Amazon’s biased AI recruiting tool
Summary: Amazon’s biased AI recruiting tool
This article explains how Amazon built an experimental AI recruiting system to score job applicants, but eventually scrapped it after discovering it showed bias against women. The system was trained on around 10 years of past resumes, most of which came from men because tech hiring had historically been male-dominated. As a result, the AI learned patterns that favored male applicants and penalized resumes containing terms like “women’s,” such as “women’s chess club captain.” It also downgraded graduates from some all-women’s colleges.
Amazon tried to adjust the tool, but the company could not guarantee the AI would not find other biased patterns. The project was ultimately shut down, and recruiters never relied only on the system’s rankings. The article is a strong example of how AI can reproduce bias from historical data, especially when used in important decisions like hiring.
Link: euronews.com/business/…/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women