Reductions in force: Delegating selection to AI creates its own risks

Your organization needs to implement a reduction in force (RIF), and you know you’re required to avoid discrimination in your selection criteria. What could be more objective and nondiscriminatory than an artificial intelligence (AI) evaluation of your employees? No surprise here, but AI assistance won’t insulate your organization from liability.

AI performance evaluations may not account for protected leave

Employees with lower productivity and efficiency may be among those identified for layoffs. AI can be used to identify these individuals, but without the proper guardrails, AI tools may be identifying employees who have requested or used protected leave or been provided an accommodation. Depending on the prompts and criteria AI analyzed, it’s possible that the performance criteria may have a disparate impact on protected individuals.

Potentially, an AI evaluation or algorithm could disproportionately identify employees based on any protected class. Under the West Virginia Human Rights Act, discrimination based on race, religion, color, national origin, ancestry, sex, age, blindness, or disability is prohibited. Consider just some of the ways protected classes could be identified as less productive:

• Sex: pregnancy- and maternity-related leave;

• Disability: reasonable accommodations and leave; and/or

• Religion: schedule changes or other accommodations to permit observance of religious beliefs.