Are crowdsourcing platforms like Amazon’s Mechanical Turk as bleak as the company’s shipping warehouses?
In the summer of 2009, the U.S. economy lost 9 million jobs: between April and October of that year, the national unemployment rate would rise to 10 percent as the stock market plummeted to nearly half its value.
Money stood at the forefront of collective anxiety: every day seemed to generate new tales of friends getting laid off or of more companies’ having closed up shop. That summer, I discovered people were resorting to making money online using a service called Amazon Mechanical Turk, or “MTurk” as it’s colloquially known. MTurk was started in 2005 by Amazon CEO Jeff Bezos and director of Amazon Web services Peter Cohen as a way to solve problems with Amazon’s ever-expanding data set. The premise was that a distributed crowd of humans could easily complete tasks that computers found too challenging. The jobs, called Human Intelligence Tasks (HITs) on the service, might be to match the name of an item—say a 95-ounce box of Tide detergent—to the correct product image. The typical pay for HITs like this range from $0.01 to $0.20 and often need to be completed within a limited amount of time.
With my curiosity piqued, I began surveying workers on MTurk, asking them to tell their stories through short memoirs. What started as research about a tool I might use in my artistic practice became a much deeper experience. The stories I heard were parables of everyday life, success, and struggle. Now, five years later, I came back to them to see if the story of crowdsourced labor has changed.
MTurk drew its name and conceptual model from the 18th- entury invention created by Hungarian nobleman Wolfgang von Kempelen, who created a mechanical chess-playing automaton that supposedly “defeated nearly every opponent it faced.” In truth, it was a hoax: A real, human chess master was hiding inside the machine.
To extend this conceit to Amazon’s platform, Bezos coined a clever, glossy euphemism, describing the service as “artificial artificial intelligence.” Looking through the FAQ page for MTurk gives a better sense of what this artificial artificial intelligence might entail:
When we think of interfaces between human beings and computers, we usually assume that the human being is the one requesting that a task be completed, and the computer is completing the task and providing the results. What if this process were reversed and a computer program could ask a human being to perform a task and return the results? What if it could coordinate many human beings to perform a task?
At first, MTurk seemed appealing as a tool that could complete work like any other software with the unique exception of being powered by an unnamed, globally distributed group of people. I envisioned doing a kind of conceptual exploration of this virtual workspace, which could then lead to future collaborative projects with the platform. But soon, I found myself preoccupied by a truly basic question: Who were the people fulfilling these requests? Who were the chess players within the machine?
My hunch was that the workers using MTurk were middle-class skilled workers like myself. To test this hypothesis, I used MTurk to hire some of them to tell me why they were on it. Since MTurk tasks needed to take only minutes to complete, I requested a brief, 250-word account and let them know that I would share their story on a Tumblr, which I titled The Mechanical Turk Diaries. I decided to pay $0.25 per story, which at the time seemed like a high rate relative to other HITs on the platform.
Soon after I posted, answers started coming in:
How many synonyms can you name for the word ‘broke’? Here are a few: unemployed; poor; desperate; discouraged. Yep. Being a displaced housewife after raising the kids is not financially healthy in this economy. Even when I worked at a bookstore (now closed), I didn’t quite make enough money to pay the bills. So, I started trying to make a little extra cash through MTurk.
The answers that followed had variety but hewed to a consistent theme. It was true; some of the unemployed were turning to MTurk to keep them going through the pains of finding a new job and surviving the crisis. In the 323 stories I received, people mentioned “money,” “cash,” “dollars,” and “bucks” more than 600 times.
The service also seemed to offer some kind of emotional reprieve from job hunting for those hit hard by the recession. It relived loneliness, and boredom at work for those who still had a job. One woman shared a story about how her husband introduced her to MTurk while she was in the hospital being treated for cancer and how it helped her “keep [her] sanity during a very hard time and helped [her] to stay busy.”
By the end of the summer, I had come to understand the Turking community as composed of pragmatists who preferred to be productive in lieu of being bored. For many, MTurk was simply a place they could work online from home without getting scammed.
I wondered, however, about the sustainability of a full-time income pegged to pennies, far from minimum wage. What would happen to workers who started to depend on MTurk, like the unemployed student and mother living in Los Angeles who described herself as “desperate for work”? While initially glad to find “an awesome website dedicated to helping people find legitimate online work without being ripped off,” she also highlighted the stress of having MTurk as one of her only sources of income, adding that well-paying HITs were hard to find. Meanwhile, she was struggling to “pay [her] late car note and get a tank of gas, rent a textbook for the summer session and buy bread.”
As more stories came in, I noticed more workers raising interesting questions about the platform and their abilities to subsist on far less than a living wage. Toward the end of the summer, on August 21, 2009, a worker posted:
The main thing I want to say is, why can’t there be a place like this that actually offers the chance of making five or seven or ten dollars an hour? A place where some of these real clerical jobs, performed for a penny or a nickel, could be done for a quarter and moms and dads and singles and jobless teens and retirees and all of us Americans could actually make Turking into the wad of gum that patches the hole in the dike. I did this HIT even though it is late, because it pays a quarter. And I appreciate that. Keep on Turkin’ and stay well.
Amazon makes no effort to set a fair wage on MTurk. In fact, their Terms of Service clearly spurn liability for what requesters post as a HIT or what workers submit when the HITs are completed. At best, a worker can flag a request that violates the site’s policies. Outside of providing tax information to set up an account as a worker, Amazon is free of any tax implications for the payment, no matter how small or great. Perhaps even more surprising is that a requester can take the worker’s submission and refuse payment without any consequence. The system frequently puts workers at a disadvantage. Lilly Irani, a professor at UC San Diego, rallied workers in 2008, which eventually led to a browser plugin called Turkopticon to address some problems, but change seems like a long-term battle.
