What do you think?
Rate this book
Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass – An Urgent Investigation into the Invisible Human Labor Powering AI and the Digital Economy
In the spirit of Nickel and Dimed, a necessary and revelatory expose of the invisible human workforce that powers the web—and that foreshadows the true future of work.Hidden beneath the surface of the web, lost in our wrong-headed debates about AI, a new menace is looming. Anthropologist Mary L. Gray and computer scientist Siddharth Suri team up to unveil how services delivered by companies like Amazon, Google, Microsoft, and Uber can only function smoothly thanks to the judgment and experience of a vast, invisible human labor force. These people doing "ghost work" make the internet seem smart. They perform high-tech flagging X-rated content, proofreading, designing engine parts, and much more. An estimated 8 percent of Americans have worked at least once in this “ghost economy,” and that number is growing. They usually earn less than legal minimums for traditional work, they have no health benefits, and they can be fired at any time for any reason, or none. There are no labor laws to govern this kind of work, and these latter-day assembly lines draw in—and all too often overwork and underpay—a surprisingly diverse range of harried young mothers, professionals forced into early retirement, recent grads who can’t get a toehold on the traditional employment ladder, and minorities shut out of the jobs they want. Gray and Suri also show how ghost workers, employers, and society at large can ensure that this new kind of work creates opportunity—rather than misery—for those who do it.
288 pages, Kindle Edition
First published January 1, 2019
128 people are currently reading
3163 people want to read
About the author
Mary L. Gray
8 books18 followersMary L. Gray is a Professor of Communication and Culture at the University of Indiana.
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 30 of 74 reviews
August 5, 2019
It's decent as a survey of ghost work, featuring stories from workers around the world.
As a critique of labour practices in Silicon Valley, it's a little disappointing - more like a business book than the "startling exposé" it's claimed to be (Tim O'Reilly and Martin Ford are listed among the blurbs on the back). Accordingly, the proposed solutions (like universal healthcare, or a different employment model) are all framed in pro-business terms, rather than in terms of workers' needs, which is already ceding ground to capital. Useful as an introduction to the problem of labour exploitation as mediated by digital technology, but I found the analysis to be wanting and the conclusion to be timid.
As a critique of labour practices in Silicon Valley, it's a little disappointing - more like a business book than the "startling exposé" it's claimed to be (Tim O'Reilly and Martin Ford are listed among the blurbs on the back). Accordingly, the proposed solutions (like universal healthcare, or a different employment model) are all framed in pro-business terms, rather than in terms of workers' needs, which is already ceding ground to capital. Useful as an introduction to the problem of labour exploitation as mediated by digital technology, but I found the analysis to be wanting and the conclusion to be timid.
May 1, 2019
Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass by Mary L. Gray and Siddharth Suri addresses a new kind of worker: one who bridges the gap between what AI can and can not do. As Gray and Suri note, “the great paradox of automation is that the desire to eliminate human labor always generates new tasks for humans.” This book is about the gray area between the robots taking over and humans.
These workers are the ones that decide if a picture that was flagged is obscene or not - is that a thumb, or something else? - after AI runs it through it’s algorithm and can’t distinguish between the two. They’re the people who have to decide if something constitutes hate speech or is a right-wing politician on a rant. They decide if a person posing as an Uber driver is really a registered Uber driver, etc. Generally their jobs are controlled by an API that spits out micro-jobs as fast as they come in, and workers clamor for this “piece work” not unlike the piece-work of the early industrial era.
Ghost work is the bridge between AI an an automated future; and as Gray and Suri attest, these jobs complicate the the dominant story of humans being replaced by robots. Yet, these workers are generally invisible, poorly paid and have few protections. (Of course, as they point out, “according to the U.S. Department of Labor’s Bureau of Labor Statistics, only 52 percent of today’s employers sponsor workplace benefits of any kind.”)
The book represents five years of anthropological study of workers in the US and India that use four “ghost work” platforms to make a living: Amazon’s Mechanical Turk (MTurk); LeadGenius; Microsoft’s lUniversal Human Relevance System (UHRS); and Amara.org.
There are a lot of stories about these workers and the opportunities and hardships this sort of piece work represents for them. For one woman in a rural village in India, whose takehome pay of $350 a month makes her the richest person in the village, ghost work is a fantastic alternative to other local options, but that doesn’t mean it is fulfilling.
As Gray and Suri note, this field of work is growing rapidly and is only slated to grow more in the future, and so it is important to understand how such ghost work could effect society and people on a larger scale. To this point they offer a number of potential ways to mitigate the experience for people doing these jobs (Step 1: Employers, realize there are humans working at the other end of the API).
This book has an excellent history of piece work and temp work, leading up to the rise of the gig economy and ghost work in general. If you’ve ever been curious why tech has built it’s foundation on contract labor, this will give you the historical context. It’s a bit wordy at times and some of the concepts are repeated more times than are necessary. But it is an interesting look at a class of workers few people realize are even there.
Read my extended review of Ghost Work here.
These workers are the ones that decide if a picture that was flagged is obscene or not - is that a thumb, or something else? - after AI runs it through it’s algorithm and can’t distinguish between the two. They’re the people who have to decide if something constitutes hate speech or is a right-wing politician on a rant. They decide if a person posing as an Uber driver is really a registered Uber driver, etc. Generally their jobs are controlled by an API that spits out micro-jobs as fast as they come in, and workers clamor for this “piece work” not unlike the piece-work of the early industrial era.
Ghost work is the bridge between AI an an automated future; and as Gray and Suri attest, these jobs complicate the the dominant story of humans being replaced by robots. Yet, these workers are generally invisible, poorly paid and have few protections. (Of course, as they point out, “according to the U.S. Department of Labor’s Bureau of Labor Statistics, only 52 percent of today’s employers sponsor workplace benefits of any kind.”)
The book represents five years of anthropological study of workers in the US and India that use four “ghost work” platforms to make a living: Amazon’s Mechanical Turk (MTurk); LeadGenius; Microsoft’s lUniversal Human Relevance System (UHRS); and Amara.org.
There are a lot of stories about these workers and the opportunities and hardships this sort of piece work represents for them. For one woman in a rural village in India, whose takehome pay of $350 a month makes her the richest person in the village, ghost work is a fantastic alternative to other local options, but that doesn’t mean it is fulfilling.
As Gray and Suri note, this field of work is growing rapidly and is only slated to grow more in the future, and so it is important to understand how such ghost work could effect society and people on a larger scale. To this point they offer a number of potential ways to mitigate the experience for people doing these jobs (Step 1: Employers, realize there are humans working at the other end of the API).
This book has an excellent history of piece work and temp work, leading up to the rise of the gig economy and ghost work in general. If you’ve ever been curious why tech has built it’s foundation on contract labor, this will give you the historical context. It’s a bit wordy at times and some of the concepts are repeated more times than are necessary. But it is an interesting look at a class of workers few people realize are even there.
Read my extended review of Ghost Work here.
April 23, 2019
Ghost workers are on-demand, disposable people who work behind the curtain to ensure the internet lives up to its promise. In Ghost Work, Mary Gray and Siddharth Suri have gone behind the curtain themselves, gathering data and interviewing the people who do the work. It is the first to penetrate this domain, which clearly needs more such efforts.
In the world of ghost work, jobs last for seconds, not years. Workers must spend far more hours searching for quick gigs than actually performing them. And despite the divide-and-conquer nature of everyone working on their own, ghost workers have found ways to communicate, link up and share knowledge of new gigs that people working in isolation might not see. The network effect is at work as much as the Pareto Effect, or 80/20 rule as it is better known. Teamwork pays, despite the dictates of the algorithm in charge. But even as the few stars grab most of the revenue on offer, it doesn’t amount to a living. They are the hamsters in the wheels that run the internet.
The workers are all homeworkers. They get no offices, uniforms, computers, software, training, supervision, encouragement or praise. There are no bonuses, vacations, promotions or awards. They sit in front of their own screens, madly searching for mini-gigs that pay a cent or two each and can, no must be completed in seconds. They are nameless. The platform they log onto gives them a coded string of letters and numbers as their ID. They have strict rules of performance, as measured by the time it takes them to complete tasks, and how seriously they fall afoul of the rules, like using a different IP address, or working alongside someone else, or using someone’s bank account (because there are none locally).
On the other side are the requestors. Companies need work done quickly and cheaply. They put out a request through the broker platform and the islaves grab them before someone else does, a feeding frenzy. If the requestor doesn’t like the results, s/he can claim they were unacceptable, and the islave doesn’t get paid. And their reputation gets dinged. The authors found that islaves lost 30% of their expected and contracted remuneration. Often, the requestor wrote the request badly, and got what s/he asked for, but it wasn’t what was desired. Garbage in, garbage out – and no pay.
The work is usually really basic. Captioning, tagging, translating, classifying, categorizing and grouping are typical examples. Some firms do a little more, like finding everyone in a city who has been convicted of a crime, so lawyers can cold call them to sell them reputation restoration services. Uber uses them to verify that drivers’ selfies when they log on are the same folks in the profile.
The conceit here is that even artificial intelligence isn’t up to a lot of tasks, so the internet actually works with millions of people doing really short tasks more cheaply and faster than AI can. Today. There are hundreds of companies, brokering tens of millions of islaves doing their bidding around the clock and the world.
