ghost work

back to index

19 results

Ghost Work: How to Stop Silicon Valley From Building a New Global Underclass

by Mary L. Gray and Siddharth Suri  · 6 May 2019  · 346pp  · 97,330 words

by Wiley Periodicals, Inc. All rights reserved. hmhbooks.com Library of Congress Cataloging-in-Publication Data Names: Gray, Mary L., author. | Suri, Siddharth, author. Title: Ghost work : how to stop Silicon Valley from building a new global underclass / Mary L. Gray and Siddharth Suri. Description: Boston : Houghton Mifflin Harcourt, 2019. | Includes bibliographical

these two types of analysis, combining their strengths to shed more light on those who work in the on-demand economy. We examined four different ghost work platforms: Amazon.com’s Mechanical Turk (MTurk); Microsoft’s internal Universal Human Relevance System (UHRS); the socially minded startup LeadGenius; and Amara.org, a

that our observations and conclusions hold broadly across the on-demand economy, as opposed to being specific to one category of ghost work. MTurk, as one of the first commercially available ghost work platforms, set the norms for how others would apply human computation to business solutions. UHRS stands in for the internal

networks. This book describes the thoughtless processing of human effort through APIs as algorithmic cruelty—literally, computation incapable of thought, let alone empathy. People doing ghost work understand the perils and potential of on-demand work better than any engineer, tech company CEO, policy maker, or labor advocate. They live it every

profit from this mix of simple computer programming, a web interface, and unregulated hiring practices, producing a powerful new way to automate recruiting humans for ghost work. Ghost work fueled a revolution in artificial intelligence through millions of people carrying out billions of small tasks hidden from technology consumers. Other companies soon figured out

suggests, it takes human intelligence to push the boundary of what machines can learn. APIs are the perfect taskmasters for teaching machines to advance AI. Ghost Work, Machine Learning, and the Rise of AI Computer scientist Kevin P. Murphy defines machine learning as “a set of methods that can automatically detect

the mechanics of MTurk. Companies pushing technological innovation need legions of workers to beta-test products and check code. They also rely on people doing ghost work to improve their services’ algorithms and artificial intelligence by cleaning up training data from large stores of proprietary data. Tech companies collect and archive information

ensure it’s not peppered with adult content. Another popular task is translation. Microsoft’s strength in speech recognition and machine translation comes from the ghost work of people training algorithms with accurate data sets. They create them by listening to short audio recordings of one sentence in one language, typically

Craigslist for other copyediting jobs, she came across an ad for Amara. Amara is a translation and video-captioning service that blends the mechanisms of ghost work with automated features that manage the slicing and recombination of translated video content. Amara blurs the boundaries between repetitive micro-tasks and larger macro-tasks

seasoned technology strategist Aleli Alcala co-founded Amara On Demand (AOD) to fill this niche. Amara represents two realities folded into the growth of ghost work. First, ghost work’s simple existence and persistence belies claims that it’s ever easy to completely jettison humans from workflows that require creativity. Second, Amara speaks to

a nascent desire among some businesses to explicitly acknowledge that people, rather than software, are the more valuable component behind ghost work. Karen started out on Amara as a volunteer, subtitling YouTube videos, short documentaries, and college lectures for deaf and hard-of-hearing communities. The macro

clear time-sensitive or new projects from their desks. Commercial platforms like Upwork and a growing crowd of competitors take a hybrid approach to managing ghost work. They allow API access, which enables automated hiring, evaluation, and payment, not unlike MTurk. But they also allow individuals or companies to place tasks

Uber’s “selfie security,” discussed in the book’s introduction. The company has other large business clients, like eBay, Mozilla, Twitter, and Facebook. People doing ghost work on CrowdFlower complete micro-tasks like approving photos, customer support, and content moderation. In 2012, a CrowdFlower worker, Christopher Otey, filed suit against the company

employment” means and where that meaning comes from. They also help explain why, when industries fixate on automating jobs away, they paradoxically spike demand for ghost work, shredding the social contract between employer and worker in their wake. BEFORE WAGNER: EXPENDABLE LABOR FUELING EARLY CAPITALISM The Wagner Act was not the first

seen as valuable employees to cultivate. They were considered interchangeable, adding little to no skill, expected to leave when their projects ended. Early forms of ghost work continued to flourish throughout the 20th century. Fast-forward to the 1980s, when temporary staffing agencies like Kelly Girl Service and Manpower contracted out more

staffing service calls suggest, technological advancement has always depended on expendable, temporary labor pools. Tracing the continuities rather than radical breaks from this past put ghost work in context. As history illustrates, hiring people on the assumption that they will be around only for the duration of a finite project or that

