Not so fast, says the U.S. Equal Employment Opportunity Commission (EEOC), which recently released new guidance addressing the interplay between an employer’s use of AI and other algorithmic decision-making tools and the Americans with Disabilities Act (ADA).
In particular, the EEOC warns that reliance on AI in making hiring and other personnel decisions, or in monitoring employee activity and performance, could adversely impact and discriminate against qualified individuals with disabilities. According to the EEOC, AI may not take into account one’s disability status or need for job-related accommodation when performing its automated analysis. For example, if an individual with a disability has a gap in their employment history due to being unable to work for a period of time, AI might reject the applicant based on the period of unemployment without any further inquiry. According to the EEOC, that could have a disparate impact on qualified individuals with disabilities.
Employers should also heed the EEOC’s guidance when using AI to monitor current employees’ work performance. AI might be used to measure and analyze employee output or activities, but it might not consider whether an employee’s substandard performance relates to a limiting medical restriction and/or is covered by an approved medical workplace accommodation. Per the EEOC’s guidance, employers relying on that AI analysis might inadvertently discriminate against a qualified individual with a disability who might be able to meet the employer’s expectations with reasonable accommodation.
To avoid these risks, the EEOC encourages employers to, among other things, inform candidates and employees of the evaluation process being employed and explain the factors being taken into account. To that end, employers may consider incorporating into their AI-driven processes appropriate human “checks” for reviewing the AI’s analysis before making final determinations. They also may consider incorporating methods by which employees and candidates with disabilities can request reasonable accommodation and commence an interactive process with the employer over any reported medical restrictions that might be limiting the individual.
The EEOC’s guidance suggests that AI-related reasonable accommodations will be similar to those already provided by employers in a non-AI context. For example, the employer may need to offer the individual more time to complete an AI-driven pre-employment evaluation, or provide an audio option for a visually-impaired candidate, or provide other reasonable accommodation that might be warranted under the circumstances. As in the non-AI context, the EEOC encourages employers to engage in a meaningful interactive dialogue with employees and candidates to determine what, if any, reasonable accommodation might be available.
The EEOC’s guidance also cautions employers against relying on AI data that is not pertinent to whether a candidate can perform the essential job functions with or without reasonable accommodation. For example, AI can make inferences based on analysis of “personality tests” or personal background information. The EEOC questions whether these type of inferences will screen out candidates based on disability-related traits even though, as a practical matter, they have limited or no impact on whether the candidate can actually perform the job.
The EEOC’s guidance also cautions against relying on a vendor’s claim of its AI tool being “bias-free.” The EEOC reports that, although algorithmic decision-making tool developers have taken steps to reduce bias from the variables considered and the data produced by AI tools, such efforts have largely focused on race and gender – not disability status or medical restrictions. Moreover, a vendor’s representation that its tool is “bias-free,” even if largely (but not completely) valid, may not insulate an employer from liability for discrimination that flowed from using the AI tool.
At bottom, using AI to streamline an employer’s hiring and employee evaluation processes can carry considerable benefits, including improving efficiency, accuracy, and cost savings in the employee hiring and management processes. However, employers should understand and be mindful of the potential pitfalls identified in the EEOC’s guidance when relying on AI analysis and decision-making tools for taking personnel actions.