Large-scale English text corpora reflect real-world societal biases

Categorization

• corroborated: 2 independent passages support this claim
• XREF: connects to: Word embeddings’ gender skew correlates with real occupational gender ratios; Word embeddings serve as novel objects for sociological analysis of bias; Machine learning increasingly intertwines technical problems with human and social questions (+7 more)

Related

• Word embeddings serve as novel objects for sociological analysis of bias (similarity 0.95)
• Common language carries more gender bias than objective metrics show (similarity 0.87)
• Hundreds of papers missed blatant sexism in word embeddings (similarity 0.87)
• Word embeddings’ gender skew correlates with real occupational gender ratios (similarity 0.82)

Library Connections

• Even massive, comprehensive text datasets reflect biases present in the real world #corroborates (0.95)
• CLAUDE #complicates (0.75) — The second channel — because embeddings miss the best links

Nearest-neighbor retrieval systematically finds the *obviou…

Broader Connections

• Language can distort as well as transmit human culture #expands (0.60)
• Language is socially constructed and speaks through individuals #corroborates (0.50)