Uber Rolls Out AI To Its App Verifying To Riders That Drivers Are Wearing Masks

Computer vision inaccuracy could be a concern, but the company is betting that customers will be more inclined to begin reusing the service if drivers are wearing masks. JL

Kyle Wiggers reports in Venture Beat:

Uber will roll out technology to its mobile app to verify whether drivers are wearing a mask. The app will apply a computer vision algorithm to drivers’ selfies to detect the presence (or absence) of face coverings. Riders will be alerted via an in-app message after they hail a ride that their driver is covering their face. Drivers who don’t wear cover their face won’t be able to pick up customers beginning on May 18. Studies have found that computer vision systems tend to misidentify women and people of certain ethnicities more often than white males.

Uber today announced it will roll out technology to its mobile app to verify whether drivers are wearing a mask. As part of a new checklist designed to ensure drivers have taken safety measures to prevent the spread of coronavirus, the app will apply a computer vision algorithm to drivers’ selfies to detect the presence (or absence) of face coverings. Riders, who won’t be subject to the same process but who must confirm that they’ve taken the same precautions, will be alerted via an in-app message after they hail a ride that their driver is covering their face. Drivers who don’t wear cover their face won’t be able to pick up customers beginning on May 18, and those who repeatedly show up without a mask will be subject to losing their access to Uber.

Uber says the new system, which will be in place through at least the end of June, is intended to adapt to changing public health guidance and regulations around the pandemic. It follows the launch of a trial in France to integrate AI tools from startup DatakaLab into security cameras in the Paris metro system to check whether passengers are wearing face masks, and it speaks to something of a growing trend. Models suggest that that donning masks can significantly reduce the spread of coronavirus when a vast majority of the population participates, but compliance remains an issue in countries like the U.S., where face coverings have become a political symbol. AI systems like Uber’s and DatakaLab’s could help ensure the safety of passengers and drivers while reducing confrontation.Short of clarifying that the mask-checking system doesn’t process biometric information or compare selfies to driver photos in its database, Uber was vague about the details. This is somewhat concerning — the company’s Real-Time ID Check system, which is designed to reduce fraud by periodically prompting drivers to submit selfies before they begin a driving session, caused a transgender Iowa driver to miss out on about three days of work, including a lucrative holiday. A number of studies, such as one published by the U.S. National Institute of Standards and Technology in December, have found that computer vision systems tend to misidentify women and people of certain ethnicities more often than white males. I

n response to the incident involving the Iowa driver, Uber said that Real-Time ID Check — which is powered by Microsoft’s Azure Cognitive Services — ultimately verified 99% of drivers during a pilot in 2016. And it says that the new mask-wearing check uses a separate system that doesn’t leverage any Real-Time ID Check technology. Still, it remains to be seen whether it’s susceptible to mistakes — particularly at a time when ride-hailing drivers are struggling to make ends meet.

“During the first phase of the crisis, it’s been heartening to see our communities come together with a collective sense of responsibility,” wrote Uber CEO Dara Khosrowshahi in a statement. “As we move into the new normal, it’s our goal to encourage this same sense of shared responsibility: Uber, riders, drivers, delivery people and restaurants — we all have a role to play in keeping each other safe and healthy.”