The mother in L.A. finished her story by saying she “hopes Amazon appreciates their Turks because we so are not getting any love in our inbox. No thanks for remotely accessing the sweatshop at a cost higher than I can afford. Oh well back to Turking, Turking, Turking day and night.”
Her complaint was real. It costs more to have an Internet connection in your house per month than the casual worker on MTurk can make in 30 days. She also wasn’t the first to describe the service as a “sweatshop”: In a 2006 essay about MTurk, Salon’s Katharine Mieszkowski asked if the service was “a boon for the bored or a virtual sweatshop?” More recently, The Nation declared Turkers “one of the most exploited workforces no one has ever seen.”
It wasn’t supposed to be like this. Journalist Fred Howe, who coined the term crowdsourcing, imagined that part of the innovation of crowdsourcing would be the “ability for communities to form out of shared interest, a shared passion for a hobby, a craft, an art.” In reality, because workers and requesters are anonymous on MTurk, it seems to lack a key component for community nurturing.
In a 2010 panel on crowdsourcing, Lukas Biewald, CEO of MTurk competitor CrowdFlower, highlighted this aspect of the platform’s relationship to workers, reframing it as an attractive feature for employers: “With technology you can find them, pay them the tiny amount of money, and then get rid of them when you don’t need them anymore,” he said.
***
Recently, I resumed collecting memoirs from MTurkers. Had the people working on the platform changed since 2010? I found my original survey from five years ago, logged back into my MTurk account, and posted a few HITs. Were people still thinking about the economy?
My first batch of HITs returned nothing. I set the pay to $0.50, double the original rate, which many workers had thought was generous. Then, looking at the configuration I noticed default settings that restricted the workers’ level to “Master Qualifications.” This was new. According to Amazon, Masters are “workers who have demonstrated accuracy on specific types of HITs on the Mechanical Turk marketplace.” This was part of a change Amazon made in 2012 to improve the pay for work on the site and prevent better workers from exposure to scammers. If it works, it’s a bright spot for experience Turkers, but also a potential barrier to higher quality HITs for new workers.
My experience tracks the findings of Panos Ipeirotis, a business professor at NYU, about the changing nature of the MTurk workforce. In 2008, he conducted a survey on the platform to disprove common beliefs about crowdsourcing. He discovered the workforce on MTurk was predominantly U.S.-based and, surprisingly, almost 60% of the users were from 21 to 40 years old and female. Only about 13% of the HITs were completed by workers outside the U.S.: 8% from India, and a little more than 5% from the U.K. and Canada.
In 2010, Ipeirotis updated his demographic study to reveal some interesting shifts. He discovered that the percentage of U.S. workers had dropped below 50% while workers from India had risen to 34%. He also found that the Indian workers often considered their work on MTurk a primary source of income. Since India’s minimum wage is around $0.28 per hour, as reported by Business Insider last year, the shift from it being a source of extra money to a primary source of income wasn’t too surprising. As platforms like MTurk proliferate, the nearly invisible microtasks that underpin so many sites on the Internet are being outsourced to the now more accessible and cost-effective crowd around the globe. This made the stories about minimum wage from workers on MTurk in the U.S. ring even louder.
After raising my rate to $1.00, I was able to get responses to my requests and more stories from the workers on MTurk. They were very different this time around:
It was four years ago when the economic crisis hit the world badly and it was at the same time that the Telangana Agitation hit our state (Andhra Pradesh, India) very badly. Market in general in our state touched its lowest point as people were very wary about what would happen next. There were continuous strikes, boycotts, agitations etc. to the extent that all that had become an order of the day. I am into real estate and this is the sector which was largely affected by the ongoing agitation and world economic crisis. I was vigorously searching for any part time job which would support my finances and take care of my overheads during the period of this agitation. I happened to find an ad in the newspaper highlighting “WORK FROM HOME” and I immediately contacted the person who published the ad.
I had never thought of crowdsourced work on MTurk as way to continue working while coping with economic and civil unrest. It seemed the platform had turned into a way to become telepresent in the global economy when conditions were too rough to leave the house. But other stories from the new survey felt similar to the ones in 2009: people finding ways to make small amounts of extra cash or build up hope to get through to the next big job.
The conversation around these platforms is only going to grow louder as more people all over the world turn to them as a primary source of income. In a recent article for The Nation, Moshe Z. Marvit interviewed struggling workers on MTurk and mentioned an upcoming and critical class action lawsuit against CrowdFlower. Early initiatives, like Irani’s Turkopticon, are still helping draw attention to the ethics of paid microlabor. Yet the future of crowdsourced labor still looks strange. More mobile apps are connecting to crowdsourcing platforms, putting microtasks at your fingertips in your idle time. Imagine a future wherein an app like Candy Crush Saga is a way to earn money for anniversary presents while simultaneously working in the global labor market for pennies.
Since my first survey, newer platforms had emerged on which people completed tasks for money. I surveyed workers on TaskRabbit, which promises to allow users to “Get just about anything done by friendly, trustworthy people.” I quickly found a woman who was using the service as her primary source of income. She chose to remain anonymous but said her days start by checking “TaskRabbit, Craigslist, and Voices.com for gigs.” She said she tries to do “mostly virtual tasks on TaskRabbit—data entry, transcription, research, etc.” She’s an actor and has been on the platform for three years. Her only complaint: “The hiring process has become so lax. There are way too many of us competing for jobs now.”
