Study finds hiring LLMs can form biases stronger than humans
Turns out, AI isn't just copying human biases: it can actually invent its own.
A new study from Princeton and the University of Chicago found that large language models (LLMs) used for hiring could form even stronger biases than people when picking candidates based on feedback from simulated tasks.
This has experts worried about how much influence AI has in workplace decisions.
LLMs favor groups despite equal qualifications
The researchers noticed that these AI systems sometimes favored certain groups just because of past performance, even if everyone was equally qualified.
The researchers wrote that "LLMs are not merely passive mirrors of human social biases, but can actively create new ones from experience."
With more than 90% of companies using AI for hiring (and legal cases already popping up), there's a real need to build better safeguards so tech doesn't accidentally make things less fair.