Algorithms are used to dictate speed, behaviour and, ultimately, the wages of gig workers, resulting in different payments for the same work conducted at the same time, with the same skills

Jude Okoye was tired and annoyed. On a blazing hot Thursday in June, he had just spent three hours driving around an area of midtown Toronto within a two-kilometre radius, delivering food and groceries using the DoorDash app.

But the latest delivery request he accepted had directed him six kilometres out of the midtown delivery zone, forcing him to contend with traffic and, more critically, lose precious delivery minutes. Because of traffic and distance, the job would take him 45 minutes in total, for an estimated payout of \$10.50. Mr. Okoye’s three previous food deliveries – each less than a two-kilometre drive – had taken exactly 52 minutes to complete, for which he received \$47.43 in total, including tips.

“You see I don’t know why it [DoorDash] did that. I put midtown Toronto as my delivery area. Now I’m wasting my time … one hour is gone!” he said, switching lanes every few minutes and braking abruptly in an attempt to get to his destination as quickly as possible.

Apps like Uber and DoorDash use AI to determine pay. Workers say this makes it impossible to predict wages

He also complains about the cost of gasoline. This particular delivery would drink up more gas than he was comfortable with. He constantly performs mental calculations about time, dollars per km and the cost of gas, trying to optimize his own pay in a four-hour shift.

But it is virtually impossible.

Every time Mr. Okoye thinks he has figured out an area or route that will give him some semblance of consistency in his wages, he says the food delivery apps (he uses SkipTheDishes and DoorDash) will throw him for a loop. They will either give him a job that takes him out too far for a wage that does not correspond with the time spent and distance travelled, or he will just stop getting delivery orders altogether, forcing him to sit idle and unpaid for up to 20 minutes while he waits for the app to find him an order.

What Mr. Okoye experiences in his job as an independent gig worker for online food delivery companies is the result of artificial intelligence technology, designed to determine what a worker should get paid for their labour based on a set of algorithms unbeknownst to the worker.

App-based ride-share and food delivery platforms such as Uber, Lyft and DoorDash depend heavily on algorithms to govern how quickly a customer can find a ride or get a food order, with the ultimate goals of boosting efficiency for customers and accumulating greater revenue for the company.

But algorithms are also used to dictate speed, behaviour and, ultimately, the wages of gig workers, resulting in different payments for the same work conducted at the same time, with the same skills. It is a concept that has become widely known as algorithmic wage discrimination – a system of pricing workers for their labour, but in a manner that is variable, opaque and unpredictable.

The ascent of digitalized pay practices is profoundly unsettling, labour experts warn, because it undermines the notion that workers should be paid in a way that is predictable and sustainable when they create value for a company.

“Algorithmic management of wages ultimately keeps the worker working for the company as long as possible, for as little as possible,” said Veena Dubal, a law professor at the University of California, known for her research on the gig economy and who coined the term “algorithmic wage discrimination.”

“It sets the worker up to believe if he works harder or faster, he will be rewarded appropriately. That might be true at first, so he keeps working. Eventually, unpredictability sets in, but the worker is now committed to the platform and convinced he can win or master the system. That rarely happens,” said Prof. Dubal.

Mr. Okoye migrated to Canada from Nigeria as a young adult, 25 years ago. He spent much of the past few decades working in the warehousing industry – driving forklifts and managing logistics for a mid-sized factory in Toronto. With few breaks and long shifts, the jobs he performed were exhausting, prompting him to seek out more flexible work.

In 2018, Mr. Okoye started driving for Uber full time. The money was decent, he said, but within eight months he got deactivated from the app for reasons that are still unclear to him. “I think maybe a passenger complained about me,” he said. “I don’t know why. Uber did not tell me the reason.”

He moved on to food delivery with DoorDash and SkipTheDishes. Given his account history with Uber, he was not able to deliver for Uber Eats, the most popular food delivery app in Toronto. Regardless, he was particularly happy with DoorDash. At least in the beginning, the work was consistent, and he was routinely bringing in up to \$250 a day for eight hours of being logged on to the app.

