AI Screening Tools Under Scrutiny: Federal Court Preliminarily Certifies ADEA Collective Action
In an important decision for employers who use AI software in making hiring decisions, a California federal district court granted preliminary collective certification under the Age Discrimination in Employment Act (ADEA) to a group of job applicants alleging they were rejected for discriminatory reasons by an AI-based hiring recommendation tool provided by Workday, Inc.[1] The court ruled that claims based on disparate impact from algorithmic hiring systems are suitable for collective treatment at least at this early stage—even when individual applicants applied for different roles at different companies and each company may have utilized the AI-based features and output in their hiring process differently. Now, the parties will engage in class discovery and the plaintiffs may serve notice of the lawsuit on similarly situated individuals, providing them an opportunity to opt in to the proceedings and have their claims heard on a collective basis.
The court's decision marks what we believe to be the first in the country to preliminarily certify a collective action based on alleged age discrimination stemming from the use of artificial intelligence in hiring. As detailed below, the decision may significantly impact employers and vendors that rely on AI-driven recruiting technologies and certainly raises the specter of greater compliance risk and scrutiny. Accordingly, we encourage companies that develop and deploy AI tools at any phase of the hiring process to review their practices regarding how AI is being used and to minimize their potential liability in this area.
About the AI-Based Tools and the Federal Court Case, Mobley v. Workday, Inc.
Workday, Inc. offers cloud-based human resources and applicant screening tools to employers across the country. Its recruiting products offer AI- and machine-learning-based tools that may evaluate applicants by scoring, ranking, or recommending candidates based on resume data and employer preferences. These tools offer employers the possibility of streamlining the talent acquisition process.
However, Derek Mobley and four other plaintiffs allege that Workday's AI tools reinforce existing employer bias and, as a result, they and other job applicants suffered from race, age, and disability discrimination. Specifically, the plaintiffs claim that the AI tools rely on biased training data or mimic (and exacerbate) prior discriminatory hiring practices that disfavor certain classes of job candidates. The plaintiffs also allege that, in many cases, applicants cannot advance in the hiring process unless they satisfy Workday's screening algorithms.
In line with his allegations, Mobley sought to preliminarily certify a collective action on behalf of "[a]ll individuals aged 40 and over" who "applied for job opportunities using Workday, Inc.'s job application platform and were denied employment recommendations." Preliminary collective certification is the first step of a two-step process to certify a collective action and paves the way for the parties to conduct class discovery. Following discovery, employers typically move to decertify the collective, at which point the court will take a more exacting look at the allegations in the complaint and record evidence. As a result, the bar to obtain preliminary collective certification at this first stage is low.
To meet this (typically low) bar, Mobley was required to argue that he was similarly situated to the potential opt-in plaintiffs. The court's decision centered on the "similarly situated" analysis and found enough similarity there to allow notice to the entire collective—which could consist of millions of putative plaintiffs—and to proceed to discovery.
The Court's Holdings
The decision, which only addresses Mobley's claims regarding age discrimination, centers on whether Mobley sufficiently alleged that he is similarly situated to opt-in plaintiffs in the proposed collective. The court first rejected Workday's request for Mobley to face a heightened evidentiary burden because the parties had already conducted some discovery, since doing so would create inconsistent results in future cases. Instead, the court adhered to 9th Circuit precedent by evaluating the issue under the "substantial allegations" of similarity in the complaint.
Second, the court determined that the proposed collective was similarly situated to Mobley because the plausible allegations in the complaint described a uniform policy applicable to all job applicants that generated a disparate impact. More specifically, the court held that Mobley plausibly alleged the mechanisms in Workday's AI recommendation system may cause disparate impact amongst job candidates over the age of 40, and all such applicants' "claims rise and fall together."
Addressing Workday's Challenges
Workday asserted numerous challenges to preliminary collective certification, including that: (i) Workday does not offer employment recommendations, so there should be no collective members; (ii) contrary to Mobley's contentions, the underlying policy at issue is not uniform; and (iii) "natural variation" amongst proposed collective members' qualifications, the number of jobs they applied to, and their individual rejection rates mean no collective could be similarly situated.
On the first point, the court distilled Workday's challenge as arguing that Workday itself does not recommend candidates for hire, since its AI recommendation system cannot auto-reject candidates without participation by the employer. The court found Mobley's allegation that Workday participated in a practice resulting in disparate impact was sufficient to warrant preliminary collective certification. As examples, the court highlighted two specific Workday AI tools that grade and recommend applicants.
On the second point, Workday argued that no uniform policy applies to applications submitted through Workday, since employers can decide whether or not to use Workday's AI features. However, the court found the proposed collective already reflected the limitation that the employer at issue must have used Workday AI features in evaluating the individual's candidacy. The court also rejected Workday's argument that individual AI features could vary in their impact across different employers, finding unit-level differences do not defeat Mobley's claims where a unified policy creates a net disparate impact provable by statistical evidence.
On the third point, the court held that individual variation in application experiences amongst the collective does not defeat preliminary collective certification. Specifically, the court ruled that Mobley was not required to prove that each member of the proposed collective is identically situated, but rather to identify similarities material to the resolution of the case. However, the court highlighted that individual differences amongst collective members may be relevant at the merits stage of the case.
Main Takeaways
Although preliminary certification is almost routinely granted, the decision is still notable because it reinforces the court's previous holding denying Workday's motion to dismiss and finding that vendors of AI tools may be directly liable under federal anti-discrimination laws if their tools function as gatekeepers in hiring decisions. It is also clear that courts will not necessarily apply more rigorous standards to preliminary certification of a putative collective challenging AI screening tools where uniform algorithms are used across multiple, broad applicant pools. Notably, unlike disparate treatment cases, disparate impact cases rely on statistics and do not require a showing of employer intent. This underscores the need for companies that use AI in the hiring process to:
(1) audit their automated decision-making tools,
(2) track how those tools impact the hiring process, and
(3) ensure human oversight where appropriate.
We will continue to track this and other cases in the algorithmic hiring space. There is an evolving legal landscape created by AI and ML tools for talent acquisition, as this case demonstrates. Stay tuned for more!