When the COVID-19 pandemic started to spread from country to country, and governments began to close borders, headlines erupted with speculations about deglobalization. The implications felt uncommonly visceral, far beyond glib corporate abstractions about ruptured supply chains or the stilted flow of capital. Ordinary people worried about the present, about being cut off from loved ones, about being isolated and left without resources. And about a future of worsened nationalism and ever-more rigid borders.
Cheap language instruction is possible because of the same things that give us cheap coffee, budget car rides, and bottom-dollar high-street fashion—that is, outsourcing, underpaid Global South labor, and the gig economy.
Bright spots amidst these dire headlines were the culture stories suggesting that being shut up indoors presented a golden opportunity for learning a new language. What’s more, a chorus of company promotions, which the media rushed to celebrate, meant that doing so was not just a remedy against boredom, but also a cheap, expedient substitute for travel that doubled as a form of self-improvement. True, a journey with Duolingo or Babbel or Memrise might fall short of actually “swanning around, ordering meals at sidewalk cafes, shopping at street markets, slipping into darkened theaters,” the New York Times noted—but, it is “free, or at the very least won’t break the bank.”
The language industry’s web pages and app interfaces, some already modeled after airline industry templates, obliged with package offers for “unlimited languages” and “community” hangouts. And thousands did click on them, booking a “vacation” of for-profit multilingualism. Some did so by downloading a gamified app. Others signed up for a one-on-one tutoring platform or opted for some semi-automated combination of both.
As a result, in a moment of tanking prosperity, language-learning companies’ profits continued to climb; new investor offers poured in. Duolingo is now certainly on track to go public in 2021, and Preply, a tutoring company founded in Ukraine but operating globally, says it has doubled its fundraising totals. Amidst this, however, few have stopped to wonder what was making language instruction so affordable. Perhaps because the answer is too obvious: the same things that have given us cheap coffee, budget car rides, and bottom-dollar high-street fashion—that is, outsourcing, underpaid Global South labor, and the gig economy.
The fast language industry, in other words, has not been substantially fairer than most other fast industries, be it fast food, fast fashion, or fast journalism. And yet, it has largely escaped the critiques, let alone outrage, often leveled at those other industries. While these exploitative labor practices are certainly not new—the companies have built their entire model around them—COVID-19 has thrown many entrenched labor inequities into sharper relief, offering a unique opportunity to reflect on why we have not eyed the fast language industry with the same skepticism as, say, Uber or Lyft. Such a reflection prompts even more fundamental questions: Is multilingualism always more desirable than monolingualism, regardless the human cost? And can multilingualism on a budget be equitable—and how?
• • •
English, a legacy of U.S. imperialism in the Philippines, has made the country one of Asia’s largest providers of proficient speakers, but it is certainly not showering online language teachers with riches.
It wasn’t for lack of reportage that the media’s endorsements of COVID-19 language learning—from Fast Company to the New York Times—has skirted the question of labor conditions and pay. On March 23, 2020, Reuters ran an article on two Chinese companies that staffed their one-on-one English-learning tutorials with tens of thousands of language teachers based in the Philippines. One Filipino freelancer reported working from home for ten hours a day, giving twenty lessons per day, and never taking days off. The company she works for, 51Talk, charges learners an average of 45 yuan ($6.36) per 25-minute session. Yet teachers receive on average only 130 pesos ($2.53) per hour. English, a legacy of U.S. imperialism in the country, may have made the Philippines one of Asia’s largest providers of proficient speakers, but it is certainly not showering online language teachers with riches. In short, COVID-19 has made old inequities resonate under new circumstances.
Alas, 51Talk’s model is not exceptional in today’s language industry. Whether a company oversees a language learning app, a live tutoring platform, or some hybrid, a perusal of its job listings and employee roster will lead inevitably to the conclusion that the majority of its permanent employees are software engineers and interface designers, followed by marketing professionals, data analysts, product managers, accountants, linguistics researchers, and leadership. The labor force that generates the actual language content, by contrast, is mostly freelance and therefore ineligible for benefits such as health care.
The freelance arrangement may work for some; there is no lack of semi-professionals and non-professionals who are looking for extra income. However, for professionals, the outcomes can be disastrous, spelling long workdays, precarity, and an endless scramble to compete in a global marketplace against an unlimited supply of qualified candidates. The same goes for teachers of “rarely taught” languages and professionals who are non-“native speakers.” And those who speak language variants, including African American English, which are devalued by contrast to what linguist April Baker-Bell calls Dominant American English. These people often find themselves at the bottom of the pay scale, which only perpetuates existing racial and financial inequities.
• • •
“Learners” on Duolingo are actually just performing the free labor—or even paying for the privilege—of helping the company improve its proprietary algorithms.
These failures are congenital to the language industry reinvented for the digital age, which has dispensed with the methods and signature teaching styles once fundamental for brick-and-mortar schools. Of this, Duolingo has been a paragon. Launched in 2011 by Guatemalan programmer and entrepreneur Luis von Ahn, Duolingo has built its emporium on what the Guardian called “democratization of language learning.” Its “freemium” business model has brought it 300 million users.