Their pay is pathetic. One American woman who excels at it says she can make as much as $40 for a ten hour day. A young woman in southern India has the highest income of anyone in her village: $350 a month. The money gets transferred to the islave’s bank account – for a $2 fee. There are no benefits, no taxes withheld, and of course the workers have no rights. They can be terminated at any time, and their accumulated balance forefeited, for any reason and no reason, without explanation or recourse. There are no humans to appeal to, no HR department to set people straight, no payroll department to correct seeming errors. There is no directory, no contact information.
Ironically, these isolated workers are finding each other. They are creating forums, direct messaging each other, sharing leads, sharing strategies, tactics, tips and stories. They are creating the work environment their “employer” denies them. But every minute spent on chat is a minute not paid. It is a brutal, unfulfilling life.
The authors say 60% of all work will be on-demand by 2055 at the rate we are going. They cite the paradox of automation, which demonstrates that for every automated process, the need for human intervention actually increases. It just doesn’t pay a living wage. They call for stronger unions, a registry where islaves’ work records can help maintain or restore their reputations when the algorithms they work for cut them off for, say, not being logged into the system long enough or often enough. The sweatshop of the 21st century is the comfort of your own home. Laws need to catch up.
There are, of course, lots of people who benefit from on-demand work. Those caring for a loved one, the disabled, the housebound, the shy, the insecure, the unemployed, those unable to work with others, those with few skills or no experience. The list is endless.
Ironically again, the authors did not use ghost work for their research. They actually visited and studied the ghost workers in the USA and India to inform their book. Ghost Work points to many little companies that seem to care. They try to improve life for their on-demand workers. But they are few and far between. And they can’t get too close to the workers, or the law would consider them employees with rights. Mostly, the companies are huge beasts like Amazon’s Mechanical Turk, totally anonymous, impersonal, uncaring and driving for ever bigger tasks for ever less pay.
For all the pain and difficulties it describes, the book is remarkably positive about it all. Ghost Work is a glass half full book.
David Wineberg
In the world of ghost work, jobs last for seconds, not years. Workers must spend far more hours searching for quick gigs than actually performing them. And despite the divide-and-conquer nature of everyone working on their own, ghost workers have found ways to communicate, link up and share knowledge of new gigs that people working in isolation might not see. The network effect is at work as much as the Pareto Effect, or 80/20 rule as it is better known. Teamwork pays, despite the dictates of the algorithm in charge. But even as the few stars grab most of the revenue on offer, it doesn’t amount to a living. They are the hamsters in the wheels that run the internet.
The workers are all homeworkers. They get no offices, uniforms, computers, software, training, supervision, encouragement or praise. There are no bonuses, vacations, promotions or awards. They sit in front of their own screens, madly searching for mini-gigs that pay a cent or two each and can, no must be completed in seconds. They are nameless. The platform they log onto gives them a coded string of letters and numbers as their ID. They have strict rules of performance, as measured by the time it takes them to complete tasks, and how seriously they fall afoul of the rules, like using a different IP address, or working alongside someone else, or using someone’s bank account (because there are none locally).
On the other side are the requestors. Companies need work done quickly and cheaply. They put out a request through the broker platform and the islaves grab them before someone else does, a feeding frenzy. If the requestor doesn’t like the results, s/he can claim they were unacceptable, and the islave doesn’t get paid. And their reputation gets dinged. The authors found that islaves lost 30% of their expected and contracted remuneration. Often, the requestor wrote the request badly, and got what s/he asked for, but it wasn’t what was desired. Garbage in, garbage out – and no pay.
The work is usually really basic. Captioning, tagging, translating, classifying, categorizing and grouping are typical examples. Some firms do a little more, like finding everyone in a city who has been convicted of a crime, so lawyers can cold call them to sell them reputation restoration services. Uber uses them to verify that drivers’ selfies when they log on are the same folks in the profile.
The conceit here is that even artificial intelligence isn’t up to a lot of tasks, so the internet actually works with millions of people doing really short tasks more cheaply and faster than AI can. Today. There are hundreds of companies, brokering tens of millions of islaves doing their bidding around the clock and the world.
Their pay is pathetic. One American woman who excels at it says she can make as much as $40 for a ten hour day. A young woman in southern India has the highest income of anyone in her village: $350 a month. The money gets transferred to the islave’s bank account – for a $2 fee. There are no benefits, no taxes withheld, and of course the workers have no rights. They can be terminated at any time, and their accumulated balance forefeited, for any reason and no reason, without explanation or recourse. There are no humans to appeal to, no HR department to set people straight, no payroll department to correct seeming errors. There is no directory, no contact information.
Ironically, these isolated workers are finding each other. They are creating forums, direct messaging each other, sharing leads, sharing strategies, tactics, tips and stories. They are creating the work environment their “employer” denies them. But every minute spent on chat is a minute not paid. It is a brutal, unfulfilling life.
The authors say 60% of all work will be on-demand by 2055 at the rate we are going. They cite the paradox of automation, which demonstrates that for every automated process, the need for human intervention actually increases. It just doesn’t pay a living wage. They call for stronger unions, a registry where islaves’ work records can help maintain or restore their reputations when the algorithms they work for cut them off for, say, not being logged into the system long enough or often enough. The sweatshop of the 21st century is the comfort of your own home. Laws need to catch up.
There are, of course, lots of people who benefit from on-demand work. Those caring for a loved one, the disabled, the housebound, the shy, the insecure, the unemployed, those unable to work with others, those with few skills or no experience. The list is endless.
Ironically again, the authors did not use ghost work for their research. They actually visited and studied the ghost workers in the USA and India to inform their book. Ghost Work points to many little companies that seem to care. They try to improve life for their on-demand workers. But they are few and far between. And they can’t get too close to the workers, or the law would consider them employees with rights. Mostly, the companies are huge beasts like Amazon’s Mechanical Turk, totally anonymous, impersonal, uncaring and driving for ever bigger tasks for ever less pay.
For all the pain and difficulties it describes, the book is remarkably positive about it all. Ghost Work is a glass half full book.
David Wineberg
May 11, 2019
If you are expecting a deep economical and political analysis of the gig economy or "ghost work", you will be disappointed.
The book does provides an interesting field research focusing on workers on platforms that focus on piece work requested through APIs, like Amazon's Mechanical Turk, which is a less visible workforce than Uber drivers, and that alone is the value of the book, but that's kind of it.
There is no in depth economical analysis, no political considerations besides a very bland mainstream neoliberal veneer, and an approach that I would classify us naive.
They point to some of the platform issues, like work isolation, lack of visibility of the whole job, no incentives to work in groups or the off loading of transaction costs to workers as "bugs in the system", "algorithmic cruelty" inflicted by accident by lack of better planning, and what they are: features of the system. Bezos and company don't want workers to feel that they are part of a group, to organize, to be paid good wages, to have a decent "work life balance". They want them atomized, alienated from work, paying bottom dollar and demanding them to show up for their software slave drivers. This is the real reason for these systems to exist.
The last chapter is dedicated to suggestions on how we can improve these workers lives, and I do believe they mean well, but they are just band-aids for an amputated arm. All criticism to the core of the system are superficial and bland, they just assume that this is the future of work and we might as well adapt, and provide a few nice ideas to make the bondage a little bit less bad.
I was extremely disappointed, I remember getting a reference to this book from a podcast, I don't remember which, but it's probably from a tame centrist one like Erza Klein's. If you lean to the left at least a little, you will likely come out of this reading fuming like me. Well meaning people giving improvement tips for the next startup to exploit workers.
The book does provides an interesting field research focusing on workers on platforms that focus on piece work requested through APIs, like Amazon's Mechanical Turk, which is a less visible workforce than Uber drivers, and that alone is the value of the book, but that's kind of it.
There is no in depth economical analysis, no political considerations besides a very bland mainstream neoliberal veneer, and an approach that I would classify us naive.
They point to some of the platform issues, like work isolation, lack of visibility of the whole job, no incentives to work in groups or the off loading of transaction costs to workers as "bugs in the system", "algorithmic cruelty" inflicted by accident by lack of better planning, and what they are: features of the system. Bezos and company don't want workers to feel that they are part of a group, to organize, to be paid good wages, to have a decent "work life balance". They want them atomized, alienated from work, paying bottom dollar and demanding them to show up for their software slave drivers. This is the real reason for these systems to exist.
The last chapter is dedicated to suggestions on how we can improve these workers lives, and I do believe they mean well, but they are just band-aids for an amputated arm. All criticism to the core of the system are superficial and bland, they just assume that this is the future of work and we might as well adapt, and provide a few nice ideas to make the bondage a little bit less bad.
I was extremely disappointed, I remember getting a reference to this book from a podcast, I don't remember which, but it's probably from a tame centrist one like Erza Klein's. If you lean to the left at least a little, you will likely come out of this reading fuming like me. Well meaning people giving improvement tips for the next startup to exploit workers.
November 4, 2024
This book argues that automation vs human labor is a false dichotomy, and shows how AI has gone on to create jobs for a labor force which it calls ‘ghost workers’. They are the invisible assembly line picking up the pieces where technology fails. A completely unregulated industry.
