Artificial intelligence systems used to screen job applications may be more prone to bias than human recruiters, according to new research reported by MIT Technology Review. As AI increasingly handles initial resume reviews before applications reach human eyes, concerns are growing about whether these systems can evaluate candidates fairly.
While researchers have previously established that large language models (LLMs) absorb human biases present in their training data, the new research suggests a more complex problem. According to MIT Technology Review, LLMs can develop their own distinct biases beyond those inherited from training materials, raising fresh questions about the reliability of AI-powered hiring tools.
The findings add to mounting concerns about algorithmic fairness in employment decisions. With many companies adopting AI screening tools to manage high volumes of applications, the research highlights potential risks in delegating these critical decisions to systems that may introduce or amplify discriminatory patterns in ways that differ from human bias.