Ride-sharing apps reduce impact of racial discrimination, new study suggests

Racial discrimination against black passengers seeking to hail rides has been a problem since the taxi era. A new study by researchers at Carnegie Mellon University’s College of Engineering aimed to determine whether the rise of ride-hailing apps like Uber and Lyft has changed this dynamic, for better or worse.

The work appears in the Proceedings of the National Academy of Sciences.

A previous study in which researchers requested rides at specific times and locations, changing only the potential passenger’s name, showed that using a black-sounding name resulted in up to twice the rate of cancellation versus using a white-sounding name. Yet despite this substantial difference, wait times were the same or reflected a difference of just a few seconds, and the research team wanted to find out more.

They ran simulations of all trips made in Chicago, before and after the COVID-19 pandemic, over several days. Research estimates that at least 3% of drivers must discriminate based on race in order to produce the cancellation disparities observed by previous studies. But the study also showed that the ability of these services to quickly rematch riders with new drivers nearly eliminates the effects of driver racial discrimination on disparities in rider wait times.

“Technology alleviates a social problem, which is quite rare,” said Jeremy Michalek, professor of engineering and public policy (EPP) and mechanical engineering and leader of the study. “Discrimination has little effect on average wait times, at least in part because these apps are able to respond quickly when someone cancels, whereas with taxis it was a very difficult problem to solve.”

“In the absence of these apps, some populations facing extremely long wait times could be lost because it is a hidden injustice where people just pass by on the street,” Destenie said Nock, professor of EPP and civil and environmental engineering. “You can now get reconnected quickly, allowing people to arrive at work on time, make hospital appointments and actively participate in the transportation system.”

Individual racism is only part of the equation, and the broader systemic problem of residential segregation led the team to focus on Chicago, one of the most residentially segregated cities in the United States, which also makes a lot of travel data available. -travel while calling.

Even when drivers treat everyone the same, Black passengers in Chicago experience significantly longer wait times because of where people live, the study found. Residential patterns in Chicago are influenced by a long history of discriminatory practices, including redlining, and other factors such as house inheritances and wealth. Today, black residents are concentrated on Chicago’s South Side, which is farther from busy downtown areas, meaning there are fewer drivers in the area to pick up passengers.

“One thing that makes this research unique is that it distinguishes between two types of discrimination,” said Anna Cobb, the study’s first author and a Ph.D. student in EPP.

Types of discrimination are “direct, such as when a driver cancels a traveler because of their race, and systemic, where history has informed people’s residential patterns, so that even when the effects of direct discrimination are weak or disappear completely, disparities may persist. “, explained Cobb. “Being able to distinguish these effects can help us determine how we address the disparities we see in the real world.”

“It’s encouraging to see how much this technology has mitigated the effects of driver discrimination on drivers,” Michalek said. “But the bigger picture is more complicated. In a society where disparities are deeply entrenched, even service without direct racial bias can nonetheless produce significant gaps in service quality that can reflect, or even exacerbate, existing disparities .”

Heinz College assistant professor of civil and environmental engineering Corey Harper and PPE alumnus Aniruddh Mohan also contributed to the research.