Uber's facial recognition system used for driver verification in India locked out nearly half of 150 surveyed drivers from their accounts, causing loss of income due to false rejections from appearance changes, poor lighting, and algorithmic bias.
Uber deployed a facial recognition system called 'Real-Time ID Check' in India in 2017, using Microsoft's Face API to verify driver identity through selfies before allowing access to the app. A survey by MIT Technology Review of 150 Uber drivers found that almost half were temporarily or permanently locked out of their accounts due to facial recognition failures. Many drivers suspected appearance changes like facial hair growth, haircuts, or shaved heads caused the rejections, while a quarter blamed poor lighting conditions. The system particularly struggled with Indian faces, as research showed Microsoft's Face API had over 3% error rates on a database of Indian election candidates. Drivers reported having to find streetlights for better lighting and spending weeks trying to resolve account lockouts through Uber's customer service. One driver, Adnan Taqi in Mumbai, was locked out for 48 hours and then a full week, disrupting his 18-24 hour daily driving schedule. Another driver, Neradi Srikanth, claimed he was permanently banned after shaving his head for a religious visit, though Uber disputes this was due to facial recognition. The system affects approximately 600,000 Uber drivers in India, with similar issues reported for other platforms like Ola. Research suggests the facial recognition software may not have been adequately trained on diverse Indian faces, contributing to higher failure rates for drivers in the region.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Accuracy and effectiveness of AI decisions and actions are dependent on group membership, where decisions in AI system design and biased training data lead to unequal outcomes, reduced benefits, increased effort, and alienation of users.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed