Digital Piecework

Digital Piecework

Homework and piece pay in the garment industry were largely abolished by the global labor struggles that preceded the New Deal. Silicon Valley capitalists have brought the model back.

Jeff Bezos has amassed over $85 billion since January, while many Amazon MTurk workers earn less than the federal minimum wage. (John Michael Snowden)

In 1975 women in Iceland went on strike, bringing the entire nation to a standstill. For one full day, referred to now as “the long Friday,” 90 percent of women didn’t show up to their jobs, and refused to cook, clean, or look after children and the elderly. Men scrambled—overwhelming restaurants with food orders and working longer and harder than usual in an attempt to do both care work and their paid work. The point of the strike was to draw attention to what socialist feminists had been arguing for decades: economies are built upon women’s unpaid labor. Their action showed how capitalism has a propensity to make invisible the labor of people with little political power—both by refusing to recognize it as work and by refusing to pay for it.

In the late nineteenth and early twentieth centuries, for example, U.S. industrialists exploited women’s subordinate position in both the family and the labor market to develop and extend “homework.” Garment manufacturers distributed tasks to immigrant women living in crowded tenements, paying them by the piece, not the hour. This piecework was advertised as “pleasure,” where a woman might make supplemental income while talking with friends. In reality, women homeworkers labored for eight to ten hours a day finishing the majority of all garments produced in the United States. That work took place in between, during, and after unpaid domestic work, at rates that were roughly one half of what women factory workers made. Homework and piece pay in the garment industry were largely abolished by the global labor struggles that preceded the New Deal and the legal standardization of the minimum wage. Though women, people of color, and immigrants continued to earn less than their white, male counterparts while laboring outside the home, the notion of a “living wage” became understood as a prerequisite for citizenship and freedom in a democracy.

Silicon Valley capitalists have brought back piecework, using legal gray zones and digital machinery to accelerate the amount of work that goes unpaid. But, bedazzled by the technology and corporate narratives, few people have noticed. When venture-funded labor platform companies like Uber, Lyft, and Amazon Mechanical Turk (MTurk) rose in popularity during the Great Recession, they promised to provide a source of flexible work and “freedom for people of all walks of life,” as one Uber ad put it. In a time of high unemployment and stagnant wages, jobs that people could get by downloading software or creating a profile seemed like a magical solution for precarious lives. But the corporate assurances were deceptive. While the companies might have created new ways for people to earn income, workers in the gig economy today labor for longer and earn far less.

This shift back to an earlier era of U.S. capitalism has been disguised by a rhetoric of technological advancement and innovation. Indeed, instead of discussing how to regulate piecework as a resurgent industrial practice, much of the contemporary debate about the future of work focuses on impending automation and the growth of the alternative workforce. Pundits across the political spectrum claim both that automation is an existential threat to paid work and that a vast majority of the U.S. workforce consists, or will soon consist, of independent contractor workers. Neither claim is empirically true, but these related ideas have shaped the perceptions of policymakers, who have tended to favor responses that help companies to profit at the expense of workers’ lives.

The historical relationship between capitalism, workers, and machines suggests that automation does not make labor obsolete; it reorders it, often rendering it invisible. Time spent laboring that was once accounted for through wages and legal protections becomes unpaid and unprotected. Understanding this difference suggests different policy priorities, like regulation and enforcement of existing labor protections. Gig workers, for example, are central to the supply chain of work that makes vehicular automation possible, often conducting the most time-intensive labor. In interviews I conducted with ride-hail drivers—who drive for companies like Lyft and Uber—and data processors—who sort and categorize data through platforms like MTurk—workers described a variety of unpaid work that they must do.

Like women working in crowded tenements and finishing sweaters for a pittance, gig workers work in between, during, and after other forms of paid and unpaid work. The workers insist that the application of technology has increased the amount of labor that goes unremunerated, and exacerbated feelings of anxiety and uncertainty. When they are unable to attain promised or desired earnings, it feels like a personal failure. Black-box algorithms, meanwhile, use psychological inducements to manipulate the price per piece, encouraging workers to work harder for reduced wages. As gig workers tell us, these mechanisms bely corporate tales of flexibility and independence—the idea that digital pieceworkers are free to decide how to use their time to earn. This gap—between the official narrative and real working conditions—has fueled growing collective organization in the gig economy. As workers communicate about the everyday problems they face on the job, many are also beginning to develop a political response—and finding ways to fight against the most insidious elements of the labor practices and algorithms that control them.

