The following op-ed is based on the results of the Fairwork Cloudwork 2021 report by Kelle Howson, Hannah Johnston, Nancy Salem, Robbie Warin, Fabian Ferrari, Yihan Zhu, Pablo Aguera Reneses, Funda Ustek-Spilda, Daniel Arubayi, Srujana Katta, Shelly Steward, Matthew Cole, Adam Badger, and Mark Graham.
When Melissa graduated from high school in Sao Paulo, she was eager to start working. But in one of the countries hardest hit by the pandemic, finding a job proved nearly impossible. Until one day, a Reddit post led her to Amazon Mechanical Turk (MTurk). MTurk is one of the largest online work (or cloudwork) platforms in the world, owned by global giant Amazon. Anyone can sign up to be a ‘Turker’, and complete short-term jobs, also known as microtasks, for money. The tasks often require little expertise or training, and can range from labelling data, which might then be used to train AI systems, to completing surveys. More and more Brazilians like Melissa are joining this global virtual labour market.
Working for MTurk and other online work platforms, has its benefits. This type of work can be performed from anywhere in the world with an internet connection. Workers are able to work flexible hours, choosing when and how to work, as well as the type of jobs they apply for. “My favourite thing about the work is the flexibility” says Melissa who likes to work late at night. She also feels safer working from home, which allows her to tend to her caring responsibilities: “I have an old grandma who I live with, and I don’t have to put her at risk of getting COVID by going to work.”
However, though it may be flexible, work on these platforms is also volatile. Melissa found herself competing against millions of workers for a limited number of jobs. At the Fairwork Project, we surveyed 792 cloudworkers in 75 countries and found most of them spend long periods of time searching and applying for available jobs —time they are not compensated for. Pay also varies widely between jobs, and can be really unpredictable. Only 2 of the 17 platforms analysed by Fairwork could prove that all their workers earn above their local minimum wage. Even worse, most workers are not guaranteed to be paid on time, or even if they will get paid at all for their work. As Melissa said:
“I’ve been scammed a few times with people just refusing to pay me even though the work was perfect. There’s not much to do […] I tried emailing Amazon, but I never got a response and this is really common.”
In the absence of the necessary protections by the platform, Turkers have been relying on each other to flag exploitative practices by employers through tools like Turkopticon. Turkopticon is a plugin that allows workers to rate jobs on MTurk, and which has grown into a vibrant community advocating for better working conditions on the platform.
Cloudwork platforms might appear as neutral intermediaries, connecting workers with clients. But they are best understood as the latest iteration of a longstanding trend of business process outsourcing (BPO). Research suggests a majority of cloudworkers are based in the Global South, while most of the clients are corporations in Europe and the US. All sorts of organisations, from consulting firms and banks to top ranking universities, increasingly rely on these platforms to outsource work to freelance workers thousands of kilometres away. Strong English language skills and its American time zone make Brazil a popular market for these outsourcing practices. But while the cloudwork labour market provides valuable income to many in the country, it is also defined by unfair pay and conditions. The seeming geographical disembeddedness of cloudwork platforms, which are often based in industrialised countries, enables them to circumvent local regulations and worker protections and makes it harder for their scattered workforce to appeal unfair practices or organise for better conditions. In fact, only 3 platforms involved in the study could evidence that their workers have access to fair representation processes and that freedom of association was not inhibited.
However, not all cloudwork platforms are the same. At the Fairwork project, we rated the working of the 17 platforms based on 5 principles of fair cloudwork (fair pay, fair conditions, fair contracts, fair management, and fair representation). The third best scoring platform, with 5 out of 10 points, was the Latin American platform Workana. Workana is a freelancing, or microtask, platform headquartered in Buenos Aires, with offices also in Brazil, Colombia, Mexico, and Malaysia. If the more than 2.5 million workers registered to the platform were classified as employees, it would be the biggest private employer in the world. Through our research we found workers on Workana were often usure of how to communicate with management or how to appeal unfair decisions. After engaging with Workana managers for over 10 months, the platform agreed to implement multiple changes and ratify policies that will have a direct impact in the lives of their workers. For instance, Workana adopted a policy stating that jobs would be removed from the platform if they paid below the local minimum wage, and made clearer the process to appeal account terminations. These changes prove how platforms like Workana or MTurk, which scored just 1 out of 10 on our ratings, have the ability to improve the working conditions of their workers.
The problem is that platforms have few check and balances holding them to account. Worker-power is suppressed in cloudwork, and platforms evade the reach of regulators. Platforms can do so much more to protect workers, but cloudwork conditions will not meaningfully improve until platforms are held accountable for their practices by national and global regulations, and by workers’ collective power. Projects like Fairwork can support this drive towards a fairer future of cloudwork. Given the growing number of people, like Melissa, who rely on this type of work for a living, it is a priority that platforms, regulators and worker collectives work together to ensure basic standards of fairness.