Slave to the keyboard: the broken promises of the gig economy

I’ve never held a job where I could feel secure in my position. From my two decades as entrepreneur to precarious contract gigs at large companies, I’ve faced having no income at all time and time again. Yet it wasn’t until I began to work at Amazon Mechanical Turk full-time that I realised how bad work can be. In fact, I believe that if we don’t move quickly to ensure gig and crowdwork is regulated and workers protected, not only will the decline of decent work opportunities increase exponentially, but our economies will begin to collapse as workers make less and less and demand dries up. It is in the best interests of corporations, governments, and citizens alike to stop the relentless push for cheaper, faster, more exploited labour.

Most people haven’t heard of Amazon Mechanical Turk, or ‘mTurk’, but getting the word out about the working conditions there is a necessity for those of us who want to stop its labour blueprint from being replicated elsewhere. A simple explanation of how the site functions is to imagine it as a giant cork board to which anyone can post a job. Normally, we would imagine someone who has work to do, known as a Requester on the platform, to tack up the entire job to be done. mTurk works a bit differently, in that they tear up that job into tiny pieces, known as micro-tasks or Human Intelligence Tasks (HITs), and then post each up to the board, one by one. For example, if you have an article you would like to have translated, you wouldn’t post the whole article to the board. Instead, you would divide it sentence-by-sentence, piling them on top of each other and then putting them on the board. This permits thousands of workers each to grab as many sentences as possible, completing them at the same time. While it might have taken one translator hours or days to complete a 1000-sentence article, on mTurk that same article would be completed in as little as 10 minutes. On top of the easiness of posting work online and the speed at which you can get the work done, mTurk has become a platform where low pay is the norm. Most HITs pay less than a dime, and the average pay reported by workers is US$2 per hour or less. That means that your 1000-sentence article might cost you US$20 to have translated, a bargain compared to traditional methods.

Delving deeper, mTurk’s processes raise very serious ethical questions. For example, when your translation project is complete there is no way for you to rate your workers, nor can they rate you, but you have the option either to accept their work or to reject it. If you accept, their time is paid for and their approval rating increases, giving them the ability to continue working into the future. If you reject, not only do you get to keep the work they did, but they go unpaid as their pay returns to your account, and their approval rating decreases. This not only potentially bars them from working on other projects that require a higher approval rating than they now possess, but Amazon can suspend them if their approval rating gets too low. There are Requesters who reject as a method of saving money, and many workers are harmed by their actions. Silberman and Irani (2016) make the point that this possibility for Requesters to reject ‘has effectively legalized wage theft in crowd work, as there is no way to distinguish between wage theft and legitimate and normal use of an intentionally designed platform feature’ (Silberman and Irani, 2016: 518).

Since Amazon refuses to step in and mediate disputes over rejections or suspensions, they are by proxy letting this unethical behaviour continue. Now workers walk on eggshells wondering whether today will be the day a Requester rejects their work en masse and they end up being blocked from using the platform, leaving them unemployed.

Amazon sells the platform to workers as being flexible, permitting you to work when you want, doing whichever piece of work you choose, and allowing you to pick your hourly rate, but on top of the stress of wage theft, workers also lack this promised flexibility. You can only work when work is available, so you are totally beholden to the schedule of those who post work. That could mean you have to get up at 2 am to work if tasks are posted on the platform at that time. Which work you do is dependent on how desperate you are for money, as those who cannot afford to be choosy end up doing the lowest paying and most distressing work. Lastly, Amazon doesn’t supply any metrics on the hourly rate of the work posted, so you have to complete some and calculate it yourself to know whether the task is worth doing or not. With thousands of people testing the work out to determine the hourly pay, even poorly paying work gets done just through this process. Flexibility on the platform is only afforded to those who don’t need the money, leaving the most desperate workers – those with a disability, no access to jobs locally, a criminal record, or otherwise unable to work outside the home – chained to their computer.

Yet the worst aspect of working on mTurk is the content of the work itself. Requesters are anonymous, selecting any name they wish. That means that workers have no idea who they work for, and therefore also don’t know the motive behind the work itself. For example, recently a Requester posted still shots taken from security camera footage next to a photo and asked the workers if they were the same person. They claimed that the video was taped in public schools and the photos were mug shots of sexual offenders. But how do we know that this was true? In his talk ‘Minds for Sale’, Jonathan Zittrain described a similar but imaginary scenario where the still shots presented were of protestors. Every time we chose ‘yes,’ we could have been condemning such a protestor to death. While some may believe that content moderation, tagging still shots of baskets full of heads from Daesh videos or selecting images of children being raped, is the worst work that we do, I believe that the work where we don’t know the outcome of our labour causes me more distress.

Of course, a strong union can help us in our fight for a fairer workplace, but traditionally unionising invisible online workers is difficult. Amazon doesn’t reveal how many workers are active on the platform, let alone their names or locations, nor do they offer any function on mTurk for workers to communicate with each other. Even if a group were formed and were able to organise, it is exceptionally difficult to use strikes or work disruptions as you never know which workers are actually participating. Collective bargaining for freelancers is also problematic as legislation deems it illegal cabal-like activity. So, how can a union help out crowdworkers? To begin, organisations like the Canadian Freelancer Union, associated with Unifor, are offering services to anyone who works without an employment contract. For example, they can provide legal and tax advice to workers who are often left to deal with such issues on their own, or group policies for health, home, business, and life insurance, or even pension plans. Unions can also offer tools, such as IG Metall’s FairCrowdwork.org website that provides workers with the ability to rate platforms and critique their actions. Union members are also experts in lobbying, and providing that expertise to crowdworkers can enable us to have a voice in legislative issues, pushing governments to enact protective legislation. While some methods unions leverage offline may be out of our reach, there are other resources and services such organisations can offer to crowdworkers, even new ones that we can dream up together, that will help us secure the future of work.

My hope is that unions will continue to reach out to all freelance workers to help them better their work environment. This means creating more online tools, holding rallies, and approaching workers through the platforms themselves, or even through surveys of their current members, in order to get in touch with those who need their help. Once organised, I believe these partnerships will forge new avenues of resistance, and the pressure we can apply on governments will help to push new legislation that offers crowdworkers similar protections to what employees enjoy today. If we don’t work together, I see a future where union membership dwindles as full-time positions are replaced by A.I. and crowds of underpaid labourers. It is as vital to unions that they begin to include freelancers in their ranks as it is for workers to leverage the abilities and resources of those unions. This is a beautiful partnership that can bring about great change, but only if we get together as soon as possible. Technology moves swiftly, and no one can predict what might come next which further erodes labour rights. If we stand together now against crowdworker exploitation, we can promise our children that when they are old enough to work, they won’t be facing the same struggle we are now. That is a promise worth keeping.