Derek Mobley applied for at least 80 jobs over several years but was rejected every time.
A Black worker over the age of 40, he says he has anxiety and depression, and believes there’s one main reason he wasn’t hired: employers used Workday software. The allegations, revealed in a class-action lawsuit in a California court, alleges Workday discriminated against him and others as part of the hiring process.
Mobley alleges in the 2023 suit filed with the U.S. District for Northern California that Workday provided screening tools that used subjective judgments in reviewing prospective job applicants. It also claims Workday allowed the preselection of applicants outside of protected categories. This case is making its way through the federal court system. Workday says the claims are false, and that its artificial intelligence recruiting tools don’t make hiring decisions, nor are they trained to use or identify protected characteristics.
AI and automated systems are now increasingly under scrutiny by states for potentially allowing discrimination in employment. California is the latest state to implement regulation that could give workers more power to push back on decisions made or aided by automated systems and AI. Other states that have passed similar regulations include New York, Illinois and Colorado, with the enforcement of Colorado’s legislation set to begin next year. As more states jump on the bandwagon, the outcome of Mobley’s case may have implications on how courts handle claims of workplace discrimination tied to AI and other technologies, attorneys said.
“The larger the employer is, the bigger the target on their back,” said Michael Manoukian, a partner at Lathrop GPM’s San Jose office. “I think we’ll see a flood of litigation here for companies using” these systems.
In California, a rule that took effect in October provides new guidance on how existing antidiscrimination laws apply to the use of AI and other automated systems.
Many states protect employees from harassment or discrimination in the workplace due to a person’s age, color, disability, race, religion, gender or sexual orientation, among other characteristics. But the update in California now also specifies that job applicants and employees are similarly protected from discrimination as a result of “automated-decision systems” meaning any system that uses AI, machine-learning, algorithms, statistics or other data processing techniques to make a decision or facilitate human decision-making.
“There’s a plethora of tools for employers to use for the employment life cycle — for hiring, onboarding, performance management, and on the back end separations and terminations,” Manoukian said. “This is the legal process trying to catch up with systems being used by businesses.”
California’s Civil Rights Department implemented the new rules because AI and other automated systems are increasingly used in employment “to facilitate a wide range of decisions” and they can “exacerbate existing biases,” it said in a statement announcing the regulation.
“They want to put some humanity into the employment process,” Alexander Reich, Chicago-based partner at law firm Saul Ewing, said about the state agency. “While applicant tracking systems and AI can be useful tools to increase productivity, they can also increase discrimination.”
Worker power
Regulations like those in California, New York, Illinois and Colorado may give workers more power to push back on employers if they suspect discriminatory practices as they relate to software systems used in the decision-making process, attorneys said.
Job candidates and employees in California who have a hunch that they were weeded out or unfairly treated due to an automated system can file a complaint with the California Civil Rights Department and also ask the employer to provide documentation, for example, Reich said. The state agency will then evaluate it to determine whether to launch an investigation.
There are still some gray areas that will likely play out in court should cases be filed, attorneys said. For example, there could be some debate as to what counts as an automated-decision system, what data needs to be preserved, where the line is drawn in terms of bias with AI, how an employee can prove their case, and how an employer can defend its processes. If a mid-manager uses AI to write an employee’s evaluation, which is negative and has adverse effects for the employee, an employee may have grounds to argue that this was an unlawful, discriminatory practice, Reich said.
Companies carry the burden
Employees will be burdened with providing enough evidence to spur an investigation, but if they do, it can create a big mess for an unprepared employer.
“If the employee believes there’s an unlawful practice, the employer will have to be the one to defend themselves,” Manoukian said.
Companies using AI or automated-decision systems in any part of their employment practices will have to be proactive in keeping records, vetting third-parties, auditing their tools and understanding how proprietary or licensed software works.
“The biggest thing is that employers just need to know what they’re getting themselves into,” Reich said about applying software to workplace decisions. “You have to know what your system is doing and do proactive compliance.”
This means employers should run audits of all the software they use, ensure vendors have antibias protocols, and that liabilities and indemnification obligations related to the new laws are spelled out in all software contracts, attorneys said. They should also test software to see if it produces any concerning results before implementing it at the workplace.
“The legislation acknowledges the reality that [AI] is somewhat of a black box at this point,” said Scott Jang, San Francisco-based principal of the law firm Jackson Lewis. “They’re putting the pressure and liability on the employer to force light into the black box.”
Employers also need to keep clear documentation of all personnel-related decisions for four years. And if they haven’t already, they should create and clearly communicate AI policies, attorneys said.
While AI may be the biggest concern, employers shouldn’t overlook other automated software like applicant tracking systems, which may also lead to biased decisions. For example, an employer might use a filter to cull a large pool of candidates to more experienced workers but accidentally cut protected classes like pregnant women, Reich said.
“I would be wary of using any type of filters where you’re giving control over to software,” he said. “Take steps to make sure that doing it that way doesn’t have any unintended consequences of weeding out certain people.”
As part of his lawsuit, Mobley hopes to recover back pay and other monetary damages he claims he suffered as a result of the alleged discrimination, and to stop Workday from perpetuating these practices. Workday AI systems and screening tools rely on algorithms and inputs from humans who often have built-in motivations to discriminate, he alleges.
Tools like these “allow employers to consciously or unconsciously discriminate against individuals on the basis of race and/or disability in making hiring decisions,” the lawsuit reads. “There is no business necessity justifying the disparate impact on individuals.”
The post New rule targets AI discrimination. Here’s what workers need to know. appeared first on Washington Post.
