Gigged Read online

  This event was set for February 26, 2016. Drivers began to organize not just along city lines, but among groups with which they shared culture and language. And all of these groups connected with each other. “We have the Ethiopian community, we have Arabic community, we have Brazilian,” Mario said. “We have everything.”

  But February 26 came and went without news about a strike. Mario stopped returning my calls and text messages. Abdoul Diallo, who had been organizing drivers in New York, told me he’d decided not to participate in the February 26 protest. “Doing this for two years,” he said, “you learn what works and does not work, and honestly, we’ve had very effective strikes in New York. But is it causing Uber to change its policies? Is it causing them to change what is harming us? And the truth is, it’s not. We’ve seen them change things before, but then they find five other ways to still screw the drivers. They change one policy, they’ll find five other ways to implement the same thing that is so-called changed.”

  * * *

  Traditional unions had two options if they wanted to organize Uber drivers. The first was to argue that drivers were being treated as employees, and as employees, they could organize under federal collective bargaining laws. The second was to find some way outside of those laws to organize independent contractors.

  A local chapter of the Teamsters in Seattle lobbied for a law that would allow Uber drivers to form a union. It passed. (Shortly later, the US Chamber of Commerce sued the city, saying the law conflicted with anti-trust law.) In Germany, a group of workers’ organizations created a list of best practices that, as of 2017, eight crowdsourcing companies had pledged to follow. Together they established an office where workers could report violations of this code.16

  Regardless of the strategy, long-established unions faced the same problem as novice organizers like Abe, which was that they didn’t know who worked for gig economy companies. The International Association of Machinists and Aerospace Workers union in New York City solved this problem by striking a deal with Uber in which Uber officials agreed to hand over contact information for its New York City drivers. This enabled the union to form an “Independent Drivers’ Guild” through which drivers could appeal Uber’s decisions to deactivate them.

  But the reaction of the labor movement to this announcement was anything but unified. In its partnership with Uber, the union agreed to forgo challenges to Uber’s classification of those drivers as independent contractors. The not-quite-union Drivers’ Guild would still not bargain over a contract, which wasn’t surprising, as independent contractors working together to set prices for their services could be considered collusion. While some saw Uber’s agreement with the Guild as a step in the right direction, others saw it as a public relations effort that protected Uber rather than its drivers. New York Taxi Workers Alliance executive director Bhairavi Desai called it a “historic betrayal” and promptly filed a lawsuit with ten Uber drivers that accused Uber of misclassifying workers as independent contractors.17 Abdoul Diallo, the Uber driver who helped organize other drivers in New York, said the Drivers’ Guild sounded “bogus,”18 and he encouraged drivers to instead sign cards that would allow the Amalgamated Transit Union to represent them.

  As unions searched for the best approach to organizing Uber’s drivers, Uber noted it was making improvements to its gig. In the summer of 2016, the company promoted a partnership with the online radio service Pandora as a benefit to workers. The new deal allowed Uber drivers to use the premium version of Pandora for free. Another feature allowed drivers to pause ride requests when they wanted to take a lunch break, instead of simply denying a request (which might eventually lead to being deactivated from the platform if it happened too often), and another paid drivers for waiting to pick up a passenger for longer than two minutes. Needless to say, these efforts seemed shallow.

  At that point, Uber still hadn’t added a tipping feature to its apps, though nearly every group of organized drivers had requested one. Tips don’t necessarily make their recipients better off, but they are standard in the livery and taxi businesses, and it would have been easy for Uber to add the option to its customer app. Uber had always resisted because it wanted transactions to be one step easier for its customers. “Riders tell us that one of the things they like most about Uber is that it’s hassle-free,” the company wrote in a 2016 blog post. “And that’s how we intend to keep it.”

