Meta, a major U.S.-based technology company, was sued by 26 employees, who allege the use of AI-assisted performance metrics tools illegally chose workers with disabilities, workers on medical leave, and pregnant workers for reductions-in-force.
Plaintiffs allege Meta used AI tools, including productivity monitoring tools and AI token use tracking, to identify employees for job cuts. These tools included a large language model assistant, a communications and documents tracker, and a productivity score derived from keystroke data, screen activity, and email/browser history.
Plaintiffs allege the use of these AI tools, which plaintiffs allege were untested, placed certain employees at a disadvantage because they took job-protected leaves. The use of the tools allegedly resulted in decisions that violate federal and state laws prohibiting discrimination and retaliation against workers with disabilities, pregnant workers, and those on medical leave.
The employees' agreements with the company may require individual arbitrations to resolve any disputes. However, the plaintiffs seek a preliminary court order to halt the layoffs while arbitrations are conducted.
Source: https://bestmediainfo.com/mediainfo/mediainfo-digital/meta-lawsuit-alleges-ai-targeted-disabled-medically-absent-employees-for-layoffs-12164960
Commentary
Use of AI for hiring, discipline, promotions, or terminations can be helpful, but only if the information provided is examined to make sure it does not discriminatorily target protected classes.
An AI model that creates a productivity score based on keystroke data, screen time, emails, browser history, and document tracking will, by its own construction, mark workers on protected leave as lower-scoring. These unchecked results leave organizations at risk for lawsuits under the Family and Medical Leave Act, the Americans with Disabilities Act, the Pregnant Workers Fairness Act, and similar state laws.
To help avoid the risks of using AI tools without any oversight, consider these steps:
· Remove periods of protected leave from any performance ranking calculation.
· Identify every activity or productivity data point used in a model and remove those that are, in practice, measures of attendance or hours.
· Require legal review of every data point used in any reduction-in-force model before utilizing it.
· Document the business necessity for each data point before decisions are made.
· Utilize a review team for every reduction-in-force decision before it is made, never relying solely on the output of AI models or tools.
· Consult with local legal counsel before, during, and after major reduction-in-force decisions to ensure compliance with federal, state, and local laws.
The final takeaway is AI as a management tool is just that - a tool. Like any tool, it should be examined, tested, maintained, and used in nondiscriminatory manners.