“employers of record” primed the so-called platform economy. Employment practices that convert workers into something “procured” to launch a product are baked into ghost work’s code. Over the past decade, companies launched on-demand services that matched individual consumers or other businesses to everything from a ride to the

smartphone app. They sell themselves as hip and innovative technologies, not sites of employment or temp agencies. But they and the scores of on-demand ghost work platforms like them, intentionally or not, are quietly taking over the $115 billion temporary staffing industry. The platforms detailed in this book, offering direct

worker is necessary in the moment but then expendable once the project ships. Part II Demanding Work 3 Algorithmic Cruelty and the Hidden Costs of Ghost Work Indifferent Design and Its Unintended Consequences Most on-demand workers accept an uneven workflow as part of the API landscape, but software bugs can

essence, he was the first economist to theorize how to produce at scale through modern private enterprise and profit from a well-oiled org chart. Ghost work economies sell themselves as software that can eliminate the expensive frictions of searching, matching, training, communicating with, and retaining workers. Yet, as Coase might

and coordination among workers, and between workers and their employers, not only is necessary but is actually money well spent. For all the claims that ghost work can combine algorithms, artificial intelligence, and platform interfaces to replace the company’s function as “the entrepreneur-coordinator, who directs production,”3 there is evidence

no opportunity to appeal any mistreatment. Many of the transaction costs passed on to requesters mirror those shouldered by workers. Each hurdle faced demonstrates that ghost work isn’t working smoothly for anyone involved. REQUESTING WORK: TRANSACTION COSTS The most commonly reported requester difficulty occurred in the process of matching a worker

on the project meeting the client’s expectations.” On the other side of the market, it’s doubtful that workers would rate ghost work “an 8 out of 10.” Ghost Work’s Hidden Pain Scale The burden of transaction costs used to fall on companies, but now it falls squarely on the people doing

. Requesters are seen as visible and valuable customers to the platform, giving them latitude to change a job request midstream without censure. Those performing ghost work are mostly invisible and treated as interchangeable. They must scramble to make do with a system designed to assume that workers could be adversaries trying

platform design. People who design on-demand job sites assume workers have fast broadband connections and reliable power sources. In reality, millions of people doing ghost work do their jobs with outdated computers, faulty internet access, and even shared IP addresses. In the chasm between the platform designers’ mental image of

19 The specific percentages of distribution vary, in many ways depending on the platform’s approach to retaining workers, but a Pareto distribution exists nonetheless. Ghost work currently organizes around a small percentage of people who turn project-driven tasks into full-time work. A slightly larger portion of people consistently contribute

obligations and the workload of more formal employment. Both views are equally valid reads of the situation. For some people we interviewed, particularly women, ghost work legitimized their contributions and gave them a way to feel valued. Women are not unaware that their options are limited in formal employment, at least

something more substantive and fulfilling, when the right mixture of workers’ needs and market demands are properly aligned and matched. It can rapidly transmogrify into ghost work when left unchecked or hidden behind software rather than recognized as a rapidly growing world of global employment. Technologies, in and of themselves, are not

government agency is tasked with changing that fact. The other challenge is that, with so much variation in workers’ schedules and commitments, the “flexibility” of ghost work’s Pareto distribution means that people don’t share a work site, hours, or a professional identity—three key ingredients to organizing workers’ interests. Lastly

works against their earning potential. The kindness and collaborations of workers helping one another are, arguably, the digital economy’s most valuable ingredient. Yet most ghost work platforms seem determined to delete it. Purposefully or not, creators of on-demand labor platforms take aim at workers’ social worlds by presuming that workers

labor as the valuable engine behind their services. But it’s businesses designed around workers’ contributions, prioritizing their schedules, project interests, and collaborations, that convert ghost work into a sustainable enterprise. Though these businesses strive to “do well by doing good,” they have learned, through close calls that nearly shuttered their startups

to society as well as to shareholders. In the on-demand marketplace, B Corps like CloudFactory push back against the prevailing notion that people doing ghost work are expendable. They prioritize workers’ schedules, interests, and collaborations. Ultimately, these platforms show how worker-focused design can improve the quality of work produced

and democratic collective decision-making in a Pareto distribution where people come and go. It turns out that people carry different mental maps to their ghost work. Nothing illustrates this more clearly than a realization that we had after completing our research and looking back over interview transcripts. We had asked

“gig economy”—an ecosystem of independent contractors and small businesses driven by short-term projects that shift to meet market demands—is quietly moving to ghost work platforms. A growing number of people are picking up on-demand gig work online — accepting project-driven tasks, with companies that assign, schedule, route,

, the survey questions did not reference the internet. No one asked workers how digital technology might be, in part, managing their labor. Because people doing ghost work might show up in the survey under various headings, including “alternative work arrangements,” “temporary staffing,” and “self-employed,” there is no easy way to

a paycheck. But humans are social animals who derive dignity and meaning from creating things together and developing professional relationships. Trying to prevent those doing ghost work from connecting is not only impossible but also misses the opportunity to capitalize on the value workers derive from connecting with one another. Our research

and break up their workday, root each other on, share information about potential work opportunities, and mentor one another in how to succeed in ghost work. We propose a professional network to build guild-like communities for on-demand workers. These guild communities could foster and facilitate communication among those doing

