SHRM, CalSHRM Concerned About California’s Proposed Automated-Decision System Rule

California’s proposed regulations on automated-decision systems could chill the adoption of artificial intelligence by HR professionals in the state, SHRM and the California State Council of SHRM (CalSHRM) cautioned in a Nov. 18 letter to David Garcia, chair of the California Civil Rights Department Civil Rights Council (CCRC). Out-of-state businesses that operate in California would also be hindered, preventing them from fully realizing AI’s potential to solve urgent challenges, the letter added.

AI’s Benefits Ignored

“CCRC’s Initial Statement of Reasons [for its proposed rule] discusses the risks of algorithmic discrimination but does not contain any statements about the potential benefits of using technology to support HR processes,” wrote Emily M. Dickens, J.D., SHRM chief of staff and head of government affairs, and Eric De Wames, government affairs director for CalSHRM. Discrimination based on categories in the California Fair Employment and Housing Act (FEHA) is already unlawful, the letter noted.

AI’s benefits include reaching out to passive candidates with diverse backgrounds who might otherwise be overlooked, as well as scanning applications for experience in addition to educational requirements, the letter pointed out. 

As AI becomes increasingly integral to business operations across industries, California workers will need the opportunity to learn, adapt, and use AI tools to remain competitive in the global job market, SHRM and CalSHRM said. By combining technology with human intelligence, HR professionals are achieving positive outcomes, the letter noted.

Flawed Cost Impact Assessment

Set against AI’s many benefits, CCRC’s assessment of the proposed regulations’ costs is flawed, SHRM and CalSHRM wrote. The proposed rule said it would result in no additional costs or savings beyond those imposed by existing laws.

“CCRC’s cost assessment ignores the direct costs imposed on California businesses deploying AI, and on their vendors, through the proposed expanded data-retention requirements,” the letter said. “CCRC’s cost assessment also ignores the indirect costs that California businesses seeking to use AI will face, and the impact to California businesses that might decide that the legal uncertainty and potential increased litigation risk mean they will defer or halt their efforts to take advantage of the benefits of AI.”

Burdensome Recordkeeping Proposal

Without additional clarification, the proposed regulations appear to issue a mandate to employers to save everything that’s considered to be automated-decision system data for four years from the date of the making of the record or the date of the staff action involved, whichever is later, SHRM and CalSHRM said. 

They noted that CCRC’s underlying definition of an automated-decision system “is breathtaking in its breadth, encompassing any ‘computational process’ that ‘otherwise makes a decision or facilitates human decision-making that impacts applicants or employees.’ ”

Even basic scheduling software that merely fills shifts or sets up job interviews would likely qualify as an automated-decision system under the proposed regulations, the letter said.

“The proposed regulations essentially dictate to businesses operating in California that nearly every application of technology for matters relating to employment will be subject to expansive recordkeeping obligations,” SHRM and CalSHRM wrote. This mandate would impose “significant burdens on California businesses that are disconnected from the scope of the algorithmic harm that CCRC says it is trying to address.”

Tension with Privacy Rights 

The proposed rule would also require employers to create and maintain extensive databases of personal information. 

“This mandate could create significant tension with employees’ privacy expectations and existing privacy rights under California law,” SHRM and CalSHRM cautioned. They also pointed out that employers complying with recordkeeping requirements would incur additional direct costs from complying with multiple notice and disclosure obligations under California’s privacy laws.

Further Concerns

The groups also expressed concern that the proposed rule would create uncertainty about whether the Uniform Guidelines on Employee Selection Procedures’ prohibition on disparate impact and disparate treatment in selection criteria would apply only when an automated-decision system is used as a selection criterion, or to any use of an automated-decision system. 

“The Uniform Guidelines provide established statistical methods for measuring adverse impact when an automated-decision system is clearly a selection procedure,” the letter stated. “[W]hat standards apply to automated-decision systems that are not selection procedures, but might still be considered to ‘facilitate’ employment decisions?”

CCRC’s proposed regulations are unclear, which could further chill the use of AI applications in California, SHRM and CalSHRM noted.

Lastly, CCRC’s proposed expansion of the private right of action under FEHA creates additional legal uncertainty, the letter said.

“SHRM and CalSHRM believe that while safeguarding individual rights is essential, it is equally important to avoid a legal environment in which employers and HR professionals feel excessively cautious or hindered in making daily operational decisions,” the groups wrote. “The expansion of private rights of action should be balanced with the need to maintain an environment conducive to business in which employers can operate effectively without the constant threat of litigation.”