The book is an anthropological study which paints a compelling picture of the humans doing ghost work. Also particularly relevant to technologists who are building algorithms and APIs reliant on human taskers.
Where the book lacks is that it doesn’t go far enough in laying bare the point that the core issue with ghost work is global inequality. And doesn’t propose policy-level suggestions, instead laying all the responsibility for action in the hands of technologists.
The book is an anthropological study which paints a compelling picture of the humans doing ghost work. Also particularly relevant to technologists who are building algorithms and APIs reliant on human taskers.
Where the book lacks is that it doesn’t go far enough in laying bare the point that the core issue with ghost work is global inequality. And doesn’t propose policy-level suggestions, instead laying all the responsibility for action in the hands of technologists.
September 30, 2021
It read more like an extended research paper than a book. I learned more about on-demand labor, and the author provided some compelling solutions to the problems outlined, but nothing stuck out to me as the main takeaway for me as a technologist to understand about “ghost work”: a term that was broadly used for any and all kinds of gig work.
April 11, 2019
This was a very interesting look into the use of platforms that distribute work into micro and macrotasks which can be completed by individual contractors instead of having full-time employees. The authors worked hard to avoid taking any particular sides on the existence of this new work type, providing benefits and issues. While these platforms have opened up opportunities for those who may need to work from home or are in developing nations, they don't offer the same types employment benefits that people expect from an office job like healthcare and regular communication with co-workers.
I'm more skeptical than the authors that this is the future of work. With only an isolated view of your work, people are more likely to be replaced by automation as they will be unable to draw connections between disparate components and communicate with other teams. It is also important that this class of individuals receives some sort of regulatory protection to shift the balance more towards workers than the mediating platforms.
Still, this book offers a critical view and does a good job of exploring the topic thoroughly.
I'm more skeptical than the authors that this is the future of work. With only an isolated view of your work, people are more likely to be replaced by automation as they will be unable to draw connections between disparate components and communicate with other teams. It is also important that this class of individuals receives some sort of regulatory protection to shift the balance more towards workers than the mediating platforms.
Still, this book offers a critical view and does a good job of exploring the topic thoroughly.
May 23, 2023
Devoured this book (an extended research report really) in a few days. Loved reading other people’s serious investigation and intellectual framework about a topic that I’ve been working on for more than a year now. Makes you kinda appreciate the value of academia paying some really fortunate people like me to think deep and hard and long about problems that corporations don’t need to or don’t have time to think about.
Standouts are the interviews with on-demand workers who enjoy the arrangement, the framing of human in the loop who are often invisible to the consumers using the software as a service product, the history linking today’s on-demand workers with contingent workforce (women and minority) from half a century ago, the platforms’ reluctance to act in any way that may make them appear in any way like an employer of record (labor classification laws need to change)
The ending is so unnecessary, kinda a lesson in what you shouldn’t do at the end of a paper
Standouts are the interviews with on-demand workers who enjoy the arrangement, the framing of human in the loop who are often invisible to the consumers using the software as a service product, the history linking today’s on-demand workers with contingent workforce (women and minority) from half a century ago, the platforms’ reluctance to act in any way that may make them appear in any way like an employer of record (labor classification laws need to change)
The ending is so unnecessary, kinda a lesson in what you shouldn’t do at the end of a paper
October 17, 2024
2.5 Stars!
“The erosion of labour protection for workers seen as “unskilled” but working outside of unionised manual jobs moved like water through cracks, exploiting and capitalizing on society’s assumptions about whose work needed protection and who was worth protecting.”
We learn that the various emerging and existing AI technologies are neither as powerful or terrifying as their creators and media outlets would have you believe, which is a semi-comforting thought.
“Our employment classification systems, won in the 1930s to make full-time assembly line work sustainable, were not built for this future.”
This is one of those tech books, which like many in that genre has the feel of a fleshed out article intended for an airplane magazine. This gives a basic and informative history of the origins of the gig economy and ghost work, though at times it felt like it was working as much as a stealth advocate for certain companies, but in the end this is a decent enough if uneven account.
November 1, 2021
Interesting look into some of the realities of doing ghost work and consideration of how that kind of work might be necessary, but lacks analysis and proper discussion of it as a political and social issue, focusing more on making ghost work seem a bit "nicer".
September 26, 2019
I don't follow Silicon Valley much, so this book offers a very clear picture from one angle: ghost workers; I also learned the role labor laws played, especially by Taft-Hartley.
However, the book doesn't quite fit the subtitle "how to stop Silicon Valley from building a new global underclass." I have the feeling that the authors mostly talked from a technical bubble (I guess that's their expertise), and a lot of the answers are quite shallow. Only in the last few page, the book covers universal healthcare and universal basic income. I wanted to give it 3.5, but 3 in the end.
However, the book doesn't quite fit the subtitle "how to stop Silicon Valley from building a new global underclass." I have the feeling that the authors mostly talked from a technical bubble (I guess that's their expertise), and a lot of the answers are quite shallow. Only in the last few page, the book covers universal healthcare and universal basic income. I wanted to give it 3.5, but 3 in the end.
July 8, 2019
Works as a 5 minute short article in the new yorker or something, but expanding it out into a full book didn't add much (beyond a bunch of words.) Never really got into the how to promised in the subtitle.
October 11, 2019
This is a book about the piece rate workers who interact and support many of the internet sites that most of us see as completely automated rather than the product of human labor. Indeed, one of the more interesting perspectives of this book is that the conceptual split between human labor and automation is often a faulty one. As AI based and other automated systems progress, there will always be a “last mile” to traverse during which the human-machine interaction will not just persist but will be crucial. The focal activities of the book center on the large systems operated by Microsoft or Amazon (Mturk), although there are myriad others. This sort of “Ghost Work” is similar to more commonly recognized actors in the “Gig Economy” such as Uber or Lyft, but involves micro level and less involved tasks, often taking only seconds to complete. Individuals are paid a pittance for their work and must aggregate large amounts of successfully completed tasks to ever earn much.
There is an emerging genre of books about temp work in general and temp work on the web in particular. Some discuss specific problems, such as content moderation. Others are broader, ranging from temporary office work and consultants to modern Uber-like firms. The work profiled here fits right in, albeit in its own niche.
There has long been a difficulty in counting the number of workers in this hidden economy, and the Census Bureau has had trouble getting beyond the 10% who do gig work as a primary occupation. Gray makes the good point in this book that it is common for people to do this work in addition to other often full time jobs in order to supplement their incomes. That suggests that official estimates of ghost workers is biased downwards. The trouble is that it remains unclear just what the extent of the bias is and how many workers we are really focusing on. That is OK. The phenomena are real, the number of people involved in growing, and it is worth understanding.
A related issue is the horrible conditions of this work. Part of the appeal of ghost work for employers (or requesters) is that all of those nasty features of employees such as work rules, benefits, unions, and the like are irrelevant to this sorts of temp systems. The book is informative about the legal origins of this situation in US labor law. The authors suggest a number of fixes, some of which seem reasonable. It is unclear, however, how the dynamics governing the growth of ghost work will be changed anytime soon. Even if changed, it is unclear what conditions for organizing this temporary work will replace current arrangements.
The authors raise a number of issues that could benefit from further discussion as well. For example, the fact that the largest number of ghost workers are in India and the US, brings up issues of cultural and linguistic fluency, especially in cultural norms, ethnic prejudices, and knowledge of idioms. This has implications for how workers coordinate with each other in different settings, such as in whether chat room or phone connections are used and what that means for the state of a work force. Some companies have made use of geographic presence to their advantage, such as in sending difficult problems in the US to Indian facilities for overnight work by individuals there who still follow US working hours. The book also touches on the need for better health coverage for ghost workers and the rationales for such coverage - raising the question of what the health conditions are for a large atomized workforce that spends much of its time hunches over a screen on old chairs at the back of their residences. Talk about sweat shops!
This is a thoughtful and well written book about the underside of the internet economy and is well worth reading.
There is an emerging genre of books about temp work in general and temp work on the web in particular. Some discuss specific problems, such as content moderation. Others are broader, ranging from temporary office work and consultants to modern Uber-like firms. The work profiled here fits right in, albeit in its own niche.
There has long been a difficulty in counting the number of workers in this hidden economy, and the Census Bureau has had trouble getting beyond the 10% who do gig work as a primary occupation. Gray makes the good point in this book that it is common for people to do this work in addition to other often full time jobs in order to supplement their incomes. That suggests that official estimates of ghost workers is biased downwards. The trouble is that it remains unclear just what the extent of the bias is and how many workers we are really focusing on. That is OK. The phenomena are real, the number of people involved in growing, and it is worth understanding.
A related issue is the horrible conditions of this work. Part of the appeal of ghost work for employers (or requesters) is that all of those nasty features of employees such as work rules, benefits, unions, and the like are irrelevant to this sorts of temp systems. The book is informative about the legal origins of this situation in US labor law. The authors suggest a number of fixes, some of which seem reasonable. It is unclear, however, how the dynamics governing the growth of ghost work will be changed anytime soon. Even if changed, it is unclear what conditions for organizing this temporary work will replace current arrangements.
