EEOC update: Artificial intelligence and discrimination

Excerpted from a Venable LLP blog by Thomas H. Strong and Robin L.S. Burroughs

On October 28, the U.S. Equal Employment Opportunity Commission (EEOC) announced an initiative to ensure the use of artificial intelligence (AI) in hiring and employment decisions complies with federal civil rights laws. The EEOC says the initiative will guide applicants, employees, employers and technology vendors in ensuring AI is used fairly and consistently.

In the agency’s announcement, EEOC Chair Charlotte A. Burrows said “the EEOC is aware these tools may mask and perpetuate bias or create new discriminatory barriers to jobs. The EEOC must work to ensure these new technologies do not become a high-tech pathway to discrimination.”

The EEOC’s new initiative will:

• Establish an internal group to coordinate the work;
• Launch a series of sessions with stakeholders about algorithmic tools;
• Gather information about the design of employment technologies;
• Identify promising practices; and
• Issue technical assistance to provide guidance on algorithmic fairness.

As the EEOC prepares to study AI, employers need to be aware of how it might be used in their hiring and firing processes.

AI in Recruitment and Screening
One way AI has broken into employment law is through the use of machine-learning recruiting engines. Recruiting engines learn an employer’s ideal candidate and screen résumés to match the desired characteristics. Despite being time and cost effective, the tools often fail to rise above human biases regarding gender and race.

For example, Amazon’s recruiting engine came under fire for discriminating among applicants. Amazon’s experimental AI tool was designed to comb through résumés and rank candidates.

Within one year, it was clear the program was not ranking candidates in a gender-neutral manner. Amazon’s AI model was trained to vet current candidates based on patterns in résumés accepted over the previous ten years. Given male dominance in the tech industry, a large majority of the historical applications came from men. As a result, the AI began downgrading résumés that included the word “women’s” and even penalized applicants from women’s colleges. In effect, Amazon’s AI had learned to prefer male candidates based on Amazon’s prior hiring practices. Amazon ultimately disbanded the project.

The use of AI in recruiting and screening candidates must be carefully monitored to ensure no patterns of bias emerge. Such bias is not limited to overt discrimination based on protected characteristics: machines can just as easily prefer candidates from only a small group of universities, people who use similar language on their résumés, or people with similar professional backgrounds. Although these may seem like neutral qualifications, they can also work to screen out candidates based on legally protected characteristics.

Speech Recognition Models
In addition to candidate screening AI, it has become increasingly common for employers to rely on speech recognition technology in the hiring process. Employers have turned to speech-to-text technology such as conversational AI and video interviewing to make hiring practices and other HR tasks more efficient and cost-effective.

A recent Stanford University study found that speech-to-text AI developed by Amazon, IMB, Google, Microsoft and Apple misidentified the words of African Americans at double the rate of whites. Furthermore, a full 20% of transcriptions from black speakers were misinterpreted so drastically as to render them unusable – compared to only 2% for white speakers. If an employer uses such technology – for example, to streamline answering employee questions about HR benefits or to gauge employee morale through a survey – it may be inadvertently favoring employees of one race over another.

Conclusion
Employers are increasingly adopting AI technologies to streamline and automate the hiring process. They may think the use of AI can eliminate human and implicit bias from the hiring equation. But as noted, it can do more harm than good if companies are not strategic and thoughtful about its use. As companies continue to embrace AI, employers must be aware of the EEOC’s heightened concern over discrimination within AI programs and work to ensure new technology does not create new liability.

For the full story, please click here.