SHARED WORKSPACE The ability to contract on-demand labor, programmatically, will change knowledge work as we know it. Companies are increasingly turning to on-demand ghost work platforms for deadline-driven projects, and this creates a continuous ebb and flow of workers across enterprise boundaries. Enterprises could assemble and equip combinations of

not whether the transaction costs disappear but, rather, who pays for them? We need large corporations, which use vendor management systems to hire ghost work in bulk, to pull ghost work out of the shadows and into the daylight. As noted above, recent economic analyses of contingent labor markets confirm a sharp increase in

technical and social fixes detailed above are not so easily divided among technologists, labor activists, and policy makers. They will require a collective effort. Bringing ghost work out of the shadows will require business owners, policy makers, customers, and citizens coming together. It will take our collective will to reorient employment toward

quantitative analyses conducted. Below, we provide readers with details of how we gathered the ethnographic and interview materials used to illustrate the lived experiences of ghost work featured in this book. ETHNOGRAPHIC FIELDWORK AND INTERVIEWS This book draws on collaborative ethnographic fieldwork in the U.S. and in India, led by Mary

reduce the need for labor,” he says. [back] 46. According to research and advisory firm ISG. [back] 3. Algorithmic Cruelty and the Hidden Costs of Ghost Work 1. Eric Meyer, “Inadvertent Algorithmic Cruelty,” Meyerweb (blog), December 24, 2014, https://meyerweb.com/eric/thoughts/2014/12/24/inadvertent-algorithmic-cruelty/; revised version published

who are usually subjected to just-in-time schedules. The recent boom of scheduling and management software businesses, like Celayix, Smartsheet, and Shiftboard, that repurpose ghost work’s tool kit of APIs and websites offer so-called predictive scheduling, suggesting that the practice of automating employees’ schedules is far from dying out

, 21–22, 87 LeadGenius and, 148–49, 224 n27 map of MTurk participants, figure 1B MTurk payment, 15 outsourcing growth in, 224 n24 reasons for ghost work, 96 shared workspaces, 180–81 underemployment, 95 women and, 106–10, 116–17 Industrial Revolution, 40–42 inequality. See discrimination; double bottom line; women

Collaborative Society

by Dariusz Jemielniak and Aleksandra Przegalinska  · 18 Feb 2020  · 187pp  · 50,083 words

racialized.33 Silicon Valley is using technology to build a new global underclass, forced to perform what Mary L. Gray and Siddharth Suri aptly call “ghost work,” dehumanized and invisible labor streamlined by apps and intermediary platforms.34 Figure 1 A Google Trends graph measures interest in the sharing economy (in black

of Low-Income Service work in the ‘On-Demand’ Economy,” Information, Communication and Society 20 (2017): 898–914. 34. M. L. Gray and S. Suri, Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass (Houghton Mifflin Harcourt, 2019). 35. N. Srnicek, Platform Capitalism (John Wiley & Sons, 2017). 36

Unbound: Selected Writings on Open Access, 2002–2011. Cambridge, MA: MIT Press, 2016. More about digital platforms and sustainability Gray, M. L., and S. Suri. Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass. Boston: Houghton Mifflin Harcourt, 2019. Rainie, L., and B. Wellman. Networked: The New Social

Wonderland: How Play Made the Modern World

by Steven Johnson  · 15 Nov 2016  · 322pp  · 88,197 words

Mechanical and Pictorial Exhibition” called the Akolouthorama; a predecessor to Philipsthal’s spook show called the Phantascopia; an exhibition called the Spectrographia, which promised “TRADITIONARY GHOST WORK!”; an influential mechanical exhibition dubbed the Eidophusikon; the Panstereomachia, “a picto-mechanical representation,” according to the Times. A virtual orchestra created by a painter and

History and Biography 84:1 (1960), 79. “The city was before us”: Ibid. “A musket machine”: Ibid. “Novel Mechanical and Pictorial Exhibition”: Altick, 131. “TRADITIONARY GHOST WORK!”: Ibid. “You are allowed to look”: Altick, 231. “no one thought of clothing”: Neal Gabler, Walt Disney (New York: Knopf Doubleday, 2006, Kindle edition), Kindle

Gilded Rage: Elon Musk and the Radicalization of Silicon Valley

by Jacob Silverman  · 9 Oct 2025  · 312pp  · 103,645 words

sit and wait for orders. Maybe write an op-ed or two. Whatever I would do in the job, I imagined I would be a ghost, working to advance some yet-unspoken goal of decision-makers far above. It was deep into the dinner hour, but a smattering of young employees in