The authors raise a number of issues that could benefit from further discussion as well. For example, the fact that the largest number of ghost workers are in India and the US, brings up issues of cultural and linguistic fluency, especially in cultural norms, ethnic prejudices, and knowledge of idioms. This has implications for how workers coordinate with each other in different settings, such as in whether chat room or phone connections are used and what that means for the state of a work force. Some companies have made use of geographic presence to their advantage, such as in sending difficult problems in the US to Indian facilities for overnight work by individuals there who still follow US working hours. The book also touches on the need for better health coverage for ghost workers and the rationales for such coverage - raising the question of what the health conditions are for a large atomized workforce that spends much of its time hunches over a screen on old chairs at the back of their residences. Talk about sweat shops!
This is a thoughtful and well written book about the underside of the internet economy and is well worth reading.
June 4, 2025
I had some familiarity with the exploitative labor practices underpinning some of the technologies under the umbrella term “artificial intelligence”, so I expected to learn a lot more of the grim specifics in this book. I did, and having finished i, now my perception of many facets of the tech industry has been illuminated and expanded.
The section on the history of “ghost work” and invisible labor was surprisingly relevant to costume production/garment work, as the authors investigate in-depth the industrialization of the textile industry in the 18th & 19th centuries, as well as that industry’s reliance on the jobbed-out labor practice of piecework.
In the history section, they also draw a parrallel to the teams of “human computers” (mostly women and people of color), mathematicians who contributed invaluably to NASA and the lunar space-race but whose contributions have been obscured in the historical record.
I expected a certain level of focus on a dystopian hellscape of exploited workers in remote areas of the globe completing monotonous content moderation tasks for pennies through the platform of Mechanical Turk. And yes, that’s a chunk of the book, but the authors also profile positive ghostwork companies like Amara (a translation and subtitling company) and LeadGenius (an account-based marketing company).
Additionally, the book offers a healthy amount of hopeful suggestions of how to push back against the depersonalization and exploitation of ghost workers, and how humane scaffolding of ghostworking could positively reimagine the workforce of the future. In conclusion, the final chapter offers ten very feasible potential fixes.
The section on the history of “ghost work” and invisible labor was surprisingly relevant to costume production/garment work, as the authors investigate in-depth the industrialization of the textile industry in the 18th & 19th centuries, as well as that industry’s reliance on the jobbed-out labor practice of piecework.
In the history section, they also draw a parrallel to the teams of “human computers” (mostly women and people of color), mathematicians who contributed invaluably to NASA and the lunar space-race but whose contributions have been obscured in the historical record.
I expected a certain level of focus on a dystopian hellscape of exploited workers in remote areas of the globe completing monotonous content moderation tasks for pennies through the platform of Mechanical Turk. And yes, that’s a chunk of the book, but the authors also profile positive ghostwork companies like Amara (a translation and subtitling company) and LeadGenius (an account-based marketing company).
Additionally, the book offers a healthy amount of hopeful suggestions of how to push back against the depersonalization and exploitation of ghost workers, and how humane scaffolding of ghostworking could positively reimagine the workforce of the future. In conclusion, the final chapter offers ten very feasible potential fixes.
April 10, 2022
Creo que me esperaba otra cosa; en vez de un estudio de cuatro plataformas de trabajo, más historias de distintas empresas que recurren a este tipo de trabajo. Pero salvo algunas partes un poco repetitivas es muy muy interesante.
July 23, 2023
Big tech gives the illusion that everything can be accomplished with the power of computation. And yet no labor sector seems to have grown more in the last couple of years than the hidden away gig economy. Humans doing the tasks that code can't. Through Amazon's Mechanical Turk and similar platforms, they type, label, categorize, moderate, proofread, translate, flag and tag. They are the hidden power that helps span the gap across A.I.'s "last mile" problem.
There's a big variety to gig work, with low-skills-required micro tasks on one end, and highly specialized freelance projects on the other end. What they have in common is, that worker and employer are connected through an API. Simplifying and dehumanizing the hiring process. Which comes with pros (no discrimination based on gender/religion/ethnicity) and cons (no security net). I love how working on these platforms seems to give a lot of women in patriarchal societies a way to achieve a level of independence. And women who are primarily caregivers, also can join the work force, due to gig work's flexibility in scheduling.
This is a very well researched book based on lots of interviews and data surveys (luckily gig workers love filling out surveys AND getting paid for doing so).
There are estimates that the on-demand gig-economy will employ 65% of the total workforce by 2055. I kept thinking of the portrayal of that in S.B. Divya's Machinehood.
There's a big variety to gig work, with low-skills-required micro tasks on one end, and highly specialized freelance projects on the other end. What they have in common is, that worker and employer are connected through an API. Simplifying and dehumanizing the hiring process. Which comes with pros (no discrimination based on gender/religion/ethnicity) and cons (no security net). I love how working on these platforms seems to give a lot of women in patriarchal societies a way to achieve a level of independence. And women who are primarily caregivers, also can join the work force, due to gig work's flexibility in scheduling.
This is a very well researched book based on lots of interviews and data surveys (luckily gig workers love filling out surveys AND getting paid for doing so).
There are estimates that the on-demand gig-economy will employ 65% of the total workforce by 2055. I kept thinking of the portrayal of that in S.B. Divya's Machinehood.
February 2, 2024
Important read to learn about the hidden labour economy and its cruel systematization, banking on the disposability and desperation of cheap labour.
September 3, 2020
Team-based to remove systematic cruelty. Difficult to implement though
May 9, 2024
There was a new item recently about the Amazon supermarket, where you walk around just basically grabbing things off the shelves and AI registers all of your ‘purchases’ and deducts the correct amount from your bank account once you walk out the door. The wonders of modern technology… except that it turns out that it wasn’t run by AI at all, but somewhere in India there was a crowd of people looking at screens watching each of the supermarket customers and recording what they were putting into their baskets. As someone much funnier than I am said, “AI stands for An Indian”.
There are lots of books telling us that we are just about to reach a point where machines will do everything for us and we will no longer need to work – that we are on the cusp of fully automated luxury communism or, more likely and believably, a horrible dystopia where 90% of the population are surplus to requirements and need to find ever more degrading ways to bum money out of passers-by. This book says that as technology advances it creates new and ever stranger jobs for humans that the machines simply can’t do. However, there is also a tendency of capitalism to only want to pay employees piecemeal, by the task. And new technologies are particularly good at facilitating that – task allocation, surveillance of keystrokes and paying by the minute. And capitalism can also be summed up by that pithy little phrase Margaret Thatcher coined, ‘There’s no such thing as society. There are individual men and women and there are families.’ Capitalism likes to deal with each of us individually, and to keep us as isolated from each other as is humanly possible. An idea I’ll come back to.
Ghost Work is this kind of work. Often it is work that employs people to meet times of peak demand on some project or other – so that the company employing the ghost workers has ‘real employees’ too, but sometimes they just need some others to get themselves over a hump. Sometimes most of the company’s employees are essentially ghost employees – people who wait for certain jobs to be posted on a site and put their hand up to do that task. Effectively, they could be working for the same company for years, while never actually ‘working for them’ – never actually being an employee with all of the entitlements that would imply. It is easy to see this form of employment relationship as deeply exploitative, but as the book points out, it is that, but also sometimes it works in the employee’s favour. Although the point made about algorithm cruelty needs to be mentioned too. Where you are pitted against everyone in the world, never knowing when a job might be posted, never able to leave your screen in case some work comes in and if you go to the toilet you might miss out. Your time is worthless, company time is infinitely valuable. Everything about the relationship reminds you of your worth, or rather, your lack of worth.
Of course, neoliberal types would argue that any employment relationship freely entered into by consenting adults is non-exploitative by definition – a very cute way to ignore power differentials and the persuasive influence of starvation – but we can ignore the silly arguments of neoliberal types virtually immediately. One tires quickly of the tedious nonsense of perpetual adolescents. All the same, the author makes it clear that women and disabled people in particular like the anonymity of the platforms employing them, since these platforms don’t judge you on characteristics that are not relevant to the work you can do.
Although, a lot of the start of the book says pretty much the opposite. Since all of the power in the employment relationship is in the hands of the employer, the potential for exploitation is very large indeed. A lot of the start of the book is a chronical of people getting ripped off in one way or another – from not being paid for work because the company that employed them disappeared, to not being paid for work because it didn’t meet specifications that had never been specified, to not being paid for work because, well, who knows, no one said, it was just your account was deleted and you had no one to contact to find out what happened. As is said here, just about everyone the research team spoke to doing this kind of work had a horror story to tell. Made all the worse by the fact that the employers had next to zero interest in their employees and would have preferred to have had even less interest in them.
The jobs being performed didn’t need to be mindless. Sometimes they required complex skills – translating, checking images for sexual content (is that a penis or a thumb?), doing CAD drawings or proof-reading documents. The bit rarely changed is the relationship with the employer – this is intended to be wham-bam-thank-you-ma’am contract employment – straight sex, no kissing, no cuddling, bring your own contraceptives. And in fact, virtually always the employees were expected to have their own software, and given this can be insanely expensive, that could be a problem. Some employers also wanted the employee to have at least a sense of their in-house culture so whatever they produced might be on-brand, but some reported here acknowledged that might be a bit much to ask.
