Meta’s AI-driven layoffs may have been biased against workers on leave, according to a lawsuit by 26 former Meta employees. These ex-employees claim that Meta used a constellation of internal AI tools to rank performance data and determine who would be laid off. However, the ranking system failed to exclude those on parental or medical leave, resulting in these protected workers being disproportionately selected for layoffs. The AI scoring not only overlooked their leaves but actually penalized them for taking time off, suggesting that the AI might have been more impressed by productivity metrics than actual job performance.

One major assumption here is that the AI tools accurately captured and weighted all relevant performance data. But did Meta’s AI consider nuances like project contributions, leadership roles, or cross-functional impact? It’s possible the AI focused too heavily on quantitative metrics—like daily check-ins or task completions—and missed qualitative achievements such as innovative solutions or mentorship. Thus, employees who excelled in less measurable ways might have been unfairly lumped into the laid-off group.

Another claim is that the AI failed to account for protected leaves, implying it treated all leave periods equally regardless of type or duration. However, some AI models can differentiate between types of leaves—vacation vs. sick vs. parental leave—and adjust rankings accordingly. Did Meta’s AI simply apply a blanket penalty to any leave taker, or were there deeper inconsistencies? Perhaps employees who took longer leaves or those with higher leave-to-work ratios were singled out more harshly.

Lastly, the lawsuit suggests that the AI penalized workers for exercising their rights—taking leave should be a boon, not a strike against one’s job security. Yet, if Meta had tied leave duration directly to performance scores without a buffer, employees who took longer leaves might have seen their rankings dip more sharply. For instance, someone on an extended parental leave could have been ranked lower than a colleague who worked the same number of days but didn’t take any time off. This highlights the need for AI models to incorporate flexible adjustments for protected leaves, ensuring that workers aren’t penalized for exercising their rights.

In summary, while Meta’s AI-driven layoffs may indeed have been biased against those on leave, the root cause could lie in the AI’s inability to fully integrate and weight all performance dimensions. By overlooking qualitative contributions and applying a uniform penalty for protected leaves, the AI might have unfairly singled out workers who took time off. To truly resolve this, Meta should refine its AI tools to capture both quantitative and qualitative metrics, adjust leave penalties dynamically, and ensure that those on parental or medical leave are not disproportionately laid off based on incomplete performance data.