But Mr. Okoye says that over time, his wages became more unpredictable. Theoretically, he could make up to \$250 a day, but it had become harder to do so because of the inconsistency of what he got paid for each delivery, despite the distance. Wait times in between deliveries – that is, the time it took for the app to generate a new job for Mr. Okoye – were also strangely unpredictable.

In a statement to The Globe and Mail, DoorDash said its pay model is designed to make earnings “fair and transparent for every delivery” and provides couriers with a “clear understanding of potential earnings on each order.”

Total earnings for an order are determined by a combination of base pay, promotions and tips, according to the company. Base pay ranges from \$2 to \$10 per delivery, and is dependent on “time, distance and desirability of the order,” while promotions and tips vary by delivery.

In order to fully understand how the app compensated Mr. Okoye for a day’s worth of work, The Globe shadowed him for two four-hour shifts.

That Thursday, Mr. Okoye left his home in the Jane and Weston neighbourhood of Toronto’s west end at around 10:15 a.m. He arrived at his delivery start point in midtown Toronto (the intersection of Dupont and Spadina) a couple of minutes before 11 a.m. His plan was to deliver food between 11 a.m. and 3 p.m., the lunch hour, take a break until 5 p.m., and then resume working the dinner shift from 5 p.m. to 9 p.m.

Between 11 a.m. and 11:21 a.m., however, Mr. Okoye waited for the app to generate a delivery. “I am not being paid right now, that’s 20 minutes gone,” he quipped, subsequently pressing a prompt on the app that said “I’m Not Getting Deliveries.”

“This will work. They will now start giving me orders,” he said. Seconds later, his first order came through – a McDonald’s delivery for which the app estimated he would get \$9.50. Mr. Okoye worked with sheer speed upon accepting the order – he knew exactly where to go, and exactly where to park in an area of Toronto heavy with traffic, pedestrians and cyclists.

He took 14 minutes to complete the pick-up and delivery, which was just slightly more than two km of driving, and received a base pay of \$4.50 from DoorDash, a “top-up” of \$1.00 (for delivering in a busy area) and a tip of \$5.00.

More orders came through in the same vicinity: a large sushi delivery to an office building (\$4.50 in base pay, \$1.00 top-up, \$15.00 in tips) and a single salad order to a condominium unit (\$5.00 in base pay, \$11.43 in tips). By 12:15 p.m., he had made almost \$50.

Mr. Okoye’s dependence on tips is striking. By this point, more than 60 per cent of his income had come from the generosity of his customers. By the end of the day, approximately half his income would come from tips.

He said this never used to be the case with DoorDash, and the app used to have a higher base pay. But when the COVID-19 pandemic set in and customers started tipping more generously, he said the app’s base pay started declining. It did not affect his total payout for a job, however – because people tipped so well, he often made more than he ever had before the pandemic.

DoorDash did not respond specifically to Mr. Okoye’s suggestion that base pay had declined recently. The company instead reiterated that there are “multiple factors” that determine what couriers earn on deliveries.

A study conducted in late 2022 by New York’s Department of Consumer and Worker Protection found that food delivery workers at Uber Eats, Grubhub, DoorDash and Relay earn an average of US\$14.18 an hour, split evenly between pay and tips. The study also found that, on average, base pay per delivery declined by 23 per cent between the first quarter of 2021 and the second quarter of 2022.

“We now have data demonstrating that the app companies pay the delivery workers about half of what they receive, and keep the other half for themselves. They can do that because they know that consumers using these delivery services are very generous with tips,” explained James Parrott, the director of economic and fiscal policy studies at the New School, a private research university in New York.

Indeed, using weekly aggregate data obtained from four delivery apps, the DCWP report breaks down this equation. For a food order of US\$33.09, including a tip of US\$4.11, the app will receive a total of US\$8.54 (fees charged to the consumer plus the app’s share of order subtotal). Of that amount, US\$4.32 will be paid out to the worker, while the company’s gross margin for that order will be US\$4.22. In effect, a customer’s tip almost completely subsidizes the worker’s wage.

