The NYC Department of Consumer and Worker Protection has released its “Final Rule” for NYC Local Law 144.
Just tell me what I need to know!
This post will dive into the changes made to the law; the detail may be of interest to compliance and legal professionals, but here is a quick summary for HR leaders and executives:
- The most impactful change in the new guidance is the new enforcement date of July 5, 2023 (moved back from April 15). We expect that since this is the final set of rules, that enforcement date won’t move again.
- There were not significant changes to the definition of “automated employment decision tool.” DCWP notes that the definition of “machine learning” was expanded (detail below); however, reading the actual text, the definition is not meaningfully changed. We don’t believe that tools that fell under the last guidance will be excluded by the final rules, and we don’t think that tools that were excluded previously will be in bounds now.
- The final rule solidifies the rules around use of test data for the compliance audit, particularly in cases where an employer would like to use a new tool. With these final rules, it’s clear that while employers are ultimately on the hook for compliance, working closely with vendors will be important
- Modifying the definition of “machine learning, statistical modeling, data analytics, or artificial intelligence” to expand its scope;
- This was a key point of contention in the previous public hearings, with industry looking to narrow the scope of this definition, and several advocates looking to expand it. Despite this note, reading the actual text of the final rule, there is very little change
- Adding a requirement that the bias audit indicate the number of individuals the AEDT assessed that are not included in the calculations because they fall within an unknown category, and requiring that number be included in the summary of results;
- This is a welcome clarification as most employers and vendors will not have complete gender and racial data for applicants. However, it introduces a substantial new ambiguity — how should cases where racial data is not kept for applicants at all be handled?
- Allowing an independent auditor to exclude a category that comprises less than 2% of the data being used for the bias audit from the calculations of impact ratio;
- This is another welcome change, as several categories rarely have large enough sample sizes for a meaningful audit. (For example, the Native Hawaiian population of NYC is ~0.6%. However, the new direction notes that when these groups are excluded from an audit, the auditor must still show the impact ratio and sample size for these groups — so this is more of a reflection on where the data should be shown
- Clarifying the examples of a bias audit;
- Clarifying when an employer or employment agency may rely on a bias audit conducted using the historical data of other employers or employment agencies;
- Providing examples of when an employer or employment agency may rely on a bias audit conducted with historical data, test data, or historical data from other employers and employment agencies;
- Clarifying that the number of applicants in a category and scoring rate of a category, if applicable, must be included in the summary of results.
- This is a welcome change; in what was likely just an oversight in the last guidance, the number of applicants by category was shown in some of the example charts but not others
Overall, the final rule is quite similar to the last guidance, making more “quality of life” clarifications for auditors rather than substantially enlarging or narrowing the scope of the law.
How Proceptual can help
Now that the final rule is released, employers will be scrambling to comply before the enforcement date, July 5th, 2023. After that, they may face fines up to $1500 per day, per tool.
Proceptual provides the technology and guidance you need to make compliance fast, easy, and straightforward, so you can cross it off your to-do list.
Contact our team for more information at email@example.com.