Best of CASE District VIII Award
 

Institution: University of Washington

Title of entry: AI systems for hiring can reproduce societal biases

About this entry: Artificial intelligence systems in 2024 grew increasingly commonplace — including in the hiring process. While AI systems that recommend socks might not shape users’ lives, AI is already used in every stage of hiring, which has major societal consequences.

Some experts hope AI can make hiring fairer, but in the past year, University of Washington researchers explored various ways that bias can creep into AI currently used for hiring. To help a general audience understand the biases lurking in these systems, Stefan Milne from UW News wrote three press releases about research concerning bias in AI hiring systems, with the goal of explaining the studies for the general public and demonstrating the impact of UW research to identify problems in emerging technologies.

These stories discussed how ChatGPT ranked resumes with disability credentials lower than those without the credentials; how three other AI systems exhibited significant bias when ranking applicants based on perceived gender and race; and how some AI systems, when used to discuss different job applicants, showed subtle biases in their discussions of race and caste.

In total, the press releases resulted in media features in four local and national outlets including Forbes, GeekWire and NPR affiliate Oregon Public Broadcasting, as well as attention in other venues. The stories showcased the UW’s mission to investigate possible problems and dangers of new technologies, as well as potential solutions. They also furthered UW President Ana Mari Cauce's goal to be "the number-one public university in terms of impact."