Prof. Dubal says that it is typical of gig workers, particularly food delivery couriers, to receive roughly 50 per cent of their income in the form of tips. “The apps have come to rely on the benevolence of customers, but that adds yet another level of unpredictability and stress to the worker, because the tip percentage is completely up in the air and varies from job to job,” she said.

Mr. Okoye takes pride in knowing exactly which neighbourhoods to work in because he’s more likely to get higher tips there. “This is why I work in the midtown area. There are expensive restaurants. And rich people, they like to give me big tips,” he said. But he also admits that it makes him anxious relying on customers for earnings, because he feels like he has to move with extra speed and precision in order to elicit the best tips.

At close to 2 p.m., the app stops giving Mr. Okoye orders. It could be because the lunch rush is over, he theorizes, but is not convinced. He waits about 16 minutes – unpaid time – before the app sends him a delivery that would take him about 12 km away. He immediately declines it. Then he gets a double order – two pick-ups at restaurants doors away from each other, and two deliveries. The app only tells him the location of the first delivery, which is about two km away from the restaurant.

After completing the first drop-off, the app informs Mr. Okoye that his second delivery is more than six km away, and would take 40 minutes, with traffic. The base pay for the first order, a quicker delivery because of distance, is identical to the base pay for the second order, which takes more than double the time.

But refusing the second order would greatly affect his DoorDash rating, and he is not willing to risk that. So, begrudgingly, he completes the delivery and then logs out of the app – irritated and exhausted.

“I wouldn’t have accepted that double order if I knew where the drop-off was,” he said. He believes the app deliberately does not tell drivers where their exact food drop-off location is, for fear that they will reject the job.

For roughly four hours of work that afternoon, including 36 minutes of idle time waiting for orders, Mr. Okoye took home \$99.18 – and half of that came from tips.

At night, he took home a bit more because dinnertime tends to be busier: \$136.47. He spent roughly \$15 on lunch, and \$35 on gas that day. By the time he gets home it is past 10 p.m. It was a 12-hour day, for net wages of approximately \$185.

DoorDash did not respond specifically to questions around why Mr. Okoye got paid the same base pay for orders with different distances and travel times, and why the app does not allow delivery couriers to see their exact drop-off address before accepting an order.

But the statement referred to a post on the DoorDash website that states that couriers who accept and complete the most orders ultimately earn the most money. In the United States, DoorDash recently introduced a new way for couriers to earn money called “Earn By Time,” which essentially guarantees an hourly minimum rate excluding tips. Or couriers can choose to stick with the existing compensation model.

The company also noted that in 2022, more than 250,000 Canadians earned income on the DoorDash platform. On average, they earned \$27 an hour while on delivery, including tips.

In her paper, On Algorithmic Wage Discrimination, Prof. Dubal suggests that information asymmetry between workers and the company – for example, deliberately not disclosing to workers an exact drop-off address when they first accept a job – allows the apps to calculate the exact wage rates necessary to incentivize desired behaviour.

Taking Mr. Okoye’s experience of being routed out of his delivery zone, Prof Dubal explains that app first “primed him” to keep accepting orders by giving him “good” orders – short distances, for a payout rate of more than \$9. “It set him up to feel like the last few hours were good, so this order should be good too. But it tricked him. He would not have accepted the order if he had the same information as the company did,” she said.

In its statement to The Globe, DoorDash called Prof. Dubal’s claims “speculative and inflammatory,” and noted that its platform is “fair and transparent for Dashers.”

Uber has publicly stated that it uses behavioural and social scientists to determine how to incentivize drivers to work harder and for longer hours, but in a way that drivers feel is rewarding to them. It uses tactics such as “surge” and “boost” pricing to influence where a driver or courier works from, and how long they stay in a certain area.

But these “wage manipulators,” as Prof. Dubal terms them, are rarely consistent, again leaving workers guessing as to how much they will make per hour.

Brice Sopher, a part-time Uber Eats delivery courier, who is also vice-president of Gig Workers United (a labour collective fighting for gig workers in Canada to be designated as employees under provincial employment acts), told The Globe that his wage per delivery has gotten so unpredictable he has stopped relying on Uber Eats for any semblance of steady income.

