Silicon Valley’s Shadow: The Ghost Workers Behind Amazon, Google, and Microsoft

An invisible, on-demand workforce supports everything from Facebook to Uber and beyond with project-based tasks — and has little to show for it.

A woman with children is flanked by a rideshare driver on the left and a doctor taking notes on the right as she attends to supportive tasks on computer screen, with codes and rates of just a few cents hovering around her.
Julia Kuo

Standing on a street corner in Chicago, a young woman named Emily uses the Uber app to request a ride. A phone mounted on a nearby driver’s dashboard pings. The driver, whose name is Sam, accepts Emily’s request and begins navigating toward her location. Meanwhile, a third person who is located more than eight thousand miles away becomes secretly involved in their exchange. 

Ayesha, a woman in Hyderabad, South India, is tasked with validating Sam’s identity. She has to confirm that the selfie Sam has submitted to Uber as part of the company’s Real-Time ID Check process matches his photo on file. Sam had shaved off the beard that appears in his file photo, and the mismatch automatically prompts human intervention. A timer for the task counts down on Ayesha’s computer interface as she compares the photos. Just before Sam’s car pulls up to where Emily is waiting, Ayesha validates Sam’s photo, authorizing his account. Emily gets in the car.

According to anthropologist Mary L. Gray and computer scientist Siddharth Suri, the authors of Ghost Work: How to Stop Silicon Valley From Building a New Global Underclass, this sort of invisible real-time interaction occurs about once in the course of 100 of Uber’s pickups in the US, or approximately 13,000 times a day. (An Uber spokesperson noted that the authors’ sequence above is inaccurate; a driver would not have been allowed to accept any trips before their ID had been verified.) “Ghost workers” like Ayesha, whether carrying out this particular task or something entirely different, are laborers in an almost completely hidden economy that supports the operations of some of the world’s largest corporations and services, where human input is required to fulfill tasks for “rewards” of as little as one cent. They moderate comments and review flagged photos and videos on social media, caption clips, transcribe doctor’s notes, train search engines to produce more relevant results, tag images to teach algorithms to recognize their contents, and do much else besides.

“This is literally the reorganization of employment.”

The demand for such tasks is relentless. “New companies crop up every day with business models that depend on workers around the world who respond to open calls routed through software to do this behind-the-scenes work,” Gray and Suri write.

An economic class of ghost workers began to cohere in the early 2000s, after Amazon created a platform called Amazon Mechanical Turk (aka MTurk) through which workers could perform much-needed on-demand services for the company, such as correcting titles and descriptions in its enormous database of books, embedding listings with keywords, and selecting the right cover images. Amazon borrowed the interface’s name from a chess-playing automaton that caused a sensation in the 18th century and was later revealed to be a hoax — a chess master would hide inside the machine and control its actions with levers. MTurk reproduces this illusion of technological magic for the 21st century.

Just as Amazon’s consumer offerings expanded beyond book sales to nearly every shippable good, MTurk also grew enormously. Today, Amazon and many outside companies use the platform to enlist a huge pool of laborers to complete necessary tasks at all hours. Google and Microsoft — whose research arm employs Gray and Suri — have also developed their own versions of MTurk, adding to the number of interfaces where ghost workers complete on-demand tasks. Ghost workers can also be contracted by third-party vendors. The lack of visibility makes it hard to quantify the number of people who are engaged in this kind of labor; a Pew Research Center report estimates that in 2015 there were about 20 million people doing hidden on-demand work in the US alone.

Gray and Suri suggest that this portends the dismantling of full-time jobs and formerly office-based work into projects that are swiftly devalued. If this trend continues at a steady rate, they say that ghost work could account for roughly 60 percent of global employment by 2055.

“It’s not a niche job,” Gray told me. “This is literally the reorganization of employment.”

This is bad news because ghost work is monotonous at best, and disturbing, degrading, and destructive at worst. Take the field of content moderation, for example. The Cleaners, a documentary by the German filmmakers Moritz Riesewieck and Hans Block, explores the experiences of ghost workers in the Philippines who moderate flagged photos and videos on social media accounts. They describe the traumatizing effect of repeated exposure to gore, acts of terror, and child pornography. One content moderator worker in Manila says that he vets 25,000 images daily, and has been doing this work for six years. Another recalls seeing hundreds of videos of beheadings.

In Ghost Work, Gray and Suri’s interview subjects explain how they need to be hypervigilant in order to snap up the highest-paying tasks. Task completion can be time-limited to the second. By default, ghost workers are isolated from the things that can make difficult work bearable: solidarity with other colleagues, commiseration, skill sharing. There is also a lack of job security. If something goes wrong on a platform, by way of a glitch or a misunderstanding, ghost workers may find themselves locked out of their profiles or with lost income. “According to a national survey we conducted with Pew Research, 30 percent of on-demand gig workers reported not getting paid for work they performed,” note Gray and Suri.

