Uber’s facial recognition is locking Indian drivers out of their accounts
Some people are finding their accounts permanently blocked
Correction: this story has been updated to include Uber’s response. The opening has been amended to remove an anecdote about a specific driver’s experience based on that response.
Uber drivers in India say that problems with the facial recognition they use to access their apps is costing them work. In a survey conducted by MIT Technology Review of 150 Uber drivers in the country, almost half say they have been either temporarily or permanently locked out of their accounts as a result of problems with their selfie. Many suspected that a change in their appearance, such as facial hair, a shaved head, or a haircut, was to blame. Another quarter of them believe it was due to low lighting.
India has around 600,000 Uber drivers, but many others work for the homegrown ride-sharing platform Ola and for startups such as Swiggy, Zomato, and Urban Company. All ask their platform workers to upload selfies for logins or verifications.
Uber checks that a driver’s face matches what the company has on file through a program called “Real-Time ID Check.” It was rolled out in the US in 2016, in India in 2017, and then in other markets. “This prevents fraud and protects drivers’ accounts from being compromised. It also protects riders by building another layer of accountability into the app to ensure the right person is behind the wheel,” Joe Sullivan, Uber’s chief security officer, said in a statement in 2017.
But the company’s driver verification procedures are far from seamless. Adnan Taqi, an Uber driver in Mumbai, ran into trouble with it when the app prompted him to take a selfie around dusk. He was locked out for 48 hours, a big dent in his work schedule—he says he drives 18 hours straight, sometimes as much as 24 hours, to be able to make a living. Days later, he took a selfie that locked him out of his account again, this time for a whole week. That time, Taqi suspects, it came down to hair: “I hadn’t shaved for a few days and my hair had also grown out a bit,” he says.
Others say they have struggled with scratches on their cameras and low-budget smartphones. The problem isn’t unique to Uber. Drivers with Ola, which is backed by SoftBank, face similar issues.
Uber did not initially respond to a series of questions from MIT Technology Review about its technology and drivers’ experiences. After publication, the company said that no deactivations can take place based on facial recognition alone. It said that Real-Time ID Check works by escalating flagged non-matches to at least two humans who then check the photos manually. This process takes less than 25 seconds. It says that its tool can handle changes in hair.
Nevertheless, more than a dozen drivers interviewed for this story detailed instances of having to find better lighting to avoid being locked out of their Uber accounts. “Whenever Uber asks for a selfie in the evenings or at night, I’ve had to pull over and go under a streetlight to click a clear picture—otherwise there are chances of getting rejected,” said Santosh Kumar, an Uber driver from Hyderabad.
Real-Time ID Check works by converting your face into a set of points, explains Jernej Kavka, an independent technology consultant with access to Microsoft’s Face API, which is what Uber uses to power Real-Time ID Check.
“With excessive facial hair, the points change and it may not recognize where the chin is,” Kavka says. The same thing happens when there is low lighting or the phone’s camera doesn’t have a good contrast. “This makes it difficult for the computer to detect edges,” he explains.
The software may be especially brittle in India. In December 2021, tech policy researchers Smriti Parsheera (a fellow with the CyberBRICS project) and Gaurav Jain (an economist with the International Finance Corporation) posted a preprint paper that audited four commercial facial processing tools—Amazon’s Rekognition, Microsoft Azure’s Face, Face++, and FaceX—for their performance on Indian faces. When the software was applied to a database of 32,184 election candidates, Microsoft’s Face failed to even detect the presence of a face in more than 1,000 images, throwing an error rate of more than 3%—the worst among the four.
It could be that the Uber app is failing drivers because its software was not trained on a diverse range of Indian faces, Parsheera says. But she says there may be other issues at play as well. “There could be a number of other contributing factors like lighting, angle, effects of aging, etc.,” she explained in writing. “But the lack of transparency surrounding the use of such systems makes it hard to provide a more concrete explanation.”