On Monday, May 29, Mr. Sopher travelled 3.2 km in 16 minutes on his bike to pick up and deliver a single order in downtown Toronto. He received a total of \$6.09 for it (the customer did not tip). On Wednesday, May 31, he took 10 minutes to bike 0.9 km for a single pick-up and delivery also in downtown Toronto, and received \$7.03 in total (including a \$2 tip). The “boost” promotion for the two orders varied as well: \$4.44 for the first delivery and \$3.54 for the second delivery.

“I cannot tell if my wages depend on distance, or time, or price of the order, or time of day,” he said.

In a statement, Uber Canada spokesperson Keerthana Rang said it is always hard to comment on experiences of a specific delivery person. “There are many factors that play a role in earnings like time of day, demand in the area, weather, number of delivery people on the platform at that given time, and whether they have opted in to receive Shop and Pay trip requests, etc.”

Ms. Rang also noted that Uber Eats presents earnings information to couriers up front so they have the “freedom and flexibility to choose what orders are right for them”.

Food delivery platforms, in particular, have seen their revenue surge over the past two years. In 2022, the Uber Eats portion of Uber Inc.’s business generated almost US\$11-billion in revenue, compared with US\$8-billion the previous year. From March, 2022, to March, 2023, DoorDash Inc.’s quarterly revenue soared by 40 per cent to more than US\$2-billion.

To a large extent, economists, business leaders and human-resources professionals agree that work – using a skill set to perform a task that creates value for the employer – is only sustainable if it also rewards the employee with job satisfaction and upward mobility. Algorithmic wage discrimination undermines this notion, argue Prof. Dubal and Prof. Parrott.

AI systems are designed to keep wages level no matter how efficiently a worker performs and essentially erode the idea of a minimum pay standard, says Prof. Parrott.

Prof. Dubal’s research has found that food delivery workers who work longer periods of time, get paid less per hour. “Literally by design, achieving upward mobility is impossible,” she notes. “You have different people getting paid different amounts for the same job, and receiving different amounts of work in the same time period. The idea that the development of a skill over time lends itself to economic mobility simply does not exist for this group of workers.”

A key policy solution to ensure a degree of certainty in gig worker pay that many jurisdictions are fighting for, and some have achieved, is establishing a minimum wage for workers deemed independent contractors.

In June, New York instituted a minimum pay standard of US\$17.96 an hour for app-based delivery workers. The amount took into account what workers had been getting paid, on average, including time they spent idle between orders. The hourly minimum will rise to US\$20 by April, 2025.

In Ontario, legislation passed in April, 2022, by the province called gig workers to be paid an hourly minimum wage, but excluded the time they spend idle, waiting for deliveries or rides. The exact implementation of the legislation is still being debated by labour groups, app companies and the government, so it is unclear when and how changes will come into effect.

Prof. Parrott argues that the New York pay standard does go a long way toward addressing compensation issues for gig workers, but there is always the potential for companies to reprogram their algorithms in ways that create better outcomes for them.

Can governments force tech companies to be transparent about how exactly they use AI to determine wages? Possibly, says Prof. Dubal, but because algorithms are not static, transparency will not yield much information to the worker or customer.

The better solution is to regulate the outcomes of the algorithmic wages. “Governments can draft regulations to make sure companies are ensuring efficiency for the worker and the customer – that the closest worker to the job receives the job, and that the app isn’t going to trick the worker into making decisions that are against their own interest,” she said.

Labour groups, such Gig Workers United, have been pushing for gig workers to gain employee status because it would entitle them to benefits such as employment insurance, minimum days of sick leave and a minimum wage. The app companies argue that designating gig workers as employees removes the flexibility the workers have to work on multiple platforms and at whatever time of the day they choose.

Mr. Okoye is torn between the two ways of working. He truly enjoys the flexibility of his job, and not dealing with a manager or colleagues. But he also feels he is being exploited, and wants a higher base wage, less reliance on tips, and health and dental benefits. He is convinced that if the apps keep “manipulating our pay” the way they are, they’ll lose workers.

“I used to save some money from this job. Now I save nothing. That’s why no one will do this forever. You just cannot count on it.”