In 2023, the worldwide share of employees who report being stressed at work reached a record high, according to Gallup. One contributing factor is the “fissured workplace,” in which corporate managers carve up their traditional workforce and redistribute its functions to subcontractors. Often marketed in futuristic terms as a part of the tech economy, it is, in fact, a well-worn way of reducing worker power.
In response to pressures from capital markets to improve their financial performance in the 1980s and ’90s, large corporations whittled down their directly employed staff to those concentrating on “core competencies,” freeing them to fire “non-essential” laborers, such as janitors, whom they subsequently brought back as temps at significantly reduced pay—a domestic expression of labor arbitrage. Neoliberals argue that this corporate strategy is a win-win, liberating workers from being tied to one company, which now must compete for their services. But in reality, they are thrust into unregulated forms of employment with irregular hours, low earnings, no route for advancement, muddied relationships with management, and no business enterprise ultimately responsible for their welfare.
One of the primary characteristics of the fissured workplace is its obscured authority structure. When the host company sets the hours, procedures, and duties, judges the performance, and does the firing, it retains the punitive powers of management. But it devolves the provision of pay to a subcontracted agency and is thereby absolved of practically all obligations to the worker required by labor law. The employing agency likewise has few of the traditional qualities of an employer: There is no incentive, for example, to elevate employees up through the ranks, vesting in them greater recognition and authority as their competence grows. Today, the agency provides for none of this advancement—it just gets a cut.
Superficially, the so-called sharing economy is an innovative way for technology companies to link consumers directly with gig workers. But more fundamentally, platforms like Uber and DoorDash profit off the loss of labor rights. They dress up the collapse of the general expectation of full-time employment—and its corollary assumption of a family wage—in the glamor of “Tech,” shrouding socio-economic disintegration under the mask of futuristic togetherness.
This is one of the primary contributions of Silicon Valley to the neoliberal order. While appearing to foster “connection,” Big Tech’s platforms obscure the disempowerment of labor by pushing work into increasingly fragmented, digitized environments, where the humanity of the worker is hard to keep in view.
Application programming interfaces, or APIs, which hundreds of tech companies use, bring this process to perfection. APIs are used to crowdsource tasks to workers that a full-time engineering staff and its AI algorithms can’t cover (or that firms won’t pay to have covered by a stable of full-time employees). A particularly striking example of how tech companies use APIs was revealed in a December 2022 entry of the Twitter Files. Twitter, like every other social-media platform, moderates endless flows of public statements, images, and videos posted by millions of individuals across the globe to ensure that violent and sexualized imagery and content is age-restricted or taken down. The algorithm, for all its power, can’t make prudential judgments, so where it fails, the moderation tasks are farmed out via APIs to task workers—many in places like India, where labor is cheap and English-speaking—who, for pennies, flag posts. As the journalist David Zweig reports, when Twitter ratcheted up its content moderation to full-blown censorship during the coronavirus debacle, it tasked, “contractors, in places like the Philippines … to adjudicate tweets on complex topics like myocarditis and mask efficacy data.” Americans were rigidly censored by on-demand workers around the world toiling for poverty wages.
Daily uses of APIs are far more mundane, but if one is concerned with revitalizing labor and the “future of work,” they are nonetheless important. As Mary L. Garth and Siddharth Suri noted in In Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass (2019), the two nations with the largest number of task workers are the United States and India. The tasks are distributed on platforms like Amazon’s Mechanical Turk (or MTurk), CrowdFlower, Lionbridge, and Microsoft’s Universal Human Relevance System, serving companies like Uber, TripAdvisor, Match.com, Google, Facebook, among hundreds of others. “Requesters” post assignments that an engineer can’t tackle (for a variety of reasons) or are beyond the effectiveness of algorithms, such as when a task worker invisibly connects an Uber rider with a driver because the algorithm has difficulty confirming the identity of one or the other. These kinds of surreptitious actions in the algorithmic gaps happen routinely to make platforms like Uber work, but they are hidden to maintain the appearance that they run without human aid, purely on the magic of tech.