The bit of this book I found most interesting was where the employees found ways to create their own means of interacting with each other. So, if there was a single platform where they learned of the tasks available to them, they would set up a companion site where employees could get together to just chat or to discuss jobs or to get suggestions and so on. Humans are fundamentally social creatures, despite capitalism’s preference for us to be atoms, and so we will often go out of our ways to create a means to communicate with each other. This was often frowned upon by the employers, but the need for human contact is often greater than the threat to one’s livelihood.
The book ends with a series of recommendations for companies. The book believes this form of employment relationship will continue to become normalised and perhaps even preferred. I’ve heard such grossly exploitative forms of employment referred to as ‘entrepreneurial’ – which is nothing if not funny. The author believes that companies should be encouraged to understand that it is in their own best interests to ensure they have better relationships with their employees, even employees on short-term, piecemeal contracts. The author points to the potential benefits that would come from these better relationships. While all this was being said I kept hearing that saying I don’t quite remember about no advance in human dignity every being achieved without first a demand and second some form of collective action. Having worked for decades as a trade union organiser, the idea that employers would improve working conditions due to sentiment nearly made me laugh out loud while reading this. Yeah, off you go, appeal to their better natures – what could go wrong?
All the same, this book was based on lots of research and a lot of this was new to me. The idea you might do work for Amazon for free – as I am doing now – is not new to me, but the idea you might do work for Amazon, real work, and get paid in essentially ‘tokens’ to purchase from the Amazon site – now that is appalling. I wish I could believe in hell.
There are lots of books telling us that we are just about to reach a point where machines will do everything for us and we will no longer need to work – that we are on the cusp of fully automated luxury communism or, more likely and believably, a horrible dystopia where 90% of the population are surplus to requirements and need to find ever more degrading ways to bum money out of passers-by. This book says that as technology advances it creates new and ever stranger jobs for humans that the machines simply can’t do. However, there is also a tendency of capitalism to only want to pay employees piecemeal, by the task. And new technologies are particularly good at facilitating that – task allocation, surveillance of keystrokes and paying by the minute. And capitalism can also be summed up by that pithy little phrase Margaret Thatcher coined, ‘There’s no such thing as society. There are individual men and women and there are families.’ Capitalism likes to deal with each of us individually, and to keep us as isolated from each other as is humanly possible. An idea I’ll come back to.
Ghost Work is this kind of work. Often it is work that employs people to meet times of peak demand on some project or other – so that the company employing the ghost workers has ‘real employees’ too, but sometimes they just need some others to get themselves over a hump. Sometimes most of the company’s employees are essentially ghost employees – people who wait for certain jobs to be posted on a site and put their hand up to do that task. Effectively, they could be working for the same company for years, while never actually ‘working for them’ – never actually being an employee with all of the entitlements that would imply. It is easy to see this form of employment relationship as deeply exploitative, but as the book points out, it is that, but also sometimes it works in the employee’s favour. Although the point made about algorithm cruelty needs to be mentioned too. Where you are pitted against everyone in the world, never knowing when a job might be posted, never able to leave your screen in case some work comes in and if you go to the toilet you might miss out. Your time is worthless, company time is infinitely valuable. Everything about the relationship reminds you of your worth, or rather, your lack of worth.
Of course, neoliberal types would argue that any employment relationship freely entered into by consenting adults is non-exploitative by definition – a very cute way to ignore power differentials and the persuasive influence of starvation – but we can ignore the silly arguments of neoliberal types virtually immediately. One tires quickly of the tedious nonsense of perpetual adolescents. All the same, the author makes it clear that women and disabled people in particular like the anonymity of the platforms employing them, since these platforms don’t judge you on characteristics that are not relevant to the work you can do.
Although, a lot of the start of the book says pretty much the opposite. Since all of the power in the employment relationship is in the hands of the employer, the potential for exploitation is very large indeed. A lot of the start of the book is a chronical of people getting ripped off in one way or another – from not being paid for work because the company that employed them disappeared, to not being paid for work because it didn’t meet specifications that had never been specified, to not being paid for work because, well, who knows, no one said, it was just your account was deleted and you had no one to contact to find out what happened. As is said here, just about everyone the research team spoke to doing this kind of work had a horror story to tell. Made all the worse by the fact that the employers had next to zero interest in their employees and would have preferred to have had even less interest in them.
The jobs being performed didn’t need to be mindless. Sometimes they required complex skills – translating, checking images for sexual content (is that a penis or a thumb?), doing CAD drawings or proof-reading documents. The bit rarely changed is the relationship with the employer – this is intended to be wham-bam-thank-you-ma’am contract employment – straight sex, no kissing, no cuddling, bring your own contraceptives. And in fact, virtually always the employees were expected to have their own software, and given this can be insanely expensive, that could be a problem. Some employers also wanted the employee to have at least a sense of their in-house culture so whatever they produced might be on-brand, but some reported here acknowledged that might be a bit much to ask.
The bit of this book I found most interesting was where the employees found ways to create their own means of interacting with each other. So, if there was a single platform where they learned of the tasks available to them, they would set up a companion site where employees could get together to just chat or to discuss jobs or to get suggestions and so on. Humans are fundamentally social creatures, despite capitalism’s preference for us to be atoms, and so we will often go out of our ways to create a means to communicate with each other. This was often frowned upon by the employers, but the need for human contact is often greater than the threat to one’s livelihood.
The book ends with a series of recommendations for companies. The book believes this form of employment relationship will continue to become normalised and perhaps even preferred. I’ve heard such grossly exploitative forms of employment referred to as ‘entrepreneurial’ – which is nothing if not funny. The author believes that companies should be encouraged to understand that it is in their own best interests to ensure they have better relationships with their employees, even employees on short-term, piecemeal contracts. The author points to the potential benefits that would come from these better relationships. While all this was being said I kept hearing that saying I don’t quite remember about no advance in human dignity every being achieved without first a demand and second some form of collective action. Having worked for decades as a trade union organiser, the idea that employers would improve working conditions due to sentiment nearly made me laugh out loud while reading this. Yeah, off you go, appeal to their better natures – what could go wrong?
All the same, this book was based on lots of research and a lot of this was new to me. The idea you might do work for Amazon for free – as I am doing now – is not new to me, but the idea you might do work for Amazon, real work, and get paid in essentially ‘tokens’ to purchase from the Amazon site – now that is appalling. I wish I could believe in hell.
December 19, 2020
This is a well-researched account of ghost work that people do for large corporations. Much of the content moderation work for user-generated text and video is outsourced to ghost work. In addition to this, much of the annotation for text classification, image tagging, and video tagging and moderation (for family safe content and to train AI/ML algorithms) is done by gig workers. The authors have covered the pain points of two-sided marketplaces very well delving into issues of safety, equity, trust, and community.
The middle part of the book focuses on marketplace and business model design which can cater to both sides of the equation to bring positive change rather than the exploitative practice that this sector is known for. Overall the storytelling with individual stores gives a human face to the problem. At the same time, data visualizations are also used very well to cover the issues qualitatively and quantitively. I felt the complex dynamics of gig work and marketplaces with different types of participants with differing incentives on both the demand and supply side (divided into always-on, regulars, and experimentalists) was very interesting.
On the flip side, there is very little coverage of ride-hailing and food-delivery companies across the globe such as Uber, Lyft, Zomato, Doordash, and more. Secondly, the book seems mostly focused on India and the US (and leaves out Europe and LatAm completely). The book doesn't criticize Silicon Valley companies and their malpractice and evil behavior in the book, giving a more business friendly version (especially in the final conclusion section). Even with these flaws, the book is still quite interesting to read.
The middle part of the book focuses on marketplace and business model design which can cater to both sides of the equation to bring positive change rather than the exploitative practice that this sector is known for. Overall the storytelling with individual stores gives a human face to the problem. At the same time, data visualizations are also used very well to cover the issues qualitatively and quantitively. I felt the complex dynamics of gig work and marketplaces with different types of participants with differing incentives on both the demand and supply side (divided into always-on, regulars, and experimentalists) was very interesting.
On the flip side, there is very little coverage of ride-hailing and food-delivery companies across the globe such as Uber, Lyft, Zomato, Doordash, and more. Secondly, the book seems mostly focused on India and the US (and leaves out Europe and LatAm completely). The book doesn't criticize Silicon Valley companies and their malpractice and evil behavior in the book, giving a more business friendly version (especially in the final conclusion section). Even with these flaws, the book is still quite interesting to read.
December 29, 2019
This is an outstanding example of the power of insight into new ways of working by melding the power of qualitative and quantitative methods. They are very transparent and tell you how they did field research, ethnography, and data science - an example for students out there. As an educator, this is exemplary of the work teams necessary to explore a problem.
If you think automation has replaced manufacturing factory work, then this book introduces you to how tech has gutted traditional middle-class job pathways and now has transformed into what is known as ghost work. This is what is missing in Andrew Yang's speech on automation, it is not just the truck drivers or retail workers, but tech is now redefining work for college-educated youth and adults with very little protection, benefits, and social interaction. This is an urgent issue for my American friends.
