{"version":1,"type":"story","url":"https://digestai.news/story/university-of-washington-study-finds-ai-models-penalize-chinese-and-ni","json":"https://digestai.news/story/university-of-washington-study-finds-ai-models-penalize-chinese-and-ni.json","markdown":"https://digestai.news/story/university-of-washington-study-finds-ai-models-penalize-chinese-and-ni.md","slug":"university-of-washington-study-finds-ai-models-penalize-chinese-and-ni","headline":"University of Washington study finds AI models penalize Chinese and Nigerian accents","summary":"A University of Washington research team evaluated how leading AI models from Google, OpenAI and Alibaba scored spoken English with five accents – American, British, Chinese, Indian and Nigerian – across job interviews, conference presentations and English‑proficiency tests. The models consistently gave the highest scores to American and British speakers, placed Indian speakers in the middle, and ranked Chinese and Nigerian speakers lowest. Most of the penalty came from delivery, though content scores for Chinese and Nigerian speakers also slipped despite identical scripts.\n\nThe study, led by doctoral student Orevaoghene Ahia and associate professor Yulia Tsvetkov, notes that the models transcribed the accented speech accurately but still judged it as “mumbled” or “difficult to understand.” Funding came in part from gifts from Google and OpenAI. The researchers point to training data that under‑represents certain accents or describes them as unclear as possible sources of bias. They stress the stakes for companies that use AI‑based interview screening and for English tests such as Pearson’s PTE Academic, which was administered about 1.2 million times in 2023.\n\nThe team plans to extend the work to more regional accents and to explore AI handling of audio in medical settings, urging developers to publish evaluations that break performance down by dialect and subpopulation.","keyPoints":["AI models from Google, OpenAI and Alibaba ranked American and British accents highest, Chinese and Nigerian lowest across interview, presentation, test tasks","Study tested five accents—American, British, Chinese, Indian, Nigerian—and found delivery penalties drove most of the score gap","Research funded partly by gifts from Google and OpenAI; 1.2 million PTE Academic spoken answers were scored by AI in 2023"],"whyItMatters":"If AI scoring systems favor certain accents, hiring decisions and immigration tests could systematically disadvantage speakers of under‑represented dialects, reinforcing inequity.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":["Google","OpenAI","Alibaba","Pearson"],"models":[],"people":["Orevaoghene Ahia","Yulia Tsvetkov"]},"firstPublishedAt":"2026-11-02T06:50:00Z","updatedAt":"2026-11-02T06:50:00Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"geekwire.com","title":"Accent bias: AI models penalize certain speakers in interviews and tests, UW study finds","url":"https://geekwire.com/2026/accent-bias-ai-models-penalize-certain-speakers-in-interviews-and-tests-uw-study-finds","publishedAt":"2026-11-02T06:50:00Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"University of Washington study finds AI models penalize Chinese and Nigerian accents\", 2 November 2026, https://digestai.news/story/university-of-washington-study-finds-ai-models-penalize-chinese-and-ni","publisher":"Digest AI","title":"University of Washington study finds AI models penalize Chinese and Nigerian accents","datePublished":"2026-11-02T06:50:00Z","url":"https://digestai.news/story/university-of-washington-study-finds-ai-models-penalize-chinese-and-ni"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}