BiasMedium
Common Biases of Vector Embeddings
- Occurred
- Jul 21, 2016
- Published
- Jul 21, 2016
- Source
- AI Incident Database
Summary
Researchers from Boston University and Microsoft Research, New England demonstrated gender bias in the most common techniques used to embed words for natural language processing (NLP).
Developers
- Microsoft Research
- Boston University
Deployers
- Microsoft Research
- Boston University
Harmed parties
- Women
- Minority groups
- Women and girls
Classification
Category confidence: 95%
Severity confidence: 90%
Rationale: Research paper demonstrating gender bias in NLP embedding techniques without reported real-world harm.
Classifier: Qwen/Qwen3.5-35B-A3B-FP8