There are several important points that they raise here:
1. The necessity to redefine jobs and work for the 21st century and that includes measuring and statistics of the kinds of work currently being performed but unrecorded. It is no longer simply as "self-employed," gig, or "part-time." There are little figures and statistics on how many people perform ghost work and they pegged the number at about 20 million hidden labor unaccounted for.
Redefinition is crucial to re-working an equitable economy and fair labour markets.
2. Contingent labourers like ghost workers have a long history stretching back to the first industrial era. They explained how the rise of unions served to protect assembly line workers in the factories BUT NOT piecemeal labourers outside of the factories who are therefore subject to exploitation. The result is the status difference between "full-time" work and "part-time" work. This status difference is important in the new era in which anxiety is now fast becoming the norm since most people are "gigging" or in-between work. Tech and automation has now erased this distinction towards a single model of "on-demand" labor. However, the social difference and stigma have remained and I believe is one root of status anxiety among new grads, those currently laid off or looking for new jobs. The new marketplace is on-demand and we need new ways to re-frame this new experience as it becomes the norm.
3. The new offices are the new work platforms in which APIs now define the kind of piecemeal work or macro-work. The lack of regulation threatens the well-being of workers. This includes both US and Indian outsourced workers. This brings us to the issue of worldwide rather than nation-based regulation protections and guidelines. How do we re-design more ethical APIs is the challenge and is there a way to solve this divide? More and more you can see those in the Global North becoming less competitive facing greater threat to job loss as ghost work is being done cheaply in the Global South.
There are many more nuggets to be gotten from this book depending on your background. This is a definite read. They offer clear recommendations but what they have failed to do is to relate it to their parent company Microsoft. This research was funded and under the aegis of Microsoft Research and they failed to explain how their research impacts how Microsoft, who also has its own Turk work platform, can do better or what are their own internal blockages to ramping up change. This is like the big elephant. As a UX researcher, I think they really need to address the relationship between their own research silo with their product teams. What are the typical challenges that their own engineers face in designing an ethical API? Why aren't they spearheading this? Are their own executives listening to them? These are the questions I want to know.
This is definitely a description of the making of a global underclass beginning with its destruction in the USA. Whether or not silicon valley has financial incentives to stop it, the recommendations remain good for book conclusions. If Dr. Gray can convince Microsoft to do good by API design, they would be global leaders.
If you think automation has replaced manufacturing factory work, then this book introduces you to how tech has gutted traditional middle-class job pathways and now has transformed into what is known as ghost work. This is what is missing in Andrew Yang's speech on automation, it is not just the truck drivers or retail workers, but tech is now redefining work for college-educated youth and adults with very little protection, benefits, and social interaction. This is an urgent issue for my American friends.
There are several important points that they raise here:
1. The necessity to redefine jobs and work for the 21st century and that includes measuring and statistics of the kinds of work currently being performed but unrecorded. It is no longer simply as "self-employed," gig, or "part-time." There are little figures and statistics on how many people perform ghost work and they pegged the number at about 20 million hidden labor unaccounted for.
Redefinition is crucial to re-working an equitable economy and fair labour markets.
2. Contingent labourers like ghost workers have a long history stretching back to the first industrial era. They explained how the rise of unions served to protect assembly line workers in the factories BUT NOT piecemeal labourers outside of the factories who are therefore subject to exploitation. The result is the status difference between "full-time" work and "part-time" work. This status difference is important in the new era in which anxiety is now fast becoming the norm since most people are "gigging" or in-between work. Tech and automation has now erased this distinction towards a single model of "on-demand" labor. However, the social difference and stigma have remained and I believe is one root of status anxiety among new grads, those currently laid off or looking for new jobs. The new marketplace is on-demand and we need new ways to re-frame this new experience as it becomes the norm.
3. The new offices are the new work platforms in which APIs now define the kind of piecemeal work or macro-work. The lack of regulation threatens the well-being of workers. This includes both US and Indian outsourced workers. This brings us to the issue of worldwide rather than nation-based regulation protections and guidelines. How do we re-design more ethical APIs is the challenge and is there a way to solve this divide? More and more you can see those in the Global North becoming less competitive facing greater threat to job loss as ghost work is being done cheaply in the Global South.
There are many more nuggets to be gotten from this book depending on your background. This is a definite read. They offer clear recommendations but what they have failed to do is to relate it to their parent company Microsoft. This research was funded and under the aegis of Microsoft Research and they failed to explain how their research impacts how Microsoft, who also has its own Turk work platform, can do better or what are their own internal blockages to ramping up change. This is like the big elephant. As a UX researcher, I think they really need to address the relationship between their own research silo with their product teams. What are the typical challenges that their own engineers face in designing an ethical API? Why aren't they spearheading this? Are their own executives listening to them? These are the questions I want to know.
This is definitely a description of the making of a global underclass beginning with its destruction in the USA. Whether or not silicon valley has financial incentives to stop it, the recommendations remain good for book conclusions. If Dr. Gray can convince Microsoft to do good by API design, they would be global leaders.
December 6, 2020
Starting from the following research questions,
- Who are ghost works, and how does their work differ from traditional nine-to-five jobs?
- Who benefits from this veneer of automation?
- And who might be harmed?
The authors spent five years and conducted hundreds of interviews and surveys, dozens of behavioral experiments and social network analysis, and case studies of other stakeholders.
Gists of the book
- AI can't, and will not, solve all the problems. Human labor is irreplaceable.
- The future is not that AI will serve all human beings, but more human workers and AI systems collaborate.
- As part of human needs, on-demand workers like to collaborate and support others even in the absence of a physical workplace. The presence and effort of people behind the scenes are the true currency and value of the labor market. However, some platforms ignore this need and treat the workers cruelly as a piece of a machine (e.g. MTurk).
Facing the problems, the authors proposed some fixes to the current ecosystem.
- Encourage collaboration
- Build guild-like communities for on-demand workers.
- Define two streams of work: flash teams and independent work
- Requesters and companies provide the employer of record.
- Create a “good-work code” for ghost work supply chains, which need the collaboration of multiple stakeholders
- Create a “safety net” for future workers: universal worker welfare and retainer base wage for all working adults
- Consumer actions to protect workers’ benefits
I appreciate the breadth and depth of this book, such that it not only help us understand the work and life of ghost workers, reviews the history of on-demand work, also analyzes the flaws and drawbacks of the platforms and the human-AI collaboration ecosystem.
Some critics.
First, the authors keep talking about “ghost workers”, they also mention gig economics. But they do not differentiate the two. They just simply see them as the same thing, using “on-demand ghost worker” to describe them. But freelancers and ghost workers are different. At least the former are not “ghosts”. They have public profiles and clients recognize them. The nature of their work is not to serve machine learning algorithms. Instead, what they do is almost the same as full-time workers, design, programming, writing, etc. They just choose an approach with more freedom. Therefore, some arguments and fixes proposed by the authors do not hold for all the cases.
Second, when I read the book, I feel a sense of subjective preference instead of objective, academic description. Especially when the authors talk about MTurk (very negative) and CloudFactory (very positive).
Third, I believe the authors and good intentions and want to create a new future for all the on-demand workers, but some proposals and fixes sound a bit unnecessary. I like the idea of building guild-like communities and empathy, as it is part of human need. But the idea that companies should provide resources such as health insurance, training, and working equipment sounds not easy. If it works, what is the difference with contractors? So it goes back to the first point, the authors should give a more thorough classification and definition of workers, Also, they did not talk much about how the companies think. I understand it’s hard to get information from them as it may be commercially confidential. But without understanding the willingness and capacities of these influencers, the proposals sound not quite feasible.
Despite the critics, I still like this book pretty much. It is an interesting and enjoyable ethnographic work to read through.
- Who are ghost works, and how does their work differ from traditional nine-to-five jobs?
- Who benefits from this veneer of automation?
- And who might be harmed?
The authors spent five years and conducted hundreds of interviews and surveys, dozens of behavioral experiments and social network analysis, and case studies of other stakeholders.
Gists of the book
- AI can't, and will not, solve all the problems. Human labor is irreplaceable.
- The future is not that AI will serve all human beings, but more human workers and AI systems collaborate.
- As part of human needs, on-demand workers like to collaborate and support others even in the absence of a physical workplace. The presence and effort of people behind the scenes are the true currency and value of the labor market. However, some platforms ignore this need and treat the workers cruelly as a piece of a machine (e.g. MTurk).
Facing the problems, the authors proposed some fixes to the current ecosystem.
- Encourage collaboration
- Build guild-like communities for on-demand workers.
- Define two streams of work: flash teams and independent work
- Requesters and companies provide the employer of record.
- Create a “good-work code” for ghost work supply chains, which need the collaboration of multiple stakeholders
- Create a “safety net” for future workers: universal worker welfare and retainer base wage for all working adults
- Consumer actions to protect workers’ benefits
I appreciate the breadth and depth of this book, such that it not only help us understand the work and life of ghost workers, reviews the history of on-demand work, also analyzes the flaws and drawbacks of the platforms and the human-AI collaboration ecosystem.
Some critics.
First, the authors keep talking about “ghost workers”, they also mention gig economics. But they do not differentiate the two. They just simply see them as the same thing, using “on-demand ghost worker” to describe them. But freelancers and ghost workers are different. At least the former are not “ghosts”. They have public profiles and clients recognize them. The nature of their work is not to serve machine learning algorithms. Instead, what they do is almost the same as full-time workers, design, programming, writing, etc. They just choose an approach with more freedom. Therefore, some arguments and fixes proposed by the authors do not hold for all the cases.
