AI Hiring Tools May Treat Boys and Girls Differently
Computers are now helping people find jobs. Some smart programs write resumes. Others read those resumes and pick who gets an interview. Researchers wanted to know if these tools treat men and women the same. They ran tests where gender was the only thing that changed. Everything else stayed identical.
The results showed clear patterns. The computer programs often wrote resumes that matched old gender stereotypes. When scoring applicants, the tools gave lower marks to women for jobs usually done by men. They gave higher marks to women for jobs usually done by women. This happened even when all other details on the resume were exactly the same.
Pay suggestions also followed old gaps. When the prompt said whether the applicant was a man or woman, the suggested pay matched real-world wage gaps. This effect got weaker when only names hinted at gender. Different computer programs acted differently too. The bias comes from two places. One is judging the same resume differently based on gender. The other is when the computer writes gendered words that then shape how the resume is judged later.
Because one tool can write the resume and also grade it, these two problems can feed each other. That means small biases can grow bigger. These findings matter because more employers are using AI every day. People may ask who gains when machines help pick workers.