A two-year study from the National Bureau of Economic Research recently found that Uber and Lyft riders with “Black-sounding” names waited longer to have trip requests approved than riders with “White-sounding” names. Now, Lyft has announced a new measure it’s taking to curb racism: a hidden score measuring driver discrimination.
In November, Senator Al Franken wrote an open letter to Lyft’s CEO, Logan Green, and Uber’s CEO, Travis Kalanick, asking them to address racism on their platforms. On Wednesday, Sen. Franken released both letters. In its response, Lyft said it will start tracking driver cancellation and quality of service for drivers in poor, minority communities. From the letter:
Also, moving forward we will enhance our regular thorough review of ride cancellations (as noted above), by including a focus on cancellation rates and quality of service in “minority census tracts” as defined in 12 USCS § 4502 (a census tract that has a minority population of at least 30 percent and a median income of less than 100 percent of the area median income).
Lyft already had a series of metrics for tracking driver behavior, but now the company will be monitoring and analyzing driver behavior in poor and minority areas specifically as well as looking at drivers’ cancellation rates and their ratings there, using real-time tracking to screen for discrimination. Lyft says that it sends alerts for each trip cancellation, and discriminatory practices will result in “immediate termination.” We reached out to Lyft for more details and will update when we hear back.
The move is very similar to Airbnb’s community overhaul after a Harvard study found, yet again, discrimination against Black users. Airbnb began tracking hosts’ rejections more closely, offered more resources to guests who felt they were discriminated against, and prevented hosts from re-booking to a new guests after cancelling against another.
Will the new monitoring process halt discrimination on the platform? That depends on when it occurs. Don MacKenzie is an Assistant Professor at the University of Washington and one of the author’s of the original study. Dr. MacKenzie immediately notes that, unlike Uber, Lyft shows drivers a picture of riders before they accept a ride request. So if a Lyft driver doesn’t want Black riders, he can just reject the initial request. With Uber, drivers would accept request then reject when they inferred they’d have a Black rider.
“If the driver’s inclined to discriminate, they’ll just not accept that ride in the first place,” he says. “So there would be no reason that someone would basically cancel in a discriminatory manner. It’s kind of measuring the wrong thing here.”
Hypothetically, the entire approach could backfire by letting racist drivers off the hook for their discrimination; they could just never accept initial rides from Black people. If cancellation alone is being tracked, it doesn’t seem like this behavior would be flagged. So would it be better to track all rejections and cancellations, even before the ride request was accepted? Dr. MacKenzie says that just tracking drivers’ upfront rejections would pick up too much noise.
“They don’t want their trigger to be too sensitive,” he says. “You can get some false positives. They don’t want to start booting drivers off the platform based on flimsy evidence. I hope and I trust that they will be fairly prudent in how they do this.”
At best, the new process would mostly flag drivers that are racist and sloppy about the cover-up. But that doesn’t mean, Dr. MacKenzie notes, that the data isn’t worth tracking or that it’s not worth Lyft’s time to root out discrimination. He stands by the study’s findings that there are racist drivers out there discriminating on Lyft, but he says it’s likely a “small subset” of the driving force.
“Go read the comments section,” he said. “Those people are out there.”