  Uber didn’t give in on tipping until it faced a public relations crisis in 2017. Susan Fowler, a former employee at the company, published a blog post that documented a culture of sexual harassment and misogyny at the company, which led to an internal investigation that ultimately resulted in the firing of 20 employees and the ousting of Travis Kalanick, who resigned amid a shareholder revolt. The company was so desperate to mitigate the impact of this news that on a steamy summer day in New York City, it plopped a giant block of ice near Union Square, inside of which were “collectable cones” that could be taken to McDonald’s for ice cream (Uber had in previous years delivered free ice cream in the summer, but this was a truly over-the-top display). Nearby, Uber staff distributed happy-toned pastel shirts to random people on the street.

  Around the same time as this summer branding campaign, Uber declared a commitment to “180 days of change” in which it would improve its experience for drivers—in addition to a new tipping option, it added extra payments for drivers who traveled a long distance to pick up a rider (previously time that went unpaid) and notifications that told drivers, before they agreed to accept a ride, when the trip would take longer than 45 minutes (which meant they could avoid ending up in the middle of nowhere, with a long return trip they would have to make on their own dime).

  These improvements still fell far short of what some drivers had demanded. But the distributed nature and independent status of Uber’s workforce made it difficult to organize, and aside from the PR crisis that made the company desperate to improve its image, Uber had little business reason to give drivers more control over their micro-businesses or to honor their other requests. Uber’s service simply moved people from point A to point B, and it wasn’t clear that drivers could provide a meaningfully better service if they were happier with Uber. Nor was it clear that customers would pay more for it. As long as Uber could find as many drivers as it needed (which was easier in a down economy), the company didn’t have a financial incentive to make the driver experience a huge priority.

  An executive who Uber unsuccessfully tried to recruit in 2016 told The Guardian that during his job interview, Uber’s chief product officer had responded to a question about how the company would handle the discontent among its drivers by saying, “Well, we’re just going to replace them all with robots” (an Uber spokesman told the paper that its executive did not recall making the statement).19

  * * *

  On her applications to universities, Kristy had described her Mechanical Turk work as a “crowdsourcing micro-contractor” position, a job that she noted included working with several Fortune 500 companies. She hoped to study psychology.

  Mechanical Turk had shown Kristy how close many jobs were to being automated. She’d been part of a crowd that helped train machines to do things like recognize images and diagnose diseases, and she knew that someday those algorithms wouldn’t need training anymore. They’d replace the humans currently doing the work. As far as she could tell, though, people would always want a therapist to offer a real human connection.

  With her husband back at work, Kristy had been able to save the money she earned from Mechanical Turk for her $10,000 annual tuition. She hoped that a degree would finally lead to work that would allow her to save for retirement.

  Kristy started classes at Ryerson University in 2012. Most of her classmates were in their early twenties, and she sometimes feared she was the weird old person in the class, the one who sat in the front row and always raised her hand. Sometimes she participated in her classmates’ Facebook groups, using an anonymous account she had s
et up for the purpose of entering contests. In person, she had a more difficult time making friends with them.

  Though Kristy had stopped working on Mechanical Turk full-time when she started school, she had continued working as the moderator of Turker Nation. The forum kept her in contact with her fellow Turkers, her friends. It also sparked her interest in activism.

  On the forum, she’d connected with academic researchers who were interested in Mechanical Turk work. Some, like Stanford’s Michael Bernstein, studied how crowd work might be applied to bigger and more complicated problems in the future. Others, like Six Silberman and Lilly Irani, who had built a tool used by Mechanical Turk workers to rate and review clients, studied what the emergence of crowd work might mean for workers. Kristy agreed to speak with a group of these researchers when they visited Toronto for a conference.

  They met over dinner at Kristy’s favorite Indian restaurant. And as they discussed Mechanical Turk, it wasn’t long before Oscar Smith came up. “Smith” (probably a pseudonym) used Mechanical Turk to transcribe thousands and thousands of business cards. He was, according to Kristy, notorious for bad pay—he only offered a penny per card—and Mechanical Turk workers suspected he worked at a professional social networking site that, at the time, offered a service in which customers could take photos of business cards in order to automatically add them to their contacts. The casual discussion turned to what Turkers could do to convince Oscar Smith to raise his pay.

