tech
AI is more likely than humans to form biases when hiring
AI doesn’t just learn stereotypes from its training. It can cook up new ones, too.
TL;DR
- LLMs can develop their own biases from experience, not just their training data.
- In a simulated hiring game, LLMs stereotyped job applicants more than human participants.
- LLMs tend to generalize rapidly from limited data, which can lead to premature stereotyping.
- Newer models with higher reasoning capabilities showed stronger biases.
- Incentivizing diverse hiring and providing relevant personal information reduced bias in LLMs.
- The real-world implications are significant as AI is increasingly used in hiring, loan, and parole decisions.