CrowdWorking: Exploitation on the Web
The Birth of Artificial Artificial Intelligence
Berlin, Germany They handle the tasks a computer can’t. Millions of clickworkers rate products, test beta versions, approve comments and delete pornographic content. With the easy money comes precarious low-wage jobs without benefits or security.
In 2001, NASA found an innovative answer to an age-old problem. The U.S. space agency wanted to map mass quantities of photo imaging from Mars. Machines alone could not solve the problem and it would have taken months for the agency’s employees to complete the task. NASA therefore opted for a new sort of outsourcing: crowdsourcing.
In a project called “ClickWorkers,” internet users – i.e., the “crowd” – were tasked with clicking on craters and other noteworthy features in photos NASA had made available online. With each mouse click, they helped map the surface of Mars. What was in NASA’s case touted as “citizens helping science” is now a business model of gargantuan proportions.
The crowd of the NASA project worked for free, but paid work on the basis of the crowdsourcing model is becoming ever more common. There are millions of digital wageworkers, found on crowdsourcing platforms, who take on jobs similar to that of NASA’s volunteers – jobs that cannot be computerized but can be done efficiently by a massive number of individuals.
Using People Like Computers
The best known internet platform is Mechanical Turk, abbreviated mTurk, run by Amazon. Over half a million workers are registered there, primarily from the United States and India.
mTurk was founded when Amazon incorporated CDs into its product line. The company was seeking internet users to work for a small monetary incentive, checking if album titles were correct or covers were suitable for minors. Today other companies can also have tasks completed by the crowd via mTurk. This is how the Germany energy firm EnBW managed to digitalize its customers’ handwritten meter readings. Computers often have problems trying to decipher handwriting.
Since then, thousands of such platforms have popped up worldwide. In Germany there are approximately forty. IG Metall estimates that it employs approximately 1 million crowdworkers. The biggest German platform is called clickworker. It advertises that it has 700,000 on-demand workers, as well as customers such as Honda and T-Mobile.
Go Online, Log In, Get to Work – Without A Contract
Crowdworkers jump in wherever a computer cannot find a solution, or at least a cheap solution. Platforms like mTurk mostly feature so-called “microtasks” – problems that in theory can be solved by anyone with a computer and internet access.
The process is simple: You log yourself onto the site and accept individual jobs. You are compensated per job that you finish. Deeming an image to be suitable for minors, for example, earns you 2 to 5 cents. Most tasks are not done by one person, but by the algorithmically coordinated cooperation of thousands of digital workers, who are unaware of the arrangement.
Together the crowd can work quickly and effectively, simultaneously correcting each other’s mistakes. In this way the roles of humans are computers are reversed. Usually computers solve problems for humans, but here computers require assistance from living workers.
Artificial Artificial Intelligence
The name Mechanical Turk refers to this phenomenon of role reversal. It derives from the chess-playing Mechanical Turk, an alleged automaton that created a sensation in the 18th century. The machine consisted of a mannequin dressed in the style of the Turkish stereotype, various cog-wheels and other constructions. This chess computer could play remarkably well and is alleged to have beaten Frederick the Great and Napoleon.
The secret of this early computer is, however, simple: inside the machine hid a man who was small in size and who could play chess well. The mechanical apparatus was distracting enough to give credence to the illusion.
Amazon’s choice of name is as telling as the platform’s tagline: “Artificial Artificial Intelligence.” The entire setup camouflages the platform and sells human work off as computer work. The working conditions are also quite fitting.
General Terms and Conditions instead of Work Contracts
The success of digital work on crowdsourcing platforms comes not from work contracts but from the general terms and conditions of the platforms that often deny digital workers basic rights. Anyone who uses words like “health insurance” or “union representation” here might as well be speaking a foreign language.
Furthermore, employers – here referred to as “requesters” – can often decide alone whether a job performance is satisfactory and whether a worker should be compensated for it. Even in the case of non-compensation does the requester reserve all rights to the work performed, which virtually raises the issue of wage fraud. Workers are also dependent on positive feedback and reviews in order to be assigned new tasks.
Income Averaging One to Three USD per Hour
The greatest facilitator of precarious work conditions is the fact that the majority of the clickworkers on mTUrk earn an hourly rate between one and three U.S. dollars. Only very experienced users, so-called Power Turkers, earn an income within the range of the minimum wage. So it boils down to profoundly precarious low-wage jobs without social securities. Who would agree to such a thing?