Second, when I read the book, I feel a sense of subjective preference instead of objective, academic description. Especially when the authors talk about MTurk (very negative) and CloudFactory (very positive).
Third, I believe the authors and good intentions and want to create a new future for all the on-demand workers, but some proposals and fixes sound a bit unnecessary. I like the idea of building guild-like communities and empathy, as it is part of human need. But the idea that companies should provide resources such as health insurance, training, and working equipment sounds not easy. If it works, what is the difference with contractors? So it goes back to the first point, the authors should give a more thorough classification and definition of workers, Also, they did not talk much about how the companies think. I understand it’s hard to get information from them as it may be commercially confidential. But without understanding the willingness and capacities of these influencers, the proposals sound not quite feasible.
Despite the critics, I still like this book pretty much. It is an interesting and enjoyable ethnographic work to read through.
March 9, 2020
Important topic, but book could have been a blog post and felt repetitive (especially yet another paragraph introducing an Indian MTurker, his/her age, religion and some coleur locale; grew stale quite fast).
Piecework in an online environment (the human computation between augmented services). On-demand ghost work platforms and their APIs allow humans to power many of the websites, apps, online services, and algorithms most consumers think are automated.
The burden of transaction costs used to fall on companies, but now it falls squarely on the people doing ghost work.
Leading to Algorithmic cruelty
1 Requesters can upload large number of tasks, and then disappear (workers need to be hypervigilant, spun as flexibility)
2 The API governs interaction, no one to turn to for guidance or assistance (spun as autonomy)
3 platform decides on access and pay-outs, without appeal (supposedly against fraud; technical problems are also the workers’ problem; 30 percent of on-demand gig workers reported not getting paid for work they performed.
Ghost work also became more popular because of "just-in-time" scheduling, which remains common among larger retail and hospitality employers.
Building better work is both a social and a technical challenge
Some snippets:
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 nonprofit site dedicated to translating and captioning content
p54 Hatton argues that this "liability model," which treats workers as a drag on corporate profits, overtook the "asset model" of employment. It is possible that the asset model never made it out of the manufacturing sector after the pas sage of the Taft-Hartley Act in 1947. What is clear is that, by the time the internet was born, the U.S. was already moving quickly to a reliance on contract-driven services indefinitely staffed by contingent labor.
The transaction costs of ghost work don't melt away. Instead they are shifted to the shoulders of requesters and workers.
p143 the future of on-demand work, and the accountability and care for he conditions of those doing macro- or micro-task ghost work, is at a crossroads. One option is to hold companies that rely on software and a human labor pool to deliver a service to consumers legally accountable to that labor pool as their employers. Another option is for workers to continue to fend for themselves through the kindness of strangers and well-intentioned companies hiring them on contract. There might be other viable options that make ghost work a more sustainable form of employment. Finding our way to these other options that distribute the Elements of this new economy to workers, on-demand services, and consumers more equitably could come from considering how far intentional design focused on both profit and worker experience takes us, compared with the limits of narrowly focusing on a single bottom line.
Taken as a whole, the current "gig economy"- an ecosystem ecosystem of inde pendent 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 as sign, schedule, route, and bill work through websites or mobile app
Piecework in an online environment (the human computation between augmented services). On-demand ghost work platforms and their APIs allow humans to power many of the websites, apps, online services, and algorithms most consumers think are automated.
The burden of transaction costs used to fall on companies, but now it falls squarely on the people doing ghost work.
Leading to Algorithmic cruelty
1 Requesters can upload large number of tasks, and then disappear (workers need to be hypervigilant, spun as flexibility)
2 The API governs interaction, no one to turn to for guidance or assistance (spun as autonomy)
3 platform decides on access and pay-outs, without appeal (supposedly against fraud; technical problems are also the workers’ problem; 30 percent of on-demand gig workers reported not getting paid for work they performed.
Ghost work also became more popular because of "just-in-time" scheduling, which remains common among larger retail and hospitality employers.
Building better work is both a social and a technical challenge
Some snippets:
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 nonprofit site dedicated to translating and captioning content
p54 Hatton argues that this "liability model," which treats workers as a drag on corporate profits, overtook the "asset model" of employment. It is possible that the asset model never made it out of the manufacturing sector after the pas sage of the Taft-Hartley Act in 1947. What is clear is that, by the time the internet was born, the U.S. was already moving quickly to a reliance on contract-driven services indefinitely staffed by contingent labor.
The transaction costs of ghost work don't melt away. Instead they are shifted to the shoulders of requesters and workers.
p143 the future of on-demand work, and the accountability and care for he conditions of those doing macro- or micro-task ghost work, is at a crossroads. One option is to hold companies that rely on software and a human labor pool to deliver a service to consumers legally accountable to that labor pool as their employers. Another option is for workers to continue to fend for themselves through the kindness of strangers and well-intentioned companies hiring them on contract. There might be other viable options that make ghost work a more sustainable form of employment. Finding our way to these other options that distribute the Elements of this new economy to workers, on-demand services, and consumers more equitably could come from considering how far intentional design focused on both profit and worker experience takes us, compared with the limits of narrowly focusing on a single bottom line.
Taken as a whole, the current "gig economy"- an ecosystem ecosystem of inde pendent 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 as sign, schedule, route, and bill work through websites or mobile app
December 27, 2019
An exquisitely well-researched work of investigative journalism, this book goes deep into the world of the API-based internet labor force - the workers behind human computation engines like Amazon's Mechanical Turk, crowdflower, and many others, contract work sites like Upwork, Fiverr, and also transportation and delivery sites like Uber, Lyft, and instacart. What all of these sites have in common is a two-sided marketplace in which the consumers, whether developers or transportation customers, are intentionally walled off from the human identities of those doing the work for them via an API or an app. Through the stories of individual workers as well as detailed figures and statistics, the authors show how labor protections rarely extend to this class of work, and how the dehumanization of this intentional separation lead customers (and sometimes meta-customers, in the guise of algorithms) to treat workers far more harshly than they would in traditional settings (for instance, removing a worker from their site because their address changes too many times - a rule decides that they are a spammer, when in fact they're an honest worker who has just lost their only source of income, with little opportunity for appeal or recourse). In the end, I think this book makes (and brilliantly illustrates) two very important point. First, we should be mindful of the people who are behind these APIs, and argue for greater regulation and worker protections to improve their working conditions and pay. Second, while every third tech article is about how automation is about to take our jobs, a far more imminent (and evidence-based) risk is having and increasing number of jobs pieced apart into microtasks and farmed out to anonymous labor pools. While these sites tend to describe themselves as disruptive, their real innovation is finding ways to pay workers less for the same work (while avoiding paying taxes and employment benefits), which is only serving to amplify our current crisis in income inequality.
June 16, 2019
Books critically evaluating Silicon Valley are becoming a grim and established subgenre, one I'm becoming well-acquainted with at this point. There's something about the dual nature of the tech sector, promising a glittering future for all humankind while producing uncountable inequities around the world, that cries out for my attention. As a disabled person who benefits enormously from accessible technology and virtual social spaces, I feel beholden to fully understand the tools I use each day and their impact on others.
Unlike so many other books in the same vein, "Ghost Work" was surprisingly optimistic. Gray and Suri didn't shy away from sharing data detailing the ways ghost work exploits the "underclass" that does it, but they also present a compelling case for ghost work as a flexible, accessible alternative to the elusive nine-to-five. In a society where very few people are fortunate enough to land a well-paying traditional job, it makes sense to offer work that can accommodate schedules, disabilities, familial obligations and poverty-related restrictions. If people can't come to the work, why not shape the work so that it comes to them, in whatever form they need?
Of course, the afore-mentioned exploitation blights this golden view of ghost work. Invariably, companies get greedy, requesters become too demanding, and a handful of scammers complicates the process for everyone else. Humans are good at taking a solid concept and transforming it into an exploitive revenue generator, totally disregarding their own role in the mechanisms they create.
But, once again unlike so many books like this, the authors conclude by offering multiple, actionable solutions to improve ghost work for everyone. Whether anyone with power will bother to implement them is doubtful, but it was a relief to finish the book with a sense of buoyant possibility. Gray and Suri have written a balanced, solution-focused critique, and I hope their ideas will stick.
Unlike so many other books in the same vein, "Ghost Work" was surprisingly optimistic. Gray and Suri didn't shy away from sharing data detailing the ways ghost work exploits the "underclass" that does it, but they also present a compelling case for ghost work as a flexible, accessible alternative to the elusive nine-to-five. In a society where very few people are fortunate enough to land a well-paying traditional job, it makes sense to offer work that can accommodate schedules, disabilities, familial obligations and poverty-related restrictions. If people can't come to the work, why not shape the work so that it comes to them, in whatever form they need?
Of course, the afore-mentioned exploitation blights this golden view of ghost work. Invariably, companies get greedy, requesters become too demanding, and a handful of scammers complicates the process for everyone else. Humans are good at taking a solid concept and transforming it into an exploitive revenue generator, totally disregarding their own role in the mechanisms they create.