“What we have here… is a cooking of the books to really dissimulate the amount of human labor and endeavor it takes to make these systems work.”

Ghost workers can also find themselves absorbing costs that might otherwise be covered by employers in traditional hourly jobs. Because they are not classified as employees of the companies they support, they don’t receive benefits. They aren’t paid for the time that they spend sorting through tasks to find the ones they want to take on. They provide their own computers and internet connections, in some cases devising virtual social support systems for their remote workplaces on their own. Minimum wage and overtime rules don’t apply to them, and their pay is chronically low. On MTurk, for example, it’s estimated that only 4 percent of workers earn more than $7.25 an hour.

The subtitle of Gray and Suri’s book indicates that it will show “how to stop Silicon Valley from building a new global underclass,” which might suggest an anti-capitalist vision for confronting ghost work. But Gray is not anti-capitalist; she sees ghost work’s potential reorganization of labor as an opportunity for society to reorient capitalism to more equitably serve workers.

“For me, there’s nothing inevitable about these being bad jobs,” she told me. “There’s so much that could be retrieved from the value that people can get from controlling their schedules, from being able to make choices about the kinds of projects they work on, to being able to make choices about who they work with.” 

In the last chapter of Ghost Work, the authors “have a few technical and social fixes to suggest for moving the future of work forward.” Gray and Suri propose creating on-demand co-working spaces and establishing a code of standards for ghost work. They suggest that a new employment classification should be created to accommodate the changing nature of work (“Let’s shift to recognizing that people work when they can, as they can”), though they don’t explain what this might look like. They also make a case against a universal basic income, suggesting that ghost workers be paid a retainer subsidized by companies and consumers. “Retainers, like those paid to lawyers and other professionals,” they write, “acknowledge the collective need to have a healthy, available, continuously updating workforce to make on-demand economies sustainable.” But they don’t explain what such a retainer would amount to, or how companies might be encouraged to adopt such a practice.

In fact, to truly “fix” ghost work, at least in most cases, would require completely changing its inherently exploitative nature. It deliberately isolates and obscures an underclass of workers who are valuable to the companies that employ them largely because they are invisible and disenfranchised.

Sarah T. Roberts, an associate professor of information studies who co-directs UCLA’s Center for Critical Internet Inquiry, has studied the working conditions of social media content moderators, many of whom suffer some of the most severe abuses within this sector. In her book Behind the Screen: Content Moderation in the Shadows of Social Media, Roberts writes that tech companies use subcontractors for difficult work like content moderation in order to shield themselves from responsibility for its deleterious effects on mental health, which would be difficult for them to do in the case of full-time employees. 

“Just as in the textile and manufacturing sectors (such as Apple contracting with Foxconn, or H&M with various contract textile manufacturers in Bangladesh), creating layers of contractors between the firm where the work originates and the ones that actually carry it out lends plausible deniability when there are repercussions for this work,” she writes.

Roberts notes that Facebook, a major employer of third-party content moderators, could have a profound impact if it were to hire its ghost workers directly. The company has the resources and the geographical reach to do so, and outsourcing its moderation work isn’t necessarily a money-saver. But subcontracting this essential labor makes Facebook and similar companies look leaner than they otherwise would, making them more appealing to investors. “What we have here, in my opinion, is a cooking of the books to really dissimulate the amount of human labor and endeavor it takes to make these systems work,” Roberts told me.

The marginalization of ghost workers is by design. As Roberts has noted, “The invisibility of [commercial content moderation] labour allows the public to imagine that social-media production is a painless, immaterial, and inhuman — rather than inhumane — process, and that any such curation practices that might occur happen only via algorithms and computational power.” Social media companies rely on ghost workers to remain hidden in order to keep the secret that humans are actually behind the screen, curating your experience.

The real fix for stopping Silicon Valley from building a new global underclass isn’t a set of piecemeal measures, but an approach that can account for these problems comprehensively.

Follow The Reboot

Join a growing community that’s examining the state of the internet and exploring its future. Subscribe to our newsletter.

A woman with children is flanked by a rideshare driver on the left and a doctor taking notes on the right as she attends to supportive tasks on computer screen, with codes and rates of just a few cents hovering around her.

Artwork By

Julia Kuo

Contact Us

Have an idea for a story or illustration? Interested in discussing partnerships? We want to hear from you. Send us a note at info(at)thereboot(dot)com.

Recommended Reading