However, what the company has really done is apply to language learning von Ahn’s previous experience of building the bot-detection programs CAPTCHA and reCAPTCHA, both of which had been criticized for tricking Internet users into doing free transcription labor. In fact, CAPTCHA’s old site overtly names Duolingo as existing principally as a tool to help computer systems improve their natural language processing and machine learning, meaning that language “learners” on Duolingo are actually just performing the free labor—or even paying for the privilege—of helping the company improve its proprietary algorithms. In this sense, it is much like Amazon’s Mechanical Turk, another model of exploitative human-assisted computer labor critiqued extensively in Mary L. Gray and Siddharth Suri’s Ghost Work (2019).
With $138 million in venture capital investment, Duolingo has recently been valued at $1.5 billion. The provision of free labor, however, remains fundamental to its creation of value in the digital economy, in line with what theorist Tiziana Terranova has argued more broadly. The company uses its so-called Incubator to look for volunteer course developers willing to “change the world” by contributing to the “worthy cause” of teaching “as many of the world’s languages to as many people as possible, completely free of charge.” In addition to unpaid altruistic labor, Duolingo benefits from the data it produces on how to maximize user retention, which it generates by analyzing the behavior of the millions of people on its app. The findings appear in trade and peer reviewed journals, boosting the company’s clout. Meanwhile, the company’s careers page doesn’t list a single full-time language-teaching opening, despite active hiring in all other areas.
As popular as Duolingo is, it has struggled to become profitable. Initially, von Ahn tried to charge companies such as CNN for the translations assigned to app users as homework. Soon enough, he realized that translation was “a crappy business.” With the Duolingo English Test (DET), the company has now pivoted toward the English language certification market.
Language testing experts have pushed back against the use of the DET for high-stakes university admissions, arguing that the test is “woefully inadequate as a measure of a test taker’s academic English proficiency.” However, these critics miss the point. In both the language learning app and the test, Duolingo follows the principles set by the guru of “disruptive innovation,” Clayton Christensen. In the influential book The Innovator’s Dilemma (1997), Christensen claims that disruptive technologies take over established markets not by overperforming their competitors’ services and products but by underperforming yet offering other features that customers value. In Duolingo’s case, it is the significantly lower cost of its test ($49 as opposed to more than $200 for the TOEFL and the IELTS) and the fact that it can be taken online anytime, without having to travel to a test center.
Needless to say, COVID-19 has been providential. With the closure of testing centers, many universities that previously disregarded the DET now have had little choice but to accept it as an alternative to the TOEFL and the IELTS, despite its shortcomings. The low cost and convenience may mean that DET is here to stay, pandemic or not, as long as international education itself remains viable.
• • •
Although the innovation of these platforms is often portrayed as technological, the more impactful innovation is likely the evasion of labor regulations that have created twenty-first-century jobs with nineteenth-century working conditions.
Online language tutoring platforms such as italki and Preply are also reaping benefits from COVID-19. According to Forbes, since the beginning of the lockdown in the United States, Preply “has seen the number of new customers double week over week.” In contrast to language learning apps, language tutoring platforms function as marketplaces for connecting learners with teachers or conversation partners.
This approach brings to language learning the recognizable principles of platform capitalism. Customers and workers are matched online; workers create “personalized” or “bespoke” content; workers are rated by customers; although workers are considered independent contractors, a commission fee ranging from 15 percent (italki) to up to 33 percent (Preply) goes to the company that owns the platform. The overabundance of offerings encourages customers to try out new workers, instead of developing lasting relationships. Worker compensation is seemingly self-set but in reality depends on the ratings, location, the supply-and-demand dynamic for each language, as well as economic circumstances such as customer un- or underemployment. Cue, once again, the collateral damage of COVID-19.
Like Airbnb and Uber, these nimble and global platforms cut across many local markets, exploiting the opportunity to circumvent the standards to which typical certified teachers are ordinarily held (and by which they are protected) in a labor market regulated by local laws, unions, and professional associations. In doing so, they capture rents at the same time that they worsen the precarity of teachers. Although the innovation of these platforms is often portrayed as technological, the more impactful innovation is likely the evasion of labor regulations that have created twenty-first-century jobs with nineteenth-century working conditions.
This exacerbates the preexisting weaknesses that have plagued the entire language industry (translation and interpreting, language teaching) across borders. The worsening has been slow, steady, and all-encompassing. Indeed, the degree to which near-collapse is accepted as the baseline of the language industry—and the under- or defunding of many languages professions (despite lip service to multilingualism)—may account for why this more recent crisis has received little attention. It may also explain why so few people have wondered who bears the loss for their free or cheap language lessons during COVID-19.
Language teacher precarity, it is no secret, is an inbuilt feature of K–12 and higher education in more than one country. At most schools, languages are a low priority, and teachers, if hired, have to adapt to constant cost-cutting while plugging curricular gaps in more than one language on a single salary. At universities, in turn, “global learning” may be the mantra, and yet language programs are among the first to be slashed in response to fiscal exigencies. Instruction in the “rarely taught” languages, forever at the whim of geopolitics and fickle market forces, tends to be a budget-cutting casualty even in the sunniest of times. Indigenous language programs are barely in existence. Study abroad increasingly devolves into a performance of monolingual multiculturalism and now hangs by a thread due to the pandemic-induced shrinkage of higher education.