The assignments are first-come-first-serve. Workers click to secure the right to the posting, beating out thousands of others, with the task awarded to clicks at times separated by mere milliseconds. If satisfactorily completed, a small sum of money is deposited into the worker’s account. If not, it is rejected without explanation (or avenue of appeal). One MTurk worker, “Joan,” told Garth and Suri that when she is really on her game, she can complete enough assignments to net roughly $22 in an hour. That’s a nice sum, but it’s impossible to sustain, because in the span of 60 minutes, Joan must plow through about 1,100 of what Amazon calls Human Intelligence Tasks to earn it. Only 4 percent of MTurk workers make more than $7.25 an hour.
Platforms like MTurk feed off of the fissured workplace and accelerate it. Garth and Suri coined “ghost workers” to describe these laborers, because every aspect of their personhood is concealed from the API market. Instead of allowing workers to operate under their own name, it is standard to assign them random alphanumeric identities: “The anonymization…of workers can make them seem interchangeable to requesters. For example, what’s the difference between hiring A16HE9ETNPNONN and A6GQR3WXFSITY?”
They are given no opportunity for advancement, no way to distinguish themselves, no coworkers, no explanation for being de-platformed, and no human boss—just a computer interface loaded with tasks that algorithms can’t handle.
This, according to Garth and Suri, is the true future of work, in which human labor is slotted between the gaps in AI’s programming. As machines are deployed by corporations to monopolize greater shares of the labor process, new forms of work are generated that require human services at the edges of AI’s reach. “Thus,” they write, “there is an ever-moving frontier between what machines can and can’t solve…. As machines solve more and more problems, we continue to identify needs for augmenting rather than replacing human effort.”
Automation, then, will create new jobs, not mass unemployment. But it doesn’t necessarily follow that we will want the jobs it creates.
The use of technology to assert managerial control over the labor process long precedes the fissured workplace and the advent of the digital age. Even so, these developments provide new tools for extending managerial control. In his classic treatise, Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century (1974), Harry Braverman analyzed how firms, beginning in the late 19th century, implemented “scientific management,” that is, an exhaustive study of workers themselves, including close observation of their bodily motions and other dimensions, such as rhythm and timing, when undertaking a task. Through scientific management, a synoptic view of the labor process was constructed from the outside. Management could then redesign the process and impose it back on the labor force from above, transfiguring work into something totally unrecognizable. Labor lost control of its own activity. Thought was severed from action, and action was divided and subdivided into ever smaller pieces. And as for the laborer himself, he was degraded along with his work, having been de-skilled and remade according to a “view of human beings in machine terms.” Braverman summarized:
Here the entire work operation, down to its smallest motion, is conceptualized by the management and engineering staffs, laid out, measured, fitted with training and performance standards—all entirely in advance. The human instruments are adapted to the machinery of production according to specifications that resemble nothing so much as machine-capacity specifications.
In other words, under scientific management, laborers were fitted to the machine and made into something very much like machines themselves. This laid the foundation for the next stage: displacement of this machine-like man from the labor process by computers and robots, which were becoming the final repository of what was once his practical knowledge. Yes, millions of laborers were freed up for clerical work and other white-collar professions as a result of this method, but those jobs, too—Braverman argued in 1974, long before it became a commonplace of social theory—were also in the process of being proletarianized by scientific management. The clerical professions, which were epiphenomenal of scientific management’s efficiency gains over manufacturing, were becoming its next victims.
Braverman’s analysis of the corporate uses of technology to fracture the labor process describe the use of APIs to a striking degree. On platforms like MTurk, the work is so overdetermined that companies credibly consider the laborers that toil away there as mere “customers.” Uber, for example,conceives of its drivers as “end users.” In official documents and public statements, Uber takes pains to describe itself as a “technology” company to resist any suggestion that it is a transportation service. This is strategically important because, besides baffling lawmakers and regulators who worshipfully defer to the “disruptive” power of “Tech,” it allows the company to identify its drivers as consumers of the app. End users, as opposed to employees, laborers, or workers, are things to which Uber, an especially lawless and predatory company, owes nothing.
None of this would be possible, legally or socially, if Uber were not operating under the shield of technology. This technology has been designed and deployed specifically to diminish the status and rights of its drivers, on the grounds that the skills they provide are so meager that it renders their labor indistinguishable from “consumption.” This is an epochal shift in the nature of labor relations and identity. The rise of Uber and similar platforms, like APIs, signal an advance in management’s unflagging efforts to strip mine labor of what little remains of its native potency and to strip workers of their rights. Success, as we have seen, depends upon workers being made indistinguishable from machines—or as Garth and Suri might put it, ghosts in the algorithm.