Court Partially Upholds Claims Against Workday's AI Screening Tools

In a recent decision from the Northern District of California, the court had a mixed response to Workday, Inc.'s attempt to dismiss a lawsuit alleging discrimination in their AI-based hiring tools. The case, Derek Mobley v. Workday, Inc., No. 23-cv-00770-RFL (N.D. Cal. July 12, 2024), raises important questions about the role of AI in employment practices.

Derek Mobley, an African American man over forty with anxiety and depression, claims that Workday’s AI-driven applicant screening unfairly rejected him from over 100 job applications. Despite his extensive qualifications, Mobley believes that the AI tools used by Workday systematically discriminated against him based on his race, age, and disability.

The court’s ruling, issued on July 12, 2024, had several notable points:

Importantly, the court acknowledged that Workday might be liable as an agent of the employers using its software under a disparate impact theory. This means Workday could be responsible for discriminatory practices if its AI tools unfairly screen out a single applicant for many different positions to which the applicant was qualified on paper to perform and the employers made use of Workday’s AI software.

However, the court dismissed claims that Workday intentionally discriminated in the creation of their AI algorithms. Additionally, the court held that Workday is properly classified as an agent of the employer under the federal anti-discrimination law as opposed to an employment agency. The judge clarified that Workday doesn’t fit this definition because it doesn't recruit employees but rather processes and screens applications for employers.

Disparate Impact Claims Allowed to Proceed

The court allowed Mobley’s claims of disparate impact to proceed. Disparate impact claims focus on the consequences of a practice rather than the intent behind it. Mobley argued that Workday’s AI tools disproportionately disqualified applicants based on race, age, and disability, regardless of the employer's intentions.

The court found that Mobley presented enough evidence to suggest that Workday’s algorithms might indeed have a disparate impact on these protected groups. This included the high number of rejections and the nature of the automated screening process, which relied on potentially biased training data and personality assessments. The court determined that these factors were sufficient to raise a plausible claim that Workday’s tools resulted in adverse effects on certain applicants.

The Court's Analogy

The court made an interesting analogy to explain the significance of Mobley's disparate impact claims. It compared Mobley applying to over 100 jobs with one employer using Workday’s platform to a situation where 100 plaintiffs apply to a single firm. This analogy underscored the systemic nature of the alleged discrimination. Just as a pattern of discrimination would be evident if 100 similarly qualified applicants were rejected by one employer, Mobley's numerous rejections across different employers using the same AI tool highlighted the potential for a widespread issue within Workday’s screening process.

Intentional Discrimination Claims Dismissed

In contrast, Mobley's claims of intentional discrimination under Title VII, the ADEA, and Section 1981 were dismissed. The court found that while Mobley showed he was qualified and that his protected traits were likely known, he didn’t provide enough evidence to prove that Workday intended to discriminate.

To state a claim for intentional discrimination, a plaintiff must show that the defendant acted with a discriminatory motive. The court concluded that Mobley did not offer sufficient facts to demonstrate that Workday’s algorithms were designed or used with the intent to discriminate based on race, age, or disability. The allegations pointed to the existence of biases but failed to show that these biases were intentionally embedded or used by Workday to disadvantage certain groups.

Implications

This ruling highlights the evolving legal landscape around AI and hiring practices. It underscores the need for employers and AI vendors to be able to audit how their software functions and is being used by customers. In order to win the case, especially if it gets to a jury, Workday will have to explain why its software made the specific decisions that it did. Members of the jury will all be human beings worried about the opaque decisions made by computers that have real world effects on peoples lives.

This case also displays the difficulty of pleading an intentional discrimination case in a claim revolving around the use of AI decision making. Potential plaintiff’s simply do not have access to the facts and data they would need to show that there was intentional discrimination in the design of these products absent some unique circumstances.

This will be an interesting case to follow. Mobley now has 21 days to amend his complaint. The case will move forward with the remaining claims, with a case management conference set for August 7, 2024.

Previous
Previous

A Closer Look at American Alliance for Equal Rights v. Founders First

Next
Next

Free Speech and Public Accommodations Law Square Off in Court