But, once again unlike so many books like this, the authors conclude by offering multiple, actionable solutions to improve ghost work for everyone. Whether anyone with power will bother to implement them is doubtful, but it was a relief to finish the book with a sense of buoyant possibility. Gray and Suri have written a balanced, solution-focused critique, and I hope their ideas will stick.
May 9, 2020
Exploring life at what is referred to as "automation's last mile", Gray and Suri discuss the human cost to supposedly seamless technology. Their study is fascinating because while it highlights the many problems that the gig economy and invisible labor represent, they also identify that despite this very precarious work, some individuals who do it find meaningful ways of engaging in. However, the cruz of their discussion exposes the ways in which companies have made huge efforts to externalize nearly every aspect of costs to people they can pay but not consider employees and have utterly no responsibility to them (including giving useful feedback or even paying them properly). What I appreciate about Gray and Suri's work is that they don't just focus on Amazon's Mechanical Turk, the most evident example for these practices but also shows the platform Microsoft has created and also more equitable platforms like Amara. By show these different platforms, the author emphasize that there is the possibility to do create structures that are more humane and dignified. Still, their study brings readers into contact with people around the world attempting to sew together income by having to be constantly "on" and monitoring for the next job while also leveraging their self-created networks to maximize opportunities. At its core, the book elevates the creativity and determination of people, throughout the world, dire to find work to achieve a reasonable income in contrast to an increasing alienated work environment that big and small companies are embracing.
November 17, 2023
October 2 - November 17, 2023 4:00 PM
4 stars
It took me long to finish this book because it’s too long and honestly, I don’t get it. I was hooked by the title of How To Stop Silicon Valley from Building New Global Underclass. I understand that “ghost work” is the people working remotely and assisting the programmers and AI to do their jobs hiddenly and receiving low-wage rates. People have the choice to go into this kind of work and sometimes they are forced by their circumstances. Amazon created a platform for them to find jobs and the payment of wages through Amazon gift cards or something is insane. They needed to exchange that for money and that cost a transaction fee. I mean Amazon is sick but if people are still doing this job then who’s fault it is? I sound like Jeff Bezos who’s heartless and greedy. Sorry, not sorry.
The cause of this book is amazing. They wanted us to be aware that they were people working hidden from the normal jobs we know. They are receiving small compensations and all the bad stuff of unemployment. They are called ghost workers since they are literally hidden and unknown by society. This book opens everyone’s eyes that they exist. I get this part. It’s just this book is too long and too much. My 4 stars are about myself and my expectations of the book but the purpose of this book is great.
Weird that it’s only 200+ pages book but it felt way too long
4 stars
It took me long to finish this book because it’s too long and honestly, I don’t get it. I was hooked by the title of How To Stop Silicon Valley from Building New Global Underclass. I understand that “ghost work” is the people working remotely and assisting the programmers and AI to do their jobs hiddenly and receiving low-wage rates. People have the choice to go into this kind of work and sometimes they are forced by their circumstances. Amazon created a platform for them to find jobs and the payment of wages through Amazon gift cards or something is insane. They needed to exchange that for money and that cost a transaction fee. I mean Amazon is sick but if people are still doing this job then who’s fault it is? I sound like Jeff Bezos who’s heartless and greedy. Sorry, not sorry.
The cause of this book is amazing. They wanted us to be aware that they were people working hidden from the normal jobs we know. They are receiving small compensations and all the bad stuff of unemployment. They are called ghost workers since they are literally hidden and unknown by society. This book opens everyone’s eyes that they exist. I get this part. It’s just this book is too long and too much. My 4 stars are about myself and my expectations of the book but the purpose of this book is great.
Weird that it’s only 200+ pages book but it felt way too long
This entire review has been hidden because of spoilers.
August 11, 2019
This is a well-researched and interesting read, and a welcome reality check on the spin of AI in the public sphere. The authors look behind the screen and into the lives of workers on MTurk, Microsoft's Universal Human Relevance System, LeadGenius, and Amara.org. They give voice to the obscured human labor that powers digital tools we use.
It's hard to see the labor until it's been documented, empowered. Imagine an industrial system coming of age in the 19th century and not seeing a factory run, nor understanding the provision of credit, the lives of slaves, and history hasn't been written, analyzed, condensed, delivered, yet.
Ghost Work is 21st century political economics, present tense history. It is time for tech workers to build solidarity, when work can be broken down to tasks, disaggregated from employer, representation, rights, workers exist between the ages, between the protections that labor won in the last century and their absence under the new terms of employment.
When employers have no obligation to their workers, when workers aren't even acknowledged to be working for the company, rights no longer derive from the labor contract. They must result from a contract with society to provide protections for people who work multiple jobs, so many tiny jobs, that take a moment and a pay a few cents and on to the next.
It's hard to see the labor until it's been documented, empowered. Imagine an industrial system coming of age in the 19th century and not seeing a factory run, nor understanding the provision of credit, the lives of slaves, and history hasn't been written, analyzed, condensed, delivered, yet.
Ghost Work is 21st century political economics, present tense history. It is time for tech workers to build solidarity, when work can be broken down to tasks, disaggregated from employer, representation, rights, workers exist between the ages, between the protections that labor won in the last century and their absence under the new terms of employment.
When employers have no obligation to their workers, when workers aren't even acknowledged to be working for the company, rights no longer derive from the labor contract. They must result from a contract with society to provide protections for people who work multiple jobs, so many tiny jobs, that take a moment and a pay a few cents and on to the next.
January 5, 2020
“According to a 2018 report from the National Low Income Housing Coalition, there is no state, metropolitan area, or county where a worker earning that state’s prevailing state minimum wage or federal minimum wage can afford to rent a two-bedroom home for their family working a standard 40-hour week.”
If you find this quote unsettling, then you should probably read this book. Once upon a time, I sought side hustles on MTurk. It was mind numbing, tedious and extremely low pay (on the order of receiving 2 cents for researching a specific pizza restaurant and finding three words to describe it for consumers, or 10 cents for filling out a 20min survey, etc). I abandoned that project in favor of amassing a combination of real-world part-time jobs to supplement my income, but coming across this book reminded me of that time.
Ghost Work is well resourced, approachable reading and provides a good overview of the typical workings of our ever-growing gig-economy. Some of the details are multi-level disturbing (E.g., If you would like to get paid for your work on MTurk, you must wait for 10 days and then you can get paid the full value of your work in the form of an Amazon.com gift card, but if you want actual real-world money you have to pay a transfer fee to Amazon for the privilege), but then, that’s capitalism for ya.
If you find this quote unsettling, then you should probably read this book. Once upon a time, I sought side hustles on MTurk. It was mind numbing, tedious and extremely low pay (on the order of receiving 2 cents for researching a specific pizza restaurant and finding three words to describe it for consumers, or 10 cents for filling out a 20min survey, etc). I abandoned that project in favor of amassing a combination of real-world part-time jobs to supplement my income, but coming across this book reminded me of that time.
Ghost Work is well resourced, approachable reading and provides a good overview of the typical workings of our ever-growing gig-economy. Some of the details are multi-level disturbing (E.g., If you would like to get paid for your work on MTurk, you must wait for 10 days and then you can get paid the full value of your work in the form of an Amazon.com gift card, but if you want actual real-world money you have to pay a transfer fee to Amazon for the privilege), but then, that’s capitalism for ya.
October 14, 2020
I am a little biased, because in another lifetime, Mary was my professor and, to be honest, just about the only one I really liked in my whole PhD experience! I respect her greatly and apparently now she's a MacArthur fellow, which is unsurprising, but still really, really cool!
So, first, my quibbles with the book: I wished there had been a little more specific ethnography and a little more discussion of theory and impact. For example, there is a lot of discussion in this book about API's as a huge revolution, but I wish there had been a bit of discussion about why this might be significant not in terms of the ghost work labor force but also in terms of technology and tech culture, and how API's work within a narrative of coding history, IP, etc., and how that might tie into the bigger picture of ghost work. I also think that it might have been better to focus EITHER on American ghost workers OR on ghost workers in India, as either context would have made for a really deep and interesting study. Sometimes the writing was overly dry, which thwarted the authors' goals of humanizing the "ghosts" in these machines.
Either way, this is a critical but under-discussed topic relevant to labor. I really hope to see more work, academic and otherwise, on the hidden, low-paying ghost work underlying so many digital experiences.
So, first, my quibbles with the book: I wished there had been a little more specific ethnography and a little more discussion of theory and impact. For example, there is a lot of discussion in this book about API's as a huge revolution, but I wish there had been a bit of discussion about why this might be significant not in terms of the ghost work labor force but also in terms of technology and tech culture, and how API's work within a narrative of coding history, IP, etc., and how that might tie into the bigger picture of ghost work. I also think that it might have been better to focus EITHER on American ghost workers OR on ghost workers in India, as either context would have made for a really deep and interesting study. Sometimes the writing was overly dry, which thwarted the authors' goals of humanizing the "ghosts" in these machines.
Either way, this is a critical but under-discussed topic relevant to labor. I really hope to see more work, academic and otherwise, on the hidden, low-paying ghost work underlying so many digital experiences.
Displaying 1 - 30 of 74 reviews