AI Compliance

NYC Releases Final Rules for Automated Employment Decision Systems (Effective July 5, 2023)

Today, the New York City Department of Consumer and Worker Protection (DCWP) released its Notice of Adoption of the Final Rules for Local Law 144, requiring employers and employment agencies to provide a bias audit of automated employment decision tools (AEDTs).

April 7, 2023
Author(s)
Ehrik Aldana
Contributor(s)
No items found.

Today, the New York City Department of Consumer and Worker Protection (DCWP) released its Notice of Adoption of the Final Rules for Local Law 144 requiring employers and employment agencies to provide a bias audit of automated employment decision tools (AEDTs).

The enforcement date for these rules has been delayed to July 5, 2023 (previously April 15, 2023).

This gives employers and employment agencies using AEDTs ninety days to get in compliance with the regulation – including some important new changes since the latest draft rules proposed in December 2022.

If you need a reminder about NYC Local Law 144, access Credo AI’s primer here.

Prepare for AEDT Bias Audit

With July 5, 2023 soon approaching, Credo AI has already helped both employers and vendors prepare for the requirements of Local Law 144. Learn more how Credo AI can support your organization by requesting a demo today!

As part of the final rules, the DCWP have:

  • further clarified what systems are covered by the law, 
  • added requirements to the bias audit and summary report, 
  • and clarified data requirements for historical and testing data.

Let’s look at these changes in detail:

Change 1: The Final Rules expand the scope of what qualifies as an AEDT.

Local Law 144 defines an AEDT as:

“any computational process, derived from machine learning, statistical modeling, data analytics, or artificial intelligence, that issues simplified output, including a score, classification, or recommendation, that is used to substantially assist or replace discretionary decision making for making employment decisions that impact natural persons.”

The previously proposed rules had clarified the definition of “machine learning, statistical modeling, data analytics, or artificial intelligence” to include mathematical or computational techniques:

“for which the inputs and parameters are refined through cross-validation or by using training and testing data.”

The Final Rules have removed this section, expanding the scope of systems impacted by the law.

Change 2: The Final Rules require bias audits to indicate data not assessed due to missing information.

The Final Rules now require the bias audit to:

“Indicate the number of individuals the AEDT assessed that are not included in the required calculations because they fall within an unknown category.”

Employers must now disclose the number of applicants or candidates assessed by the AEDT that are missing race/ethnicity or sex data, even if they are not included in the calculation of impact ratio.

Change 3: The Final Rules allow bias audits to exclude small sample sizes (<2%).

Calculations of impact ratio for the purpose of the bias audit may now exclude categories that represent less than 2% of the data being used for the bias audit. If a category is excluded, the bias audit must include justification for the exclusion, as well as the number of applicants and scoring rate or selection rate for the excluded category.

Change 4: The Final Rules clarify data requirements for historical and testing data.

The last iteration of the rules established that bias audits are required to use historical data of the AEDT. The Final Rules further require that employers or employment agencies can rely on a bias audit of an AEDT using historical data from other employers/employment agencies (conducted by a vendor for instance) only if:

  1. “such employer or employment agency provided historical data from its own use of the AEDT to the independent auditor conducting the bias audit or”
  1. “such employer or employment agency has never used the AEDT."

If insufficient historical data is available to conduct a statistically significant bias audit of the AEDT, employers may use test data. In this case, the bias audit must additionally explain why historical data was not used and how the test data used was generated and obtained.

Change 5: The Final Rules require the bias audit to include the number of applicants or candidates and scoring rates for each category.

As part of the summary of results for the bias audit, auditors must now include the number of applicants or candidates for each category (ethnicity/race and sex), as well as the scoring rates for each category, if applicable.

Prepare for AEDT Bias Audit

With July 5, 2023 soon approaching, Credo AI has already helped both employers and vendors prepare for the requirements of Local Law 144. Learn more how Credo AI can support your organization by requesting a demo today!

DISCLAIMER. The information we provide here is for informational purposes only and is not intended in any way to represent legal advice or a legal opinion that you can rely on. It is your sole responsibility to consult an attorney to resolve any legal issues related to this information.