"You Could Get Shot Pulling Over to Take Picture Where I Work": Facial Recognition in Gig Work

• In Event: Algorithms at Work: The Practice of Prediction - II: Case Studies of Prediction in Practice

Abstract

Algorithmic decision-making in gig work has occasioned shifting relations between workers, technologies, and organizations (Rosenblat and Stark, 2016). One understudied application of this decision-making is in account security, and the use of biometric and behavioral data to “authenticate” workers’ identities. Uber in 2016 rolled out a facial-recognition security measure called “Real Time Check ID,” where a driver logged into the Uber app is sent a message to pull over and take a photo of themselves using their phone. This photo is then analyzed with third-party software (Microsoft’s Cognitive Services) and compared with the photo attached to the account. If they don’t match, the account is deactivated. Uber says this is meant to protect riders from fraudulent drivers, and drivers from account highjacking.

STS scholars have long been concerned with the sociomateriality of technology at work (Orlikowski, 1992). With this lens, how is ‘selfie verification’ (as it’s been called) thought about, acted upon, and interacted with by workers? Thinking more broadly, how do these processes play out within, or at the “secure” boundaries of, organizations integrating facial recognition, such as in public transit, education, and corporate human resources?

To examine workers’ meaning-making about Real Time Check ID I analyze comments posted to online messageboards (Greene and Shilton, 2017) by drivers over the past three years. Their accounts (one example in the title of this paper) evince wide variation in interpretations of, and practices around, this increasingly pervasive form of algorithmic decision-making.

Author

• Elizabeth Anne Watkins, Columbia University