This hits language teachers hard. Overwhelmingly graduate students or contingent faculty, they have long been among the academic precariat—well before the boom in adjuncts made national news—and thus have few employment rights, heavy workloads, and meager compensation. With short-term contracts, scarce union representation, and lacking for benefits, they are a step away from working for Duolingo, italki, or the better-paying (but, for this reason, more skeptically reviewed) Verbling. Many already do.
One would be remiss not to point out that the standardization of language education at all levels has done its share to pave the way for gamifying language learning. Since the 1980s, the American Council on the Teaching of Foreign Languages (ACTFL) has invested in envisioning language learning as the linear, easily measurable, and “readable” development of communication skills and competencies, as shown in the Proficiency Guidelines. It is no surprise that U.S. military and security agencies had pursued such compartmentalization of language into proficiency levels years earlier. A product like Duolingo, where you level up skills and gain crowns as you complete levels, is the logical extension of this narrow way of defining language learning.
And then, of course, there is the rest of the language industry. It is bigger though not much better than education. Many translation professionals are obliged to work as independent contractors, without an option of formal full-time employment or union membership. This means, among other things, that they are their own billing departments and have little leverage when clients fail to pay them. Hundreds of Dutch interpreters and translators recently went on strike to protest their government’s increasing acceptance of non-professional services at lower rates. The trend is to dip into the global labor pool, suppress pay and benefits, circumvent local regulation, and expand the gig economy while few industry outsiders pay heed.
• • •
Going forward, any broadminded politics of justice should necessarily include a recognition that, while multilingualism is laudatory, the means by which one becomes multilingual also matter.
Going forward, any broadminded politics of justice should necessarily include a recognition that, while multilingualism is laudatory, the means by which one becomes multilingual also matter. Much of language learning is now yoked to the exploitative dynamics of tourism and the predatory extraction of human capital. We have put up with “products” inspired less by linguistic justice than by game design. We have acquiesced to corporate cognitive-science experiments that harvest and mold our habits as data (“We want to help our users create a habit,” they tell us. “We want to make it like brushing your teeth.”). When it comes to acquiring languages, we have learned to ignore our agency, our sense of ethics and social justice, even as we have flexed these same faculties to support a bookstore or a clothing company. Simply put, we have consented to being the consumers that the language industry wants us to be.
To be fair, it is comforting to feel connected to other parts of the world without leaving our home, while feeling like we have helped someone put dinner on the table for showing us a sliver of the world. However, our choices matter. Free or low-cost services are seductive, but often obscure unacceptable labor conditions. Under the guise of free service and open community participation, as media scholar Mark Andrejevic has argued, digital platforms uphold the familiar logic of private ownership of unpaid and underpaid productive resources. COVID-19 has exacerbated the ultimately unsustainable dynamic of a business built on extracting value from the labor of low-paid teachers, the glut of user-generated content, and the data produced by millions of uncompensated users.
But if we must be cast as consumers, we can at least be responsible ones and ask the same kinds of questions we are increasingly learning to ask when shopping for other things. If the service is free, who bears the cost? Is the provider honest about pay rates for teachers? Are teacher positions disguised under such euphemisms as “live coach,” “guide,” or “conversation partner”? Are career openings focused on software engineering, user experience, and digital marketing strategies rather than on teaching and linguistics? Is the phrase “native speaker” privileged in descriptions of teachers, at the cost of their training or expertise? Is user data collected to run engagement experiments that fine-tune the persuasion of apps designed to steer our behavior? Mass skepticism about such offerings would give the media a reason to report more critically on the industry’s abusive practices and to heap fewer mindless praises on it. That, too, is leverage.
In addition, fairer, low-cost alternatives that carve out roles for participants rather than consumers do in fact exist. For K–12 and higher education institutions, there are co-educational projects that do not rely on outsourced labor. Individuals can also circumvent the lopsided market relations and pay-to-play transactions of the low-cost language industry in favor of informal, peer-to-peer initiatives such as the very low-tech #LockdownLanguageChat, a hashtag started by education and language acquisition expert Uju Anya to help people find learning partners for equitable arrangements.
However, like with everything else, individual consumer choices only go so far. For change to be lasting, it must be systemic. Ultimately, not only companies but entire societies need grassroots narratives about why language learning matters or is valuable. As scholar David Gramling ventures, this begins with greater suspicion about the notion of value itself. In the tech world, to which the digital-age language industry belongs, “value” has become a cipher for languages being reduced to translatable, practical, divisible, and monetizable shares to trade, instead of the messy, dissonant, non-standardizable, fiercely individual codes of expression that they are. This glib and superficial supralingualism, as Gramling dubs it, works well for governments, international organizations, banking and finance, and the tech industry. Yet multilingualism—the vaunted instrument of democratic citizenship—it is not and will never be.