Accent bias: AI models penalize certain speakers in interviews and tests, UW study finds

A UW study tested how major AI models evaluate English speech across five accents in job interviews, presentations and proficiency tests. American and British speakers consistently ranked highest, Indian speakers in the middle, and Chinese and Nigerian speakers lowest. Despite accurate transcription capabilities, models penalized non-Western accents on delivery and sometimes content, even when all speakers used identical scripts.

Summary and key points written by AI from the headline and a short excerpt, so they can be incomplete or wrong. GeekWire did the reporting; read the original for the full story.