Uber Eats - Discriminatory delivery algorithm

What Happened

Uber Eats utilized an AI-driven automated dispatch and rating algorithm to manage food delivery couriers. The system frequently suspended couriers' accounts or docked their pay based on automated "efficiency scores." Couriers' unions proved that the algorithm systematically penalized couriers from minority backgrounds and those working in congested urban areas, misclassifying delayed deliveries caused by systemic infrastructure issues or racial discrimination at restaurants as courier laziness. The automated rejections lacked transparent human review or appeals.

How MEG 1 would have acted

The delivery algorithm failed both the Dynamic Accuracy Index (DAI) and the Index of Safety and Responsibility (ISR) audits due to its systemic bias and violation of labor dignity and procedural fairness.

MEG 2 Analysis - Legal Framework

Agent level: N2 (management) - automated monitoring and decision-making system used by a commercial platform to govern labor contracts and access to work.
Cause (6.1):
(a) system defect - design flaws and bias built into the efficiency analysis and dispatch model (design defect), attributable to the manufacturer.
(b) autonomous decision error - the system's decision to autonomously suspend couriers' accounts or dock their pay.
Attachment of liability: Uber Eats as the operator is fully liable for the discriminatory outcomes and the lack of accessible human appeals; the developers are liable for design defects (6.1a).

Procedural Mechanism

6.2 (architectural human confirmation): Decisions involving labor contracts, account suspensions, or financial penalties must require architectural human confirmation. The AI must be technically blocked from autonomously executing penalties; a human officer must review and sign off.
7.1 (independent forensics): The monitoring logs, analyzed metrics, and the model's decision path must be securely logged in an independent ledger for labor audits.
9.4 (least privilege): The monitoring system was granted direct authority to execute disciplinary actions without human validation layers, violating safe operational parameters.