Unpaid Labor

Jeff Bezos launched Amazon Mechanical Turk in 2005, unveiling a plan to provision “humans as a service” through a crowdsourcing labor platform. On MTurk’s website, requesters with data-related microtasks connect to an atomized and dispersed virtual workforce that competes for and completes the tasks. Individual workers are paid not for their time, but by the piece, each of which is called a Human Intelligence Task (HIT). Unlike the garment homeworkers of a century ago, today’s digital homeworkers have to spend time competing for assignments and they must accept the risk that requesters will reject their work and decline to pay for it. These workers, or “Turkers,” are treated as independent contractors, and neither the requesters nor the labor platform companies assume the legal responsibilities of an employer. Thus, Turkers—more than half of whom are based in the United States—do not have access to the minimum wage, overtime, or any safety net protections. Since the amount of payment for each task is typically a few cents (sometimes less than one cent), data homeworkers are compelled to work swiftly through a set of tasks for extraordinarily low and unpredictable wages.

While manufacturers in the previous century claimed it was impossible to measure the time that garment homeworkers spent laboring in order to pay them by the hour and not the piece, time laboring online can be meticulously accounted for. Turkers are even advised to install accessory scripts into their browser, which calculate how much money they will earn per hour if they move through batches of HITs at a particular speed. (Although paid by the task, gigging homeworkers still think of their time through the medium of the hourly wage.) The scripts operate as tools of self-management and time discipline—pushing workers without human supervisors to maintain an exacting speed in order to increase their income. But the scripts are also the only way that these workers can even attempt to approximate how much—or how little—money they will make on a given day or week.

Digital homeworkers are acutely aware of what these scripts don’t account for: how much time they spend looking for work and doing data-processing tasks (often central to artificial intelligence and automation shifts) that go unpaid. Janey, who lives in a small former mining town in Appalachia and has been a digital pieceworker for almost five years, expressed how profoundly frustrated she was at the functional logic of MTurk, which prevented her from predicting and calculating potential income. The insecurities and demands of digital homework nagged at her through the day and even into the night. Both her conscious and unconscious time was spent looking for work without compensation:

If I work 12–16 hours a day, I’ll make maybe $5/hour. That’s when there is work, but when you’re sitting in between jobs and you consider that time, when you’re just looking for work, then the hourly wage falls dramatically. There are so many of us now, and fewer quality jobs. Sometimes I wake up in the middle of the night just to see if I can grab some good requests. Most HITs are gone if you don’t click right away.

Janey and her homeworking colleagues work long and unpredictable hours, far exceeding the traditional eight-hour shift, but without any overtime pay. When I asked Janey how she decided that she had worked enough in one day, she answered that it was only when she met her financial goals that she let herself rest.

If I need to make $50 to pay the rent, then I’ll work sixteen hours straight. Whatever I need to do. . . . But then there are those times when you don’t get paid or your work is rejected . . . so you can’t predict the time or the money, really. But you do the best you can.

As Janey articulates, the very logic of MTurk makes it impossible to make any meaningful wage calculation. While a number of studies have attempted to capture how much money workers make in the gig economy, with vigorous debates among economists on how the data should be interpreted, the larger point is that workers themselves cannot know or understand their income in relationship to time spent working or their own expenses. This lack of income predictability has great human cost for workers like Janey. But rather than address this, gig companies have used technology to manipulate it, leveraging worker anxiety to increase profits.

Automating Anxiety

Early gambling machines borrowed from behavioral management techniques adopted in factory production. Today, the borrowing goes in the other direction: digital labor platforms use behavioral insights on addiction that have long been leveraged by the digital gambling industry. It’s little surprise that gig workers often describe their work through the feelings of angst, hope, and loss endemic to gambling.

While MTurk data processors have to compete for HITs, Uber and Lyft drivers’ algorithmic bosses send them pings—which sound like a slot machine—telling them that they’ve won a ride and, if they’re lucky, enough money to make ends meet. Drivers are paid by the piece, but unlike the garment homeworkers of the previous century, they have no idea how many pieces they will be allotted or how much they will earn per piece. With changing fares and limitless competition, take-home pay remains an everyday gamble. Meanwhile, algorithms and graphic design obscure the stakes, often dangling impossible-to-reach carrots.

To keep workers in the game, Uber and Lyft utilize the expertise of behavioral psychologists. The hidden processes by which algorithms deliver wins and losses are designed to create an enchanting sense of possibility and hope. Drivers receive in-app individualized “promotions” and badges for their “performance” to make them feel special and rewarded, and yet drivers say it’s often unclear what they are getting. Text messages and alerts nudge and cajole them to work for longer than they planned. As John, a ride-hail driver in Los Angeles who has been involved in driver organizing, explained,

My background is tech and even though I’m familiar with these strategies used in gaming and casinos, I still play. I spent four years living in my car playing this game [driving for Lyft]. When I hear the sound of a ride request, it’s like winning the jackpot.

The opaque algorithms used by Uber and Lyft are a source of much conjecture. No one quite knows how the machine works, and many feel frequently frustrated and even duped. Angeline, an immigrant driver in San Francisco who lost her small business to the Great Recession, told me:

They have so many tricks! These companies, they always have something going on to compete with each other. Somedays, they will give you lots of rides—they make you drive a lot. But they are very cheap rides. $1. I got a $1 ride yesterday . . . But then, you think, I am getting so many rides, I should stay working on this app and not switch to the other because what if one of those rides is a big ride? It is a trick to keep you driving for them and not to switch over.

Drivers receive alerts about “surges” and “personal power zones,” which they understand as increased demand in a particular area. The idea, Uber and Lyft tell the public in the familiar language of economic efficiency, is to calibrate supply and demand. But it is not efficient for the drivers, who are compelled by texts, emails, and in-app pop-ups to drive toward a saturated market. By the time they arrive, they often find there is no demand, and they are left to wonder why Uber sent them in that direction. In a group text of self-organizing Uber drivers that I was included on in 2016, drivers would alert each other about false surges: “I’m here! There’s nothing.” Their efforts at collective action were an attempt to gain some power over the shrouded logic behind the company’s directives.

Many Uber and Lyft drivers, like the Amazon Mechanical Turkers, work until they make a certain amount of money—to pay rent, for example, or to buy groceries. Drivers say that Uber’s algorithms seem to know how much they need to make and then make that goal harder to achieve. And as drivers approach their goal—when they are ready to stop working—the companies sometimes give them better fares in an attempt to keep them on the job. Ivri, a ride-hail driver in the Bay Area and vocal spokesman for other drivers, explained:

I have had too many “good” things happen to me right when I start heading home. The sense is that the system knows my routine and keeps intervening in all sorts of ways to get the most out of me. Irregular and inconsistent rewards definitely do the trick. And when they happen, drivers go and tell other drivers.

The perception that a prize could be won and the system gamed is a nagging myth that agonizes many drivers as they try to make sense of their earnings. Adil, a recent refugee, said that he had friends who had been able to make Uber work for them. Why couldn’t he?

My friends, they say they make so much money, but I don’t understand how people can do it? No matter how many hours I work, I don’t make what they make. I don’t know what I am doing wrong. I don’t know how I am doing it, but I have to . . . I don’t find another option.

When we spoke, Adil was visibly anxious and overwhelmed. He wiped his forehead and occasionally welled with tears. He had to make $2,700 a month to pay his rent, and buy groceries, gas, and insurance. Adil had set a weekly goal of $1,000, but after a few months of driving, he had to work even longer to make that same amount. He spent most nights in his car, three hours away from his children and wife, in order to maximize his earnings. Adil was certain Uber’s algorithms had figured out his income target and used it to keep him working longer hours. Each day of hard work, he told me, felt like a gamble.

Decades of insights from socialist feminists tell us much of what we need to know about the resurgence of piecework. In the same way that domestic labor has long been invisible, despite its centrality to economic production, a growing amount of work in the gig economy has become, by design, unremunerated and hidden from regulation. And yet gig industry representatives deny or ignore the unpaid labor and automated anxiety experienced by contemporary pieceworkers like Janey, John, Ivri, Angeline, and Adil. Technology executives continue to argue that gig work is a dignified way for people to earn money in their spare time, enabling workers to fulfill other life obligations. This conceptualization obscures the ways in which piecework suppresses income and lowers labor standards across the board. Like garment piecework of the previous century, it also facilitates a deceptive cultural narrative that makes it appear possible to earn while simultaneously attending to unpaid domestic work.

In response to these conditions, gig workers share information—like the group text created by Uber drivers where they could warn each other about false surges. MTurk workers discuss the machinations of the platform in Facebook groups, Reddit forums, and even activist systems like Turkopticon. In some cities, Uber and Lyft drivers, many of whom are immigrants, women, and people of color, have formed and sustained collective worker groups. The budding politics of these organizations takes aim at the piecework model. Decrying black-box algorithms and information asymmetries that make their lives anxious and uncertain, many of these groups have demanded recognition of their status as employees and enforcement by states, which would guarantee a minimum wage and other protections. In California, ride-hail companies almost shut down in August, after a court ruling ordered them to recognize drivers as employees. A last-minute reprieve paused the change until October, when further arguments will be heard in the Court of Appeals. In London, the UK Supreme Court is considering a similar case, also brought by organizing workers. Acting in concert, gig workers are building collective power to reshape perspectives on why their time, labor, and lives are intrinsically valuable—and should be treated accordingly.

Veena Dubal is Professor of Law at the University of California Hastings College of the Law. Dubal’s research focuses on technology and work, and she is conducting a multi-year embedded ethnography of self-organizing Uber and Lyft drivers in California. Portions of this essay are reproduced, with permission, from “The Time Politics of Home-Based Digital Piecework” previously published in the C4E Journal.

Names have been changed to protect the identity of workers.