Millions of Workers Now Report to an AI Algorithm — Not a Boss

Complaining about your boss is an office tradition as old as work itself

But what if you don’t work for a human boss? What if your boss is an algorithm whose internal workings are totally opaque–and yet that boss is able to control everything about your life, and even determine if you’ll be able to put food on your table?

According to a new report from the Mozilla Foundation, a non-profit research organization based in Mountain View, that precarious position is increasingly common for gig workers around the world. Many toils in potentially dangerous conditions, with their every move dictated by black-box algorithms that even the algorithms’ creators don’t understand. 

These algorithms determine their pay, have the power to furlough them or lay them off, and can dramatically change their working conditions on a data-driven whim. Without the ability to work with a human supervisor–or even to connect with other human colleagues–workers often lose the ability to organize, air their grievances, or ask questions about how to do their jobs safely.

Inside the Algorithms

To learn about this shadowy world, the Foundation’s researchers connected with Eduardo Meneses. Meneses is the global head of a technology consultancy that’s working to understand the impact of AI algorithms on workers. According to Meneses, in countries like Ecuador, delivery drivers and other gig workers are often completely beholden to the decisions of apps and their underlying algorithms.

This can lead to crushing pressure for workers. Meneses told Mozilla’s researchers that on a specific app if delivery drivers “take more than 50 minutes” to complete a delivery, they’re “suspended for two days. So that means for two days, you cannot work.” Those kinds of penalties can be crushing for low-wage gig workers, Meneses says. Losing two days of work could mean the difference between a worker putting food on their family’s table or going hungry.

Unlike a human boss–who might show lenience and understanding if a worker explained that a delivery was delayed by traffic, an injury, or another unavoidable cause–algorithms are brutal in their adherence to strict rules. The Mozilla Foundation’s report shares that with certain apps, if a worker is even one minute late for a delivery, they might lose out on days of work as a penalty.

Delivery drivers are the most visible type of gig worker, but the Foundation’s report says that millions of workers worldwide now report to algorithms and apps instead of human bosses. Gig workers clean houses, drive cars, run errands, and perform all manner of other work. Gig work–and the associated issues with reporting to an algorithmic boss–are not a problem unique to small countries, either.

Over 59 million Americans participate in the gig economy in some capacity.

The Perils of ‘Being Your Own Boss’

To gloss over the challenges facing gig workers, the report says, most gig economy companies turn to corporate narratives about flexibility and empowerment. Companies entice workers with the promise that one can “Be your own boss” or “Work whenever you want.” 

Those promises are true in theory; gig workers generally don’t have formal supervisors or a position within a rigid corporate hierarchy. Barriers to participation in the gig economy are often low. In 2019, I experimented with driving for the Silicon Valley gig economy company Doordash. I signed up in the morning and was delivering food by the afternoon.

In practice, though, Mozilla says that gig economy workers simply end up substituting an algorithmic boss for the traditional, human kind. They might technically work for themselves, but they’re only able to actually get work if their apps’ algorithms choose to assign it to them. Despite apps’ entrepreneurial rhetoric, the Foundation says, that puts them in a more precocious position than traditional workers–especially since many lack the legal protections of formal W2 employment.

Faced with these challenges, many workers are beginning to fight back. According to Mozilla, gig economy workers have begun to organize, staging walkouts and protests for better pay and working conditions. In some cases, workers choose to protest not against the apps that employ them but against the gig economy algorithms themselves.

Fighting Back

Workers are taking other actions, too. Mozilla says that organizations like Driver’s Cooperative in New York City are building delivery platforms owned and led by gig workers themselves. The organization now has 6,500 members. CoopCycle, based in Europe, is another example. 

Some workers are also trying to understand better how their algorithmic bosses work so that they can engineer or demand changes. That can be an especially challenging problem, Mozilla says. Many algorithms are designed by companies here in the Bay Area, only to be deployed in myriad apps all over the world. Connecting the dots and determining who controls an algorithm can be challenging for professional researchers like Meneses. Gig workers often have almost no tools at their disposal to do this kind of sleuthing.

The design of the algorithms can frustrate attempts to understand their actions, or even to determine basic things like how much the algorithm is paying a given worker per hour for their labors. As Mozilla shares, many algorithms are “black boxes,” which means that even their own creators don’t understand how they work. 

Black box algorithms might appear to be dolling out fair wages in the aggregate, but their impacts on specific workers can be much more harmful, Mozilla says. In 2020, for example, the Foundation says that Shipt, a popular delivery app, switched to a black box algorithm. Although Shipt said the algorithm was fair, researchers show that it resulted in “a significant pay cut for 41% of workers.”

Again, workers are taking matters into their own hands with black box algorithms, too. According to Mozilla’s report, gig workers have begun to deploy tools that help them track their earnings and share data with others workers. Dan Calacci of MIT’s Media Lab, for example, helped to create Shipt Calculator, a data-sharing tool that workers used to uncover the alleged unfairness of the app’s black box algorithm.

Some workers and advocacy groups are aiming higher. Rather than reverse engineering specific black box algorithms, they’re working to outlaw such algorithms altogether. The European Commission, for example, has explored legislation that would require gig economy apps to allow auditing of their algorithms.

In many cases, though, advocates say these kinds of audits don’t do enough to protect workers. Instead, they’re turning to existing legal frameworks and seeking to classify gig workers as traditional employees. New EU laws would classify gig economy workers as “platform workers,” for example, providing “access to social security, perhaps to insurance,” advocate Aída Ponce Del Castillo told Mozilla. California’s AB5 took a similar approach in 2020, classifying gig workers as employees, although the process has been fraught.

You’re Next

This kind of redrawing of traditional models for employment, Mozilla says, is essential not just for the future of the gig economy but for the future of all workers. As the Foundation points out, gig economy workers are a “canary in the coal mine,” because their work is relatively easy to direct via algorithms.

As AI improves, though, more and more jobs will be supervised or controlled by algorithms. Advanced AI systems like GPT-3 are already capable of remarkable feats and can already do things like writing legal briefs or diagnosing medical conditions. The time when such algorithms start supervising human professionals may be sooner than we think. Imagine a future where lawyers are compensated based on whether an AI feels their arguments would stand up in court or where doctors’ earnings (or medical licenses) depend on an AI’s evaluation of their diagnoses. 

As AI bosses take on increasingly complex challenges, drawing the right boundaries between algorithms’ efficiencies and humans’ nuanced understanding of the real world–as well as their compassion for other human beings–will become more and more important. Collective action, tools that empower workers, and legislation that ensures algorithmic fairness and transparency are essential to make sure that algorithms end up serving us–and not the other way around.

// Thomas Smith is a writer and content expert based in San Francisco, California.

Feature image: Courtesy of Thomas Smith/Gado Images

Leave a Reply