Take steps to mitigate AI risks in HR
According to the Equal Employment Opportunity Commission (EEOC), nearly 8% of employers use artificial intelligence (AI) technology to make employment decisions. The prevalence of AI technology in the employment arena shouldn’t surprise HR professionals with long to-do lists. AI can assist with résumé reviews, training, and employee evaluations. While AI can be a powerful tool, it can also result in violations of Title VII of the Civil Rights Act of 1964 and the Fair Labor Standards Act (FLSA) if used incorrectly.
AI discrimination
Common Title VII issues arise with algorithms that screen résumés. Algorithms are often “trained” to compare new résumés with past résumés of high-performing employees. Potential disparate impact discrimination could occur if the past résumés consist of candidates from a predominant group.
Imagine an industry that historically has more male employees than female employees, such as engineering. In such an industry, the screening algorithm may eliminate résumés from female candidates. Gender in a résumé may be revealed through education, civic activities, or even an applicant’s name. When the AI tool finds these “female” characteristics, it may eliminate female applicants from consideration.
Based on the risks of disparate impact discrimination with AI, the EEOC issued technical guidance on May 18, 2023, entitled “Assessing Adverse Impact in Software, Algorithms, and Artificial Intelligence Used in Employment Selection Procedures Under Title VII of the Civil Rights Acts of 1964.”
According to the guidance, employers should inquire with AI vendors about what steps have been taken to prevent disparate impact discrimination. Keep in mind, however, you are still responsible for Title VII violations caused by AI tools purchased from vendors regardless of the vendor’s predictions or studies regarding disparate impact discrimination.
Disparate impact discrimination could be found if an AI tool chooses individuals of a particular race, color, religion, sex, or national origin or individuals with a combination of those characteristics at a selection rate that is less than 4/5 (or 80%) of the rate of the nonprotected group. This is known as the four-fifths rule.
For example, if your selection rate for white applicants is 70% and for Hispanic applicants 35%, you could have a problem with disparate impact. Because 35% is half of 70% and 50% is lower than 4/5 (or 80%), the four-fifths rule says the selection rate for Hispanic applicants is substantially different from the selection rate for white applicants, which could be evidence of discrimination against Hispanic applicants.
FLSA violations
In addition to potential disparate impact discrimination, AI may cause an employer to inadvertently violate the FLSA. The Department of Labor (DOL) has highlighted the following risks with AI:
- AI tools that track active and idle time, such as keystrokes or eye movements, could result in the underreporting of hours. Imagine if a supervisor telephones a remote employee for a lengthy impromptu conference, and the employee steps away from the computer for the call. Because the employee is not at the computer, the impromptu meeting may not be captured as hours worked.
- Some AI tools treat all breaks as noncompensable. Under the FLSA, however, breaks lasting 20 minutes or less are compensable.
- Location-monitoring tools may track only work performed at a certain location. What happens if an employee normally assigned to one location is sent to work at another location?
Takeaway
To help minimize the risk of inadvertent FLSA violations, it’s important to implement a reporting mechanism in which nonexempt employees can record hours worked that they feel weren’t captured by AI. In addition, you should use these reports to adjust your AI technologies to ensure all hours worked are captured. And if you use AI in recruiting and hiring, be sure you pay attention to signs of potential disparate impact. By taking these steps, you lower your risk of running afoul of the FLSA and Title VII.
