account’s profile, and you can read a short bio, as well as all of its past tweets. Accounts can mention another account by including @user_name in a tweet, which will notify the referenced account it was mentioned.


My sad attempt at self-promotion. Thank you to those who liked and promoted. This did not do numbers.
Tweets accumulate likes, and tweets can be directly re-shared via a retweet, or re-shared with commentary via a quote tweet. A tweet about another tweet or account that does not explicitly reference it is called a subtweet.
The number of likes and retweets a tweet accumulates can be thought to signify its salience or resonance. Sometimes, tweets go viral. This means a lot of accounts like and share the tweet, resulting in a rapid rise in the tweet’s popularity as it is shared and seen and liked. When a tweet resonates strongly with users for whatever reason, it can accumulate hundreds of thousands to millions of likes over the course of a day. When one has a tweet that blows up, or reaches a relatively large number of likes, the tweet is said to do numbers. I have never authored a tweet that did numbers.
The first time the word “tendies” was printed in the NYT, tweeted by a bot. This did numbers.
Chains of related tweets can be composed into threads, as in a message board or online forum. Threads are often used to get around the maximum length of a tweet, and relate longer-form content. Responses to tweets by other accounts, or replies, also accumulate under tweets and threads. These are sometimes referred to as mentions, because when someone replies to your Tweet, you receive a notification, as if they had at-mentioned you. Long chains of comments are sites of reactions, riffs, jokes, discussions and debate, as well as outrage, abuse, rogue advertising, toxicity, and a phenomenon known as reply guys, a detailed description of which you can find in this thread.

A nice reply to my tweet about my last post by a founder of Fluent. No sign of Toucan…
At its core, Twitter is a massive social graph structured by the follow relation. I follow other accounts, and then the app shows me the content those accounts produce and like and share. That stream of content is the primary interface of the app: an infinite scrolling feed called a timeline (or TL). To scroll endlessly in a helpless state of feed consumption, often when something terrible in the world is happening and your feed is filled with tweets about it, is called doomscrolling.
The app dynamically composes my timeline for me; I can curate it somewhat by choosing to follow accounts whose content I want to read, and blocking or muting accounts I don’t. But the timeline is constantly generated and refreshed by the app, based on logics of selection and ranking that are opaque to me. This timeline generation is performed by algorithms designed and trained to optimize some numerical objective function; they likely have the goal of keeping me engaged with the application, and serving me content that is relevant to me (they also serve me ads). This stream of tweets is ephemeral; if you want to save a tweet, you have to bookmark it, or it can be difficult to recover later.
With the Facebook social graph, the primary relation is that of the friend. The friend relation is bidirectional, or undirected, in math jargon. This means that all of my Facebook friends are also Facebook friends to me. But on Twitter, the follow relation is unidirectional, or directed. This means that just because I follow an account does not mean it follows me back. Its operator must explicitly choose to do so. When two accounts follow one another, they are said to be mutuals. Mutuals have a certain special status for many users, a certain kind of solidarity.


