By: Dave Schmidt and Sara Hendrick
On September 19, 2022, the New York City (NYC) Department of Consumer and Worker Protection (DCWP) issued proposed clarification to Local Law 144 that will regulate the use of automated employment decision tools (AEDTs) with which hiring decisions are made or substantially assisted by algorithmically-driven mechanisms. As an example, AEDTs can be used to score a candidate’s performance on an assessment, score a candidate’s interview responses, or automate the task of screening resumes to assist recruiters in prioritizing candidates—all of which may help determine who moves forward in a hiring process. The law, which was passed in December of 2021, is set to take effect on January 1, 2023. DCWP clarification is currently under review and open for public commentary. A public hearing is scheduled for 11:00 am EDT on Monday, October 24, 2022, and DCWP is accepting written commentary up until the hearing. Specific instructions for providing written or oral commentary are in the clarification.
Local Law 144 introduced some of the most broad-spanning regulations on the use of AEDTs to date, including requirements for notices to candidates and transparency on the data being used and the characteristics being evaluated. The law also requires that all AEDT tools be independently audited for race and gender bias and that the results of those audits be published. However, there were many elements of the law, as originally written, that required clarification. The proposed clarification from DCWP provides additional guidance on some of these elements, though there are some places where refinements or further details may be needed, or where new questions were introduced that DCWP may need to address prior to finalizing this.
The proposed clarification, includes four sections:
- § 5-300. Definitions. DCWP expands the number of terms included in the definition section to thirteen, three of which are from the original law and ten of which are newly defined terms. The three original terms—AEDT, Bias Audit, and Employment Decision— are largely left intact. However, there is additional information provided to clarify the “substantially assists” phrase in the AEDT definition. The additional terms defined include “Candidate for Employment”, “Category”, “Distribution Date”, “Employment Agency”, “Independent Auditor”, “Impact Ratio”, “Screen”, “Selection Rate”, “Simplified Output”, and “Machine Learning, Statistical Modeling, Data Analytics, or Artificial Intelligence”. Many of these definitions are clear, but some will likely draw comments requesting DCWP to refine and/or provide additional information.
- § 5-301. Bias Audit. DCWP specifies the minimum analyses that must be conducted and includes two use cases showing what might be produced for bias audit analyses. Within these examples, there are some inconsistencies and some new concepts, which may lead commenters to request further clarification from DCWP.
- § 5-302. Published Results. The original law is relatively clear on the need to disclose AEDT use and the distribution date, job qualifications and characteristics being used or considered, data types and sources being used, and a data retention policy. DCWP provides further guidance on the type of information and level of detail they expect employers to publish related to the bias audit results. Some commenters may request further details from DCWP on the specifics of what must be published.
- § 5-303. Notice to Candidates and Employees. DCWP provides guidance regarding how and where employers can provide these notices and list multiple ways to meet the notice requirements. There is also a statement that “Nothing in this subchapter requires an employer or employment agency to provide an alternative selection process.”
DCWP is hosting a public hearing on October 24th to discuss the proposed rules. The ensuing revision from DCWP on the clarifications is much anticipated. As an independent auditor conducting bias audits related to this law, DCI will monitor these developments closely and post more information as it becomes available.