Why Racial Bias is Prevalent in Facial Recognition Technology

By Harvard Journal of Law & Technology.

Summary: Explains the technical and social roots of racial bias in facial recognition systems and its discriminatory effects.

Relevance: Squarely in the Inequality & Algorithmic Harm subdomain — directly relevant to social workers whose clients may be subject to facial recognition in policing, housing, or benefits verification, since misidentification risk falls unevenly by race.

Key findings: NIST’s 2019 test of 189 algorithms from 99 developers found many were “10 to 100 times more likely to misidentify a Black or East Asian face than a white face,” with Black women faring worst; training datasets skew heavily white (Labeled Faces in the Wild is 83.5% white), and the ACLU’s own test wrongly matched 28 members of Congress, disproportionately people of color, to mugshots.

Read the original: https://jolt.law.harvard.edu/digest/why-racial-bias-is-prevalent-in-facial-recognition-technology