I have a mildly parasocial thing going on with this account. They don’t know they’re teaching me French…
On Twitter, the lion’s share of accounts have relatively few followers and produce relatively little content. My account is an example: I follow 1.4k accounts, but I have only 40 followers, and I rarely tweet. I have no audience, no status. A smaller concentration of accounts have lots of followers, and produce lots of content. So most users are consumers, a smaller set of users are producers, and an even smaller set are producers with audiences. This leads to a relationality sometimes called parasocial - a one-way sense of intimacy and attachment to media performers who know nothing about you.
A portmanteau for the home stretch?
There are hundreds of millions of accounts on Twitter, and not all of them are humans. The non-human ones are called bots. Twitter exposes tools to software developers for creating bots. Some bots do fun and artful things, like randomly post pictures of every Post Office in the United States from Google Street View, or tweet every time a word appears in the New York Times that has never been published in its entire history. Some bots capture tweets and recontextualize them, saving them for later. Other bots harass people, and spread misinformation.
Twitter is many Twitters
There is a corner of the Twitter graph for everyone. Actually, this is likely not to be true. Who is not on the app, and how its usage breaks out across social structure are both interesting questions.
But you, reading this right now - there is definitely a corner of Twitter for you. In fact, there are many corners of Twitter for you. Indeed, people refer colloquially to specific discursive communities on the app as distinct Twitters. These take terms such as “Black Twitter” and “Econ Twitter.” Sometimes they crystallize briefly as a hashtag, though hashtags don’t seem super in vogue these days, at least in my corners of the app. As an aside, it would be a fun computational social science project to find and observe the use of these names within the historical corpus of tweets.
I think just articulating some of these Twitters can help give us sense of what the app actually is. Here is a sample of the Twitters that shape my TL (nomenclature my own):
Academic Sociology Twitter
Venture Capital Twitter
Military Veteran Writers Published By Major Publishers And Outlets Twitter
Ethical AI Twitter
Economics Twitter
Ten Reasons Remote Work Is The Future Sign Up Here For My Remote Work Platform Twitter
Public Health Twitter
Cult Of Users Of A Note Taking App That Can Profoundly Change How You Think And Write Twitter
Digital Humanities Twitter
Analytics Engineering Twitter
You Should Write Online By Taking My How To Write Online Online Writing Course Twitter
Transportation Policy Wonk Twitter
Machine Learning Research Twitter
If You Were A Good Venture Capitalist You Would Move To Miami And Buy This Futuristic Bed That Automatically Optimizes Your Sleeping Temperature And Measures Your Sleep Quality Twitter
Voice-Driven Technology Enthusiast Twitter
Suburbs And Exurbs And Highways And Cars Are Truly Terrible Twitter
Tools For Thought Twitter
Data Science Twitter
Just A Person Who Tweets Interesting And At Times Intimate Things About Their Daily Life Which Occasionally Go Viral Resulting In The Accumulation Of Thousands Of Followers Twitter
Political Polling and Election Forecasting Twitter
Snarky Software Engineer Critiquing San Francisco Tech Culture From Within Twitter
Language Learning Tool Twitter
Aspiring Comedians And TV Writers Twitter
Twitter Meme Account Twitter
Can’t You All See That Nate Silver Is A Hack Twitter
Sociologist Comedian Cultural Critic Data Visualization Expert Twitter
Musicologist Comedian Making Approachable Classical Music Content Twitter
Design Twitter
The Same SNL Clip Of Daniel Craig Saying Ladies And Gentlemen,,,, The Weekend! Every Weekend Twitter
Of course, none of these Twitters are entirely distinct from one another. Some Twitters contain other Twitters. There are overlaps and intersections. Members of the various Twitters are members of many other Twitters. I try to get at that intersectionality and compositionality in some of the titles above.
A member of Language Learning Tool Twitter contributes a Suez canal meme
As such - and owing to Twitter’s recommendations - conversations, personal confessions, transgressions, memes, jokes, aphorisms, meme forms, articles, new words, and reactions to world events all diffuse, to varying degrees, across the different Twitters. This week, virtually all the Twitters have generated some spin on the Ship Blocking The Suez Canal meme. Some of these have mass appeal and go viral; others accumulate likes proportional to the size of a specific Twitter. This is a good example of a kind of hive mind, or zeitgeist, or context, and an intertextuality both within and across Twitters that I will later argue can be helpful for language learners.
Sociologist Comedian Cultural Critic Data Visualization Expert Twitter stays on-brand with a meta Suez canal meme
I suspect that in order to most effectively harness Twitter as a language learning tool, a learner should first be an L1 user. They should find some Twitters that resonate with them in their native language. They should establish themselves in the social graph - find their communities. They should spend time in the app, and actively like and engage with tweets; this gives the recommenders the numbers they need to offer more compelling accounts and content. They should smash that follow button when they find a smart or funny account.
Once we find our L1 Twitters, we can set out to find our L2 Twitters. More on this, and why I think it is a good idea, in posts to come.
I love listening to BBC Arabic when I do chores, so maybe this isn’t the best simile…