Amazon's AI-powered Flex delivery system automatically terminated contract drivers like Stephen Normandin and Neddra Lira through algorithmic decisions with minimal human oversight, causing financial hardship for workers who were fired for circumstances beyond their control.
Amazon's Flex delivery program uses algorithms to monitor, rate, and terminate contract drivers with minimal human oversight. The system tracks driver performance through metrics like punctuality, delivery completion, and following customer instructions, rating drivers as Fantastic, Great, Fair, or At Risk. Bloomberg interviewed 15 Flex drivers, including four who claimed wrongful termination. Stephen Normandin, a 63-year-old Army veteran, was terminated after nearly four years when algorithms decided he wasn't performing adequately due to issues like locked apartment complexes and malfunctioning Amazon lockers. Neddra Lira, a 42-year-old school bus driver, lost her Flex income during the pandemic after her rating dropped following a tire problem, leading to car repossession and near loss of her house. Former Amazon managers acknowledged the company knew the automated system would make mistakes but considered it cheaper than human oversight. The system processes appeals through largely automated responses, with drivers having 10 days to appeal and the option to pay $200 for arbitration. Amazon has 4 million global Flex drivers, with 2.9 million in the US, and the system maintains about 95% successful deliveries while algorithms examine the remaining 5% for problematic patterns.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Social and economic inequalities caused by widespread use of AI, such as by automating jobs, reducing the quality of employment, or producing exploitative dependencies between workers and their employers.
AI system
Due to a decision or action made by an AI system
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed