This voice is automatically generated. Please let us know if you have any feedback.
Diving overview:
- Human recruiters mainly Adopt bias in artificial intelligence tools Researchers at the University of Washington found in a recent study detailed in a Nov. 10 press release that they use this information when selecting job applicants.
- In this study, 528 participants were asked to select candidates for 16 different jobs using a simulated large-scale language model. Researchers simulated an AI model with varying levels of bias that generated hiring recommendations for resumes submitted by hypothetical white, black, Hispanic, and Asian men with the same qualifications.
- Results showed that participants who chose candidates without using AI or with “neutral” AI chose white and non-white candidates at the same rate. However, those using a moderately biased AI showed a preference for either white candidates or non-white candidates who matched the AI’s biased recommendations. Participants who used the severely biased AI model only made slightly less biased decisions than the ones recommended.
Dive Insight:
of Combining AI and human recruiters “This idea is becoming increasingly dominant,” Kyla Wilson, a doctoral student at the University of Washington and lead author of the study, said in a press release. Given this reality, the researchers set out to understand how AI technology influences hiring managers’ decisions, Wilson said.
“Our findings were clear: people were perfectly willing to accept bias in AI unless the bias was obvious,” Wilson said.
According to a recent report from Employ Inc., 65% of hiring managers Use AI in your workflows Additionally, 52% said they plan to invest in new recruitment technology, indicating that adoption is accelerating. In an even more extreme finding, a Resume.org report published in August found that one-third of U.S. workers believe their employer’s hiring process will: Completely powered by AI by 2026.
Concerns about bias in AI adoption have been expressed throughout the hype cycle surrounding AI technology in the early 2020s. A University of Washington study found that participants using a severely biased AI model made less biased decisions, but still followed the AI’s suggestions about 90% of the time. But Wilson said employers can reduce bias by implementing better models.
Another way to combat bias is to require recruiters to perform implicit association tests to help detect subconscious biases. The university announced a 13% reduction in bias among participants who began the study with such a test.
Airin Kalliskan, associate professor at the university’s School of Information Studies, said in a release that scientists who design AI models have a role to play in reducing bias, as do policy makers.
“People have agency, and that has huge impacts and consequences, and we shouldn’t lose our ability to think critically when interacting with AI,” Kalliskan said. “But I don’t want to put all the blame on the people who use AI.”
Employers may also need to note the following: Establishment of state and local laws It regulates the use of AI in recruitment. For example, California recently issued regulations requiring employers subject to the state’s consumer privacy law to provide pre-use notices to potential candidates about the use of AI tools and to respond to requests for consumer information regarding automated recruiting tools. Employers will also be required to carry out a risk assessment of such tools in accordance with regulations that come into force in January 2027.
