Yang joined Fortune’s CEOI Collaborative discussion this week to explain how the department is working alongside some of the largest corporations in the U.S. to address diversity training and hiring. She also shared concerns about the software programs companies are using to find—and weed out—potential job candidates.

After the events of last summer, when George Floyd’s murder sparked protests and an anti-racism movement, it led many CEOs to institute diversity training programs in their offices as well as diversity working groups. Still, according to Refinitiv’s ESG database, Black Americans hold just 5% of manager positions in the 80 Fortune 500 companies for which data was reported, and Hispanic Americans held just 6% of managerial positions.

The DOL is prioritizing open lines of communication with Fortune 500 companies, said Yang, so that they can share data and help direct corporate resources to where they have the greatest impact on advancing equity in the workplace. It’s important, she said, to understand employers’ strategic objectives so that the department can “align our objectives with the business objectives.” Working in unison, she said, usually increases employee engagement and leads to a greater advancement of corporate commitments.  

The DOL is also looking to collaborate across the government and with private enterprise in creating and utilizing new technologies in the hiring process that make diversity and equity a priority. 

“Long-standing hiring practices that we use today like using subjective evaluations or reviews or unstructured interviews can allow bias to enter the process,” Yang explained at the discussion for CEOs hosted by Fortune. Employers have come to her department, she said, looking for alternatives to help them make job-related decisions with less subjectivity. 

The Biden administration, said Yang, is particularly interested in understanding how technology is currently being employed in the workplace to evaluate résumés and interviews. 

“From our standpoint in the government, it’s critical that we understand how these models are making decisions, so that we can assess whether they are operating in a way that’s fair,” she said. Often technology developers express concern with sharing that information with employers because of proprietary reasons, and that’s where the government could step in, she noted: “It’s critical for employers to understand how a system works, and not just rest on claims that systems are unbiased.”  

Yang pointed to Amazon as an example. The company opted to abandon its own A.I. screening program when it discovered that the system had taught itself to prefer male candidates over women based on the company’s past hires. 

“I do think there is an important opportunity,” said Yang, “for greater transparency about how the systems work.”