  Everyone at the table well knew that it would not be easy to organize Mechanical Turk workers in such an effort. Some were hesitant to challenge Amazon’s policies, if for no other reason than Amazon could easily shut the whole site down. In its annual report, which documented $136 billion in net sales, Amazon didn’t even mention its crowd work platform. The company was known for rapidly launching and closing new businesses—grocery delivery, home cleaning, a Pinterest-like shopping platform—and one imagined it would have few qualms about shuttering Mechanical Turk. Any attempt at collective action would need to be managed carefully. It would also be made difficult by the fact that Mechanical Turk workers were distributed all over the world, spoke many languages, and had significantly different ideas about how many dollars translated to a living wage. Traditional unions wouldn’t work.

  But maybe there were other forms of collective action that would. What if workers could hold bad employers’ work hostage, as a “strike” of sorts? They’d sign up to do the job, but not actually do the work. After the allotted time expired, when the tasks would be released back to the platform, another member of the resistance could pick up the same work and continue to hold it. Because they hadn’t actually completed the work, Oscar Smith would not have a chance to reject that work, and therefore could not impact their ability to get other work in the future.

  Or maybe Mechanical Turk workers could accept all of Oscar Smith’s tasks, but do the work in a slightly wrong way. The phone number would be off by one digit. The address would have the wrong zip code. Oscar Smith might not notice in time to leave the Mechanical Turk worker a bad rating. His customers would.

  Though the conversation had started as a hypothetical exercise, the researchers and Kristy alike started to see potential in some of these ideas. Sabotaging Oscar Smith’s work might be too extreme, but the strategy of collective action made sense.

  Kristy agreed to keep in touch with the people she referred to as “the academics.” She asked for feedback on the ideas from other Turkers (not all were pleased, fearing retribution from Amazon) and eventually helped researchers launch a website called Dynamo. Lilly Irani, Michael Bernstein, and Niloufar Salehi, a Stanford graduate student, led the project along with two other Stanford students and Kristy, who was an author on the eventual paper.

  Dynamo was designed to help Mechanical Turk workers plan a movement. Here’s how it worked: Its creators posted a task on Mechanical Turk that was open to anyone who had completed 100 jobs or more. In it, the researchers explained that they were paying the Turkers for a five-minute-long vacation. Turkers could decide either to do nothing, and still collect their pay, or to visit Dynamo and sign up with a code provided. This served a dual purpose of recruitment and screening—only active Mechanical Turk workers would have the code—without violating workers’ desire to be anonymous (which was legitimate, as union laws would not protect them if Amazon decided to kick them off of Mechanical Turk in response to their participation).

  Inside Dynamo, anyone could start a “campaign” by writing a description of it. Members could vote ideas up or down, and when an idea collected at least 25 positive votes, then it would be moved to a project space with tools for outlining action steps and tracking progress. The site launched in 2014.

  Around the same time, Kristy got invited to her first big public speaking event. She had emailed a Microsoft researcher after reading about her work with crowdsourcing platforms (Kristy emailed anyone she read about who seemed interesting), and that researcher had introduced her to a Pittsburgh conference organizer. When she was first asked to speak, she didn’t really know what to say. So she set her session up as an interview, with Michael Bernstein asking her questions about Mechanical Turk.

  It was the first time she’d been to Pittsburgh, and she wasn’t sure how she’d react to speaking to a large group of people, many of whom had doctoral degrees. But as she started to talk about her experiences as a worker on Mechanical Turk, she realized that she was actually enjoying herself.

  “Oh my god,” she thought. “I’m good at this. And I’m not scared at all.”