Uber said that it regularly conducts fairness testing (both internally, and in partnership with Microsoft) to understand how this product is performing for users with different skin complexions and identify any opportunities to improve efficiency and accuracy. Microsoft declined to comment in response to questions sent by MIT Technology Review.
The problems don’t end with the algorithm’s decision. Drivers say the grievance redress mechanism that Uber follows is tedious, time-consuming, frustrating, and mostly unhelpful. They say they sometimes spend weeks trying to get their issues resolved. “We have to keep calling their help line incessantly before they unlock our accounts, constantly telling us that the server is down,” said Taqi, with a tone of frustration—but mostly a sense of defeat—in his voice. “It’s like their server is always down.”
Uber said it gives drivers a chance to appeal deactivations.
If problems persist—and protections remain limited—they could have an outsize effect, and not just on work. “Labor platforms in India are starting to become a key interface between the worker, the market, and the government—they enable loans for cars or even credit for larger household expenses,” says Aditi Surie, a senior researcher at the Indian Institute for Human Settlements, who has done research on gig work in India. In a country where such work can catapult someone from precarity to a middle-class existence (especially when estimates suggest that the majority of people worldwide who fell into poverty during the pandemic live in India), getting blocked from or kicked off a platform can have devastating consequences.
In other markets, gig workers have fought back against facial recognition. In the UK, for example, at least 35 Uber drivers claimed last year that their accounts were wrongly terminated. The Independent Workers’ Union of Great Britain has blamed a “racist algorithm.” Uber has faced at least two lawsuits in the UK because of the software. Uber said that in the UK and EU, drivers can choose whether their selfie is checked by verification technology or by human reviewers.
Some countries and regions have moved to provide better protections for gig workers. The EU proposed a directive last year to improve working conditions and provide algorithmic transparency. And in September 2021, California court struck down Proposition 22, a ballot initiative that excluded gig workers from employee benefits under state law. These regulations recognize that algorithmic systems can “negatively impact the rights of workers,” says Divij Joshi, a lawyer and a PhD candidate at University College London. But India currently has few legal protections in place for gig workers, Joshi says: “These same transparency efforts are not being seen in India from a policy or regulatory lens.”
Elizabeth Anne Watkins, an organizational sociologist from Princeton University who has extensively studied the impact of facial recognition on Uber drivers in the US, would likely find this pattern familiar. “Prone to malfunction in variable conditions, the system places a heavy burden on workers who are left with little organizational support when facial recognition fails,” Hawkins, who is now a research scientist at Intel Labs, wrote in a 2020 paper. “Further, accountability for identity verification is shifted to the workers, who bear the consequences for systemic failures.”
Samantha Dalal, who studies how workers understand algorithmic systems, says there could be more transparency about how the AI made a decision. “Including some explanation that goes beyond ‘You are deactivated’” would help, says Dalal, a doctoral candidate at the University of Colorado Boulder. “Such capabilities exist.”
Absent any insight into what the mercurial, non-human boss wants, gig workers attempt a lot of trial and error while interacting with the apps, Dalal says. In an email, Uber said it sends multiple clear warnings before account deactivation.
Varsha Bansal is a freelance journalist based in Bangalore. Reporting for this story was supported by Pulitzer Center’s AI Accountability Network.
The inside story of how ChatGPT was built from the people who made it
Exclusive conversations that take us behind the scenes of a cultural phenomenon.
AI is dreaming up drugs that no one has ever seen. Now we’ve got to see if they work.
AI automation throughout the drug development pipeline is opening up the possibility of faster, cheaper pharmaceuticals.
GPT-4 is bigger and better than ChatGPT—but OpenAI won’t say why
We got a first look at the much-anticipated big new language model from OpenAI. But this time how it works is even more deeply under wraps.
The original startup behind Stable Diffusion has launched a generative AI for video
Runway’s new model, called Gen-1, can change the visual style of existing videos and movies.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.