The Hidden Algorithms Reshaping Gig Workers’ Pay

Food delivery riders and Uber drivers are challenging the algorithms that shape their pay, job access and working conditions. They say automated systems make decisions behind a black box, while workers are left without a clear way to understand or question the results.

In Edinburgh, a group of food delivery riders says pay has fallen as platforms increase their use of automation. They also describe working conditions that have deteriorated, leaving them with less earning power for the same time on the road.

David, who has worked as a food delivery rider in Edinburgh for seven years, says he now earns half what he made four years ago for the same hours. “I am making half the money I was making four years ago, for the same amount of hours. It makes no sense,” he said.

Why riders want to see inside the system

Xabier Villares, an Uber rider who has been riding for eight years and leads the Workers’ Observatory, says the change has been clear. “There has been a dramatic change in the last three years,” he said, pointing to a reduced ability to earn.

The Workers’ Observatory is a charity founded by gig economy workers and academics. Its director, Cailean Gallagher, says the group wants to understand how the platforms’ systems operate instead of accepting unexplained decisions about work and income.

“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,” Gallagher said. He added that “there is so much infrastructure of knowledge and data that’s concealed,” leaving gig workers to work in the dark.

Trade unions, including the Confederación Sindical Internacional, are campaigning to ban dynamic pricing. They say the system leaves workers uncertain about what they will earn, since prices and payments can change through automated decisions.

Uber introduced a dynamic pricing algorithm in 2023, and drivers earned substantially less per hour afterward. Drivers from the UK, the Netherlands and other countries have launched a class action against the company, saying they live in constant fear of the algorithm used to set pay and allocate jobs.

Uber denies changing trip prices based on individual driver behavior. The company attributes differences to other features of its system, including GPS.

A wider problem across platform work

The same concerns reach beyond the UK and Europe. In Latin America, delivery workers and drivers have seen trips and earnings rise or fall according to automated decisions made by algorithms that operate as a black box.

Platforms process information such as customer ratings, accepted orders or trips, cancellations and time in service. Workers do not always know how those details affect decisions about their jobs or income.

Regulations in Mexico, Chile, Colombia and Uruguay recognize that workers should receive information about the algorithmic criteria that influence their work. Mexico’s Federal Labor Law says platforms must explain, in simple and clear language, the factors that affect task assignments and other parts of the job.

That requirement raises an important difference between transparency and explainability. Transparency shows which elements enter a system. Explainability goes further by helping people understand the reasoning behind an automated decision.

Laura Mantilla-León, an analyst at Rights Digitales, says the information provided to workers often remains too broad. “Lo que está en el papel se restringe un poco a descripciones generales de criterios y de variables y no avanza a algo que es clave, que es la explicabilidad,” she said.

In English, her point is direct: listing general criteria and variables does not show a worker why a particular decision happened or how the system weighed the available information.

Ratings, bias and the need for human review

A 2024 study from the Universidad Nacional de General Sarmiento found that ratings and acceptance rates can influence the number of orders workers receive and their access to better-paid jobs. The study also found that variables with a smaller role can affect work opportunities and income without the worker knowing how.

Research from Yale University and the National University of Singapore found that racial bias in ratings can affect decisions on home delivery platforms. Biased evaluations increased the share of ratings below five stars among workers perceived as minorities by 33.9%, producing a 6.5% income gap.

That finding puts pressure on platforms to do more than disclose the existence of automated systems. Workers need a way to challenge a decision when the data behind it is inaccurate, unfair or affected by discrimination.

Mantilla-León says explaining an automated decision is not enough. A person must also be able to request human review, especially when the decision affects access to work or income.

“La importancia de la revisión y la supervisión humana es justamente que toda decisión que tenga la potencialidad de afectarme como persona trabajadora pueda ser pasada por unos criterios de razonabilidad y de proporcionalidad de esa decisión y que no se adopten de forma estandarizada, opaca y discriminatoria,” she said.

Mexico, Colombia and Uruguay recognize different forms of human review for automated decisions. But major questions remain: How long can a review take? How independent should the reviewer be? Who checks that the process works as promised?

Those questions matter because a blocked account can remove a worker’s income at once. A review that takes days or weeks can mean days or weeks without access to work, making transparency and human oversight part of the basic conditions of platform employment.

Stay connected via Google News

Follow us for the latest AI News