  * * *

  Six months in, Dynamo had 470 registered users. Of 22 proposed ideas, two had made it to the point of action.20 Dynamo’s first campaign created a set of ethical guidelines designed to curb bad behavior by academic researchers, who used the site as a cheap way to recruit research participants (by 2015, more than 1,120 academic studies on Google Scholar would include Mechanical Turk workers as participants).21 Despite Mechanical Turk’s growing popularity as a research tool, study ethics boards did not always hold these experiments and surveys to the same standards as research conducted in person.

  In one example that particularly irked Kristy, a graduate student at the University of Washington had devised a study in which he recruited Turkers to tag photos without disclosing that they were part of a study. Unsuspecting Turkers who signed up were first met with nice pictures of animals, much like the ones I had spent hours tagging in my own Mechanical Turk experiment. Then, sometimes, the photos grew more gruesome. One Turker wrote on Turkopticon, a tool that Mechanical Turk workers use to share reviews of task posters, that the photos were “90% pretty cute kittens, rays of sunshine, cupcakes and brownies” and the other 10% were “the polar opposite—dismembered children in the streets, burn victims, amputations, decaying corpses, etc.” The graduate student and his fellow University of Washington researchers wrote a paper based on the research and in a draft explained that they were testing “differences in labor supplied depending on the ‘agreeableness’ or ‘disagreeableness.’” They’d found images for the study by searching Google Images for topics such as “amputations, autopsy, broken limbs, gangrene, and larvas to name a few.”22

  Though some Turkers said they didn’t really mind being used as lab rats, just so long as the job paid, other workers believed that studies like this one wouldn’t exist if study ethics boards (like the institutional review board that approved the disagreeable image experiment) were better informed. Throughout the first six months of the ethical guidelines campaign, 171 workers and 45 task requesters signed the formal, academically worded guidelines.23

  The list of signatures begins, thanks to Dynamo’s pseudonymous nature and automatically generated usernames (which uses the convention “adjective + animal), like this:

  Gorgeous monarch butterfly (14 August 2014)

  Courageous cockroach (14 August 2014)

  Fancy cod (15 August 2014)

  Faithful fly (15 August 2014)

>   Dark bird of paradise (15 August 2014)

  Elated sea urchin (15 August 2014)

  Lonely wombat (15 August 2014)

  Amused hedgehog (15 August 2014)

  Jolly otter (16 August 2014)

  Terrible cat (16 August 2014)

  It was not the most authoritative list of names, no, but it was something.

  Kristy proposed the second Dynamo campaign. Her biggest concern was that employers on the platform sometimes did not treat workers like humans, partly because Amazon’s description of the service didn’t frame them that way or acknowledge their individual existence. Amazon’s slogan for Mechanical Turk is “Artificial artificial intelligence.” The Mechanical Turk website homepage used a simple flowchart graphic to explain to new workers how the gig worked: Find an interesting task-> Work-> Earn Money. “Work” was represented with an image of wheels turning, rather than by a human worker. Unlike other gig economy platforms, employers never saw the name of the people who worked for them through Mechanical Turk. “When you have something that is not humanized, like your refrigerator, you have no empathy for it,” Kristy said. “You don’t care if you slam the door. You don’t care if you leave moldy food in it. It doesn’t matter. You don’t care about the refrigerator.”

  Her solution to being treated as less than human by clients on Mechanical Turk was simple: Ask Turkers to explain the problem in letters to Jeff Bezos, the CEO of Amazon. The bullet points on Kristy’s campaign memo, below a “WHAT WE WANT TO SAY” headline, read as follows:

  1:Turkers are human beings, not algorithmsb, and should be marketed accordingly.

  2:Turkers should not be sold as cheap labour, but instead skilled, flexible labour which needs to be respected.

  3:Turkers need to have a method of representing themselves to Requesters [people who hire Mechanical Turk workers] and the world via Amazon.

  Kristy’s own letter was heartfelt and polite. It made a business case for treating workers better, not unlike the case that Dan and Saman had once put together for Managed by Q.