Robert Williams and Facial-Recognition Technology
The wrongful arrest of Robert Williams demonstrates why algorithmic output should not be treated as conclusive evidence. In January 2020, Detroit police arrested Williams after facial-recognition software incorrectly matched his driver’s-license photograph to a person shown in low-quality surveillance footage. Police relied on the match while failing to obtain sufficient independent evidence connecting him to the theft. Williams was detained for approximately 30 hours before being released. The ethical issue extends beyond police conduct: what responsibilities belong to software developers, database operators, agencies purchasing these technologies, and officers interpreting the results? This case also raises questions about demographic differences in facial-recognition accuracy and the danger that people may place greater trust in a computerized result than it deserves. It highlights the need for testing, disclosure, human review, corroborating evidence, and procedures allowing people to challenge algorithmic decisions.