Let's Pretend This Never Happened (A Mostly True Memoir)

by Jenny Lawson  · 5 Mar 2013  · 308pp  · 98,022 words

. This afternoon I sauntered into Victor’s office and said smugly, “So, apparently my ‘craazy’ plan for setting off the fire alarm to appease the ghosts worked, because guess who just found the dead bodies I’ve been searching for? ME, MOTHERFUCKER. I found the dead bodies.” Then I held up my

Accelerando

by Stross, Charles  · 22 Jan 2005  · 489pp  · 148,885 words

lying to yourself," Rita sends back. "You're lying about your own goals and motivations. You may not want to know the truth your own ghost worked out, but I do. And I'm not going to let you deny it happened." "So one of your agents seduced a personality image of

Futureproof: 9 Rules for Humans in the Age of Automation

by Kevin Roose  · 9 Mar 2021  · 208pp  · 57,602 words

boring, repetitive jobs, many of which we don’t see in the West. Mary L. Gray and Siddharth Suri have written about the rise of “ghost work,” a phenomenon in which human labor, carefully concealed from the end user, is deployed to make AI and automated systems function properly. Social networks like

: Social Meaning of a New Technology (Cambridge, Mass.: MIT Press, 1990). Mary L. Gray and Siddharth Suri have written Mary L. Gray and Siddharth Suri, Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass (New York: Houghton Mifflin Harcourt, 2019). In China, “data labeling” companies Li Yuan, “How

Your Computer Is on Fire

by Thomas S. Mullaney, Benjamin Peters, Mar Hicks and Kavita Philip  · 9 Mar 2021  · 661pp  · 156,009 words

from Amazon.cult,” Seattle Weekly (October 9, 2006), https://www.seattleweekly.com/news/how-i-escaped-from-amazon-cult/. 29. Mary Gray and Siddharth Suri, Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass (Boston: Houghton Mifflin Harcourt, 2019); Sarah Roberts, Behind the Screen: Content Moderation in the

The Costs of Connection: How Data Is Colonizing Human Life and Appropriating It for Capitalism

by Nick Couldry and Ulises A. Mejias  · 19 Aug 2019  · 458pp  · 116,832 words

code unstructured data, are generally treated as below the line and so as not affecting the proxy’s validity for knowledge. See Gray and Suri, Ghost Work. 69. Brayne, “Big Data Surveillance,” 20. 70. Rouvroy, “End(s) of Critique,” 143. 71. Grosser, “What Do Metrics Want?” 72. From the Mindshare media and

Approach to Recommendations.” Netflix (blog), February 17, 2016. https://media.netflix.com/en/company-blog/a-global-approach-to-recommendations. Gray, Mary, and Siddharth Suri. Ghost Work. New York: Houghton Mifflin Harcourt, 2019. Greenberg, Andy. “Apple’s Latest Selling Point: How Little It Knows about You.” Wired, June 8, 2015. https://www

The Gig Economy: A Critical Introduction

by Jamie Woodcock and Mark Graham  · 17 Jan 2020  · 207pp  · 59,298 words

, M.A. and Friederici, N. (2017c) Digital connectivity and African knowledge economies. Questions de Communication, 32: 345–60. Gray, M.L. and Suri, S. (2019) Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass. New York: Houghton Mifflin Harcourt. Gray, M.L., Suri, S., Ali, S.S. and

Internet for the People: The Fight for Our Digital Future

by Ben Tarnoff  · 13 Jun 2022  · 234pp  · 67,589 words

Democracy's Data: The Hidden Stories in the U.S. Census and How to Read Them

by Dan Bouk  · 22 Aug 2022  · 424pp  · 123,180 words

Don't Be Evil: How Big Tech Betrayed Its Founding Principles--And All of US

by Rana Foroohar  · 5 Nov 2019  · 380pp  · 109,724 words

Code Dependent: Living in the Shadow of AI

by Madhumita Murgia  · 20 Mar 2024  · 336pp  · 91,806 words

AI in Museums: Reflections, Perspectives and Applications

by Sonja Thiel and Johannes C. Bernhardt  · 31 Dec 2023  · 321pp  · 113,564 words

The Lonely Century: How Isolation Imperils Our Future

by Noreena Hertz  · 13 May 2020  · 506pp  · 133,134 words

After the Gig: How the Sharing Economy Got Hijacked and How to Win It Back

by Juliet Schor, William Attwood-Charles and Mehmet Cansoy  · 15 Mar 2020  · 296pp  · 83,254 words

Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI

by Karen Hao  · 19 May 2025  · 660pp  · 179,531 words

Blood in the Machine: The Origins of the Rebellion Against Big Tech

by Brian Merchant  · 25 Sep 2023  · 524pp  · 154,652 words