How Can AI Resume Screening Introduce Bias?

**Name and demographic signals:**

• Names can signal gender, ethnicity, and national origin
• Address/ZIP code correlates with race and socioeconomic status
• Graduation year reveals age

**Experience and education bias:**

• Prestige-based filtering (top companies, elite universities)
• Keyword matching that favors specific communication styles
• Gap penalties that disadvantage caregivers and medical leave
• Foreign institution and credential devaluation

**Language and formatting:**

• NLP models trained on native English text
• Resume format preferences that vary by culture
• Action verb preferences that may be gendered

**Feedback loops:**

• Models trained on historical hiring decisions encode past biases
• "Successful candidate" profiles reflect who was hired, not who would perform best