Digital Restructuring of the Classes
Although many men are employed as crowdworkers, the proportion of women among Turkers is clearly higher in the general labor force. Money is the determining factor for the vast majority of crowdworkers. Kirsty Milland, a community-manager and activist at the forum turkernation.com, says, “I’m a Turker – middle-aged, an entrepreneur, a student, a mother, a wife, and I’m reliant on my income from mTurk in order to keep my family from declaring bankruptcy.” Many of the digital workers are by now full-time Turkers – that is to say, completing microtasks has become both their career and their primary source of income.
A relatively high number of digital workers have had problems on the job market due to various physical limitations and forms of discrimination, and end up working at mTurk. One worker named Carey confirms this. “I don’t get invited to many job interviews as a pregnant immigrant with no work experience. That’s why I decided to work at Mechanical Turk. My husband earns a considerable amount, but my income is what literally puts food on the table.”
Piece Salaries and Working from Home: Back to the Future?
The digital economy did not invent the concept of informal, precarious jobs that are poorly paid in piece salaries and performed from home. On the contrary, this sort of work is reminiscent of models that were widely used in the early days of capitalism and that can be found today most commonly in the Global South. A prime example is sewing and textile works, wherein women in particular are exploited. They work from home for piece salaries while simultaneously performing household tasks such as caring for their children or dependent family members.
Working from home nurtures the myth of the “idle” housewife, who actually does the sewing – or the microtasks – only to pass the time and therefore does not deserve to be well-paid. Despite the tedious and poorly paid work, many workers claim that they are happy to be earning money with the help of mTurk. A 29-year-old woman from Missouri named Christina, who has five children, says that through mTurk “I can finally pay off some of the healthcare costs and the rising electric bills. I work eight to ten hours – whenever I have time between taking care of the kids and the housework – sometimes just in order to earn ten dollars in one day.”
Crowdsourcing Turnover Has Doubled
The demand for artificial artificial intelligence has an increasingly international distribution of labor and has created an equally global cybertariat. At mTurk, the majority of workers are located in the United States and India. At the German provider clickworker, they are distributed evenly across Germany, other European countries, the United States and the rest of the world.
And the platforms are thriving. According to its own figures, the crowdsourcing industry was able to double its turnover in the last year.
Digital Organizing and Resistance
Through their terms and conditions, crowdsourcing platforms are structured to individualize their workers because they are physically separated and sometimes spread across continents. These are not good conditions for organizing and resistance. But the crowd is fighting back.
Exploitation on IT-platforms has always led to dissatisfaction, which has been expressed in several ways, including a letter campaign to Amazon’s CEO Jeff Bezos. With the slogan “We are humans, not algorithms,” digital workers demanded more recognition and better work conditions.
Workers continue, often with union support, to fight for legal rights and better pay. An important starting point for creating collective bargaining power are internet forums, where workers can exchange information and organize protests.
Turkopticon: A Tool in the Digital Class Struggle
Another starting point for tactical intervention is the technology. As the platforms are being laid out to continuously rate workers and create competition and pressure to perform, there is almost no way of rating the requesters – that is, the employers.
Turkopticon is a website and browser plug-in developed by Lilly C. Irani and M. Six Silberman as a form of activist technology. There digital workers can rate requesters at mTurk. They can issue warnings about particularly low-paid jobs or companies with a history of poor payment moral.
When a company is being avoided due to its high number of poor ratings, it can end up with costly delays in the completion of its tasks. This is how at least the very first steps toward a digital strike could be taken, which is a considerable success in light of the initial conditions.
There is much to suggest that this kind of work will gain significance. But there are limits to it. Many projects cannot be broken down into tasks suitable for the crowd, requiring more complex forms of coordination and cooperation. Computers meanwhile are continuing to be developed to handle more and more tasks, rendering living workers less necessary.
Ironically, this often happens with the help of the crowd. In order to improve image-recognition software, thousands of sample images must be categorized – a task that sounds like it was made for mTurk. However, this form of digital exploitation will not disappear any time soon. It fits far too well into the logic and the demands of digital capitalism in the 21st century.
Published February 2016
first publication (original article): Luxemburg - Gesellschaftsanalyse und linke Praxis