Amazon recently announced—and for the first time—the earnings of its cloud storage and computing business, Amazon Web Services (AWS). The report from the company pegged AWS revenue for the fourth quarter of 2014 at $1.6 billion, contributing $256 million to the company’s bottom line. The service, which drives the back-end technology of companies including Netflix, Intuit, and many others, has received much attention as the leader in the growing trend toward cloud storage.
However, less attention is being paid to another cloud-based Amazon offering that may have an even larger impact on peoples’ lives: Amazon’s Mechanical Turk. Described as a “marketplace for work,” Mechanical Turk Workers (Turkers) complete a variety of tasks (known as “HITs,” or Human Intelligence Tasks) online. Each HIT, is a discrete piece of work with an associated pay amount and once the work is submitted and approved by the requester, the worker is paid at that task-specific amount.
The tasks range from very simple data processing or image tagging—which can be completed in a matter of minutes and pay as little as 1 cent per HIT—to more complex transcription, writing, or research tasks, which can take a couple of hours and pay $25-$50 or more. Some tasks require little or no experience, while others require the worker to pass a certification test, be located in a particular region, or have specific qualifications or a certain success rate in past work. As I’m writing this, there are currently over 300,000 tasks awaiting completion, and Amazon says there are more than 500,000 workers from 190 countries in the system. From 2009-2014 more than 130 million tasks were posted to Amazon Mechanical Turk, according to a recent study.
“While it’s hard to tell just how many hours of work are available overall, it is possible to do work online and get paid.”
I recently spent several hours, spread out over the course of a few weeks, trying my hand as a Turker. In that time, I’ve completed a variety of HITs, including: entering data from scanned-in receipts, tagging images of food, completing surveys, analyzing blog posts for movie review content, and, perhaps oddest of all, identifying whether or not particular images contained a chair. I’ve completed 270 tasks and earned almost $6.50, or an average of 2.4 cents per HIT. Though I’ve taken many qualification tests, I have not (yet!) been able to break into the higher-paying tasks, many of which are reserved for “masters” who have completed 1,000 or more HITs.
Thinking about online work more generally, I’d offer a couple observations based on my experience, as well as broader labor market trends, keeping in mind that these may be more or less relevant depending on the work platform.
- There is work to be done online. While it’s hard to tell just how many hours of work are available overall, it is possible to do work online and get paid. The exact income potential for this work, however, is an open question. In my experience, the hourly wage equivalent from working through this platform is well below any reasonable developed-country minimum wage—though the hourly equivalent pay certainly increases with higher-level tasks. It’s also worth noting that other platforms may offer higher per-task rewards, as well. If and when more work moves online, there may be a growing demand for workers, and this might push up compensation. However, it’s also easy to see how developments in artificial intelligence could begin to compete with humans for these kinds of simple tasks.
- Having a job in the real world is often the best way to learn new skills, both job-specific and more generalizable to other employers and work situations. Work performed through Mechanical Turk, however, is a step towards commodification of the work itself—tasks are typically broken down to such an extent that they require little skill to complete, and thus few skills are built up in the process of working. One could potentially apply skills learned off-platform to more advanced tasks, but that responsibility—and cost—is placed on the shoulders of the workers. For online work to provide skill or training benefits, a more deliberate strategy to augment online work must accompany the work itself—one Rockefeller Foundation partner, CloudFactory, has taken this approach to supplement online work with off-line skill development for their workers in Nepal and Kenya.
- There is virtually no relationship between Turkers and task “requesters.” However, the more time I spent completing online tasks, the better sense I got of what kinds of tasks and requesters would provide better compensation for my time and effort. But something is still lost in not having any significant feedback, investment by an employer in a worker’s success, or a personal connection to the employer. Many of the tasks are broken down to such an extent that there is no real connection to the overall purpose of the work or mission of the requester. I never felt like I knew why I was doing a task or what it would ultimately accomplish. No doubt that this impacts productivity, along with virtual turnover of workers across requesters.
This experience has led me to wonder if this kind of task-based, cloud-based work can ever be a substitute for a more traditional full- or part-time job. Studies on work and happiness show that part-time employment can be nearly as rewarding as full-time—but how many hours constitute “part-time” when that work is online? Is home- or online-based work an acceptable substitute for being at a worksite? What skills are actually being developed through work? What is the overall impact of working for multiple, anonymous employers as opposed to a steady employer relationship? These are just a few questions that require further exploration in order to realize the full potential of online work for improving livelihoods—and with support from The Rockefeller Foundation, the World Bank will soon release a study looking at the challenges and opportunities for this emerging part of the labor market.
“This experience has led me to wonder if this kind of task-based, cloud-based work can ever be a substitute for a more traditional full- or part-time job.”
The Amazon Mechanical Turk case may prove to be more of an extreme example; other platforms undoubtedly provide more sophisticated work that could build skills and a solid employer-employee relationship with greater levels of engagement. But while many of the benefits of traditional jobs come along more naturally, online work efforts will need to be more deliberate in their design to ensure that people truly benefit from the work experience.