Algorithmic Management, Monitoring, and Control: Worker Classification in the Digital Age

Nowadays, it’s hard to read anything about workplace policy without running into “algorithmic management.” Companies, we’re told, are increasingly controlling workers through an array of digital “tricks.” These companies record our keystrokes, track our locations, and even watch us through our webcams. We hear this same story in academic journals, government reports, and the popular press. In fact, the story has even made its way into federal regulations—specifically, in the U.S. Department of Labor’s current rule about independent contractors. Like the more popular accounts, this rule assumes that algorithmic management is pervasive. And it treats the practice as a form of “control.”

There’s only one problem: algorithmic management isn’t a real thing. It is a term used to describe business practices so old and so normal that they would be boring if they weren’t branded with a scary label. These practices include monitoring performance, providing incentives, and tracking the work. Critics of these practices have successfully marketed them as something new and bad for American workers. And for a few years now, that misleading narrative has been shaping policy at the Department of Labor.

Fortunately, the narrative may finally have run its course. The Department of Labor recently proposed to eliminate the idea from its regulations and return to a more traditional definition of workplace “control.” This definition would recognize, at least implicitly, that incentives and monitoring are not in themselves forms of control. Properly understood, they are ways that businesses manage their suppliers when they have no control. That is, they are not evidence of control, but its opposite. And that is true even when they’re packaged under a scary label like “algorithmic management.”

Algorithmic Management: A Null Set

The term algorithmic management has always been a bit vague. It was coined in a 2015 academic paper focused on then-novel rideshare platforms. The paper studied how these platforms used digital tools to coordinate a vast fleet of independent drivers. The paper focused on three of these tools: matching riders to drivers, customer-ratings systems, and surge pricing. The paper found that these tools not only helped platforms balance driver supply and rider demand, but also kept the drivers mostly happy. Algorithmic management, it seemed, wasn’t such a bad thing.

But since then, the term has taken on a life of its own. It is now applied to tools ranging from scheduling software to chatbots to AI, the latter of which has driven the alarmism to new heights. Writers like Noam Scheiber of the New York Times, Sam Levine of the Federal Trade Commission, and Veena Dubal of UC Irvine have described algorithmic management in pernicious terms, arguing that companies are using it to fool people into working longer hours for less money. Dubal has even compared it to a modern Jim Crow.

These arguments aren’t just rhetoric. They’ve started to shape public policy. In 2024, the Department of Labor adopted a regulation spelling out the difference between employees and independent contractors. That difference is important because under the Fair Labor Standards Act (FLSA), the nation’s main wage-and-hour law, only employees are entitled to minimum wages and overtime. The FLSA distinguishes between these classes of workers based on multiple factors, including who “controls” the work. The 2024 regulation defined this type of control to include some kinds of algorithmic management. In particular, it pointed to “technological” monitoring, such as GPS tracking. That kind of monitoring, the regulation said, counted as control even if it wasn’t paired with any other steps. Technological monitoring was itself a way to control another person’s work.

That approach was a departure from the department’s traditional rules. Traditionally, control meant some affirmative action: the business had to instruct, prevent, or punish the worker for doing something. But the 2024 rule expanded the concept to include mere observation. If a business merely collected information about work, it could be deemed to control the work—as long as it did so through “technology.”

Analog Logic for a Digital World

This kind of thinking gets the issue backward. Incentives, monitoring, and similar techniques aren’t signs of control: they’re signs of its absence. In economic terms, they’re just ways that businesses solve the “principle-agent problem.” The basic issue is this: when a business hires a contractor, the two have different incentives. While the business wants the best service for the lowest price, the contractor wants the highest price for the least effort. So to protect its interests, the business has to adopt safeguards against slack work. If it were using its own employees, it could manage that risk simply by telling the employees what to do. But it can’t do that when it uses contractors, often because it doesn’t know enough details about the work itself. (That’s why it hired a contractor in the first place.) So instead, it may require the contractor to report back when it hits certain milestones (monitoring). Or it might give the contractor an extra payment for high-quality deliverables (incentives).

This kind of indirect influence has never been considered “control” for the purpose of classifying workers—nor should it. If it were, it would be hard to classify any work as independent. Because all principals monitor their agents to some degree, monitoring itself tells us nothing about how to classify the relationship. For example, when a business ships a package and pays for delivery confirmation, it is in some sense monitoring the work. But no one thinks the law should treat every logistics supplier as the employee of all its customers.

None of that changes with modern technology. Even when businesses and contractors use technology, they still face principle-agent problems. And the best tools for addressing those problems are still monitoring, incentives, and similar indirect methods. While these methods may work faster in the digital world, the dynamics are the same—even they’re wrapped in a scary label like “algorithmic management.”

Of course, this greater efficiency of algorithmic tools is exactly what makes people uneasy about them. People worry that if an employer can monitor their keystrokes or access their webcams, the employer might use the information to micromanage their work. But the important point isn’t whether the employer collects information: it is how the employer uses the information. If the employer uses keystroke data to evaluate, compensate, or punish a worker, it is the use rather than the data that counts as control. Legally, the question is whether the principal controls the work—not whether the principal knows how the work is being done. That’s true whether the data is collected through technology, a delivery receipt, or observation with one’s own eyes.

Fortunately, the Department of Labor is coming around to that view. In late February 2026, it published a new rule to distinguish between contractors and employees. This new rule says nothing about algorithmic management or any other kind of “technological” control. Instead, it returns to first principles: it says that if a person controls her own work, she’s probably a contractor. If she doesn’t, she’s probably an employee. The point is the same even when work moves into the digital world.