Who Controls the Algorithm Setting Gig Workers’ Pay?

Food delivery riders and Uber drivers are asking platforms to explain the algorithms that decide which jobs they receive and how much they earn. They describe these systems as a “black box” because workers can see the outcome but not the rules behind each decision.

In Edinburgh, a group of food delivery riders says pay has fallen while working conditions have worsened as platforms increase automation. David, a food delivery rider in the city, said, “I am making half the money I was making four years ago, for the same amount of hours. It makes no sense.”

Xabier Villares, a Deliveroo rider and lead organiser of the Workers’ Observatory, said, “There has been a dramatic change in the last three years.” The Workers’ Observatory is a charity founded by gig economy workers and academics to research the concealed parts of workers’ experiences.

The hidden rules behind pay and job offers

The observatory has carried out experiments to monitor the pay levels offered by algorithms, a practice known as “dynamic pricing.” Trades unions are campaigning to ban dynamic pricing because they say it leaves workers uncertain about their earnings.

Research from the University of Oxford and Columbia Business School found that the introduction of a dynamic pricing algorithm in 2023 resulted in Uber drivers earning “substantially less” an hour. Drivers from the UK, the Netherlands, and other countries have launched a class action against Uber, claiming they live in “constant fear” of the algorithm used to set pay and allocate jobs.

The lawsuit claims Uber’s pay-setting system breached privacy laws and pushed down earnings. Mohammed Shirwa, an Uber driver from Rotterdam, says Uber’s pay-setting mechanism preys on weakness. Uber denies adjusting trip prices based on individual driver behavior and attributes differences to other system features, such as GPS.

Cailean Gallagher, director of the Workers’ Observatory and a lecturer at St Andrews university business school, said the group wants to understand the systems that shape workers’ lives. “What we are trying to do with the observatory is to get on the ground floor in terms of understanding how the whole apparatus works,” he said.

Gallagher added that “there is so much infrastructure of knowledge and data that’s concealed” by platforms. That data can include customer ratings, accepted or canceled orders, and completion times, all of which platforms process when making decisions about work.

Transparency is not the same as an explanation

The same questions are being raised across Latin America. Apps measure location, times, and ratings, but workers often do not know how those details influence their work and income. Derechos Digitales, an organization defending digital rights in Latin America, has identified opacity in the way data shapes decisions about workers.

Latin American regulations recognize that workers should receive information about the algorithmic criteria affecting their tasks and earnings. Mexico’s Federal Labor Law, known as La Ley Federal del Trabajo, states that factors such as ratings, incentives, penalties, and criteria influencing task assignment must be explained in simple language.

Knowing which elements are involved is transparency. Understanding why a specific decision happened is explainability. Workers need both, because a list of general rules does not explain why one person received fewer orders, lost access to better-paying jobs, or faced a penalty.

A 2024 study by the Universidad Nacional de General Sarmiento analyzed 70 interviews and 750 cases involving delivery workers in Buenos Aires. It found that metrics such as ratings and acceptance rates influence the number of orders workers receive and their access to better-paying jobs.

Without clear information, workers may have to “decrypt the rules of the game” to improve their performance or income. The study also found that safety concerns and caregiving responsibilities can influence acceptance rates and earnings, which means a single metric may not show the circumstances behind a worker’s decisions.

Why human review matters

Research from Yale University and the National University of Singapore found racial bias in evaluations that affect income and job opportunities. The study analyzed 86,157 services and found that discrimination increased 33.9% in lower ratings for minority workers, producing a 6.5% income gap.

Laura Mantilla-León, an analyst of Public Policies at Derechos Digitales, says explaining automated decisions is not enough. Workers also need review mechanisms that can challenge decisions and correct errors.

Mexico, Colombia, and Uruguay recognize different forms of human review for automated decisions. The purpose is to ensure that decisions affecting workers are reasonable and proportional, rather than standardized, opaque, or discriminatory.

That protection raises practical questions. How long should a review take? Who handles it, and how independent are they? Is there outside oversight? Delays can leave workers without access to jobs for days or weeks, cutting off their income while a decision is examined.

Gallagher described the value of human supervision this way: “the importance of the review and the supervision human is exactly that every decision that has the potential to affect me as a worker can be passed through some criteria of reasonableness and proportionality of that decision and that decisions are not adopted in a standardized, opaque and discriminatory way”.

Even if platforms explain the variables they use, one question remains: who verifies that the system works as the platform claims? For delivery riders and drivers, that answer affects more than an abstract debate about artificial intelligence. It can determine which jobs they see, how much they earn, and whether they can keep working.

Stay connected via Google News

Follow us for the latest AI News