On April 6, 2023, the New York City Department of Consumer and Worker Protection (“DCWP”) adopted final rules related to new legislation aimed at regulating the use of artificial intelligence in hiring and promotions. bottom.[1] The law requires employers or staffing agencies to conduct bias audits to ensure that artificial intelligence (“AI”) tools infect the hiring and selection process with unlawful biases and unfairly differ from certain groups of applicants. It requires us to evaluate whether we are likely to make decisions that affect us. employee. The AI used in this process is called an Automated Employment Decision Tool, or ‘AEDT’, hence the name ‘AEDT Act’. DCWP will begin enforcing his AEDT Act on July 5, 2023.[2] For more information on the basics of AEDT law, check out our previous warning here.[3]
The final rule was created after lengthy discussions regarding the regulation of AEDTs in recruitment. The New York City Council originally enacted a bill in November 2021 with an effective date of January 1, 2023.[4] However, since its enactment, the DCWP has received numerous comments from employers, employment agencies, law firms, AEDT developers, and advocacy groups calling for clarification and revision of the law. In response to a series of public comments, DCWP has clarified what constitutes her AEDT, biased audits, and independent auditors.
1. What are AEDTs?
Generally speaking, AEDT refers to an automated screening or selection process previously performed by humans. To clear up confusion and add concreteness to the definition, DCWP explained that AEDT is any computational process “derived from machine learning, statistical modeling, data analysis, or artificial intelligence.”[5] Provides employers with scores, classifications, or recommendations that are used to significantly assist or replace the human decision-making process.[6] DCWP further states that substantially assisting or replacing a natural person’s decision-making process means: 2) Use simplified output as a criteria list consideration, but give more weight to output than other criteria in the set. or 3) use the output to reverse the conclusions of human decision making.[7]
2. What is required for a bias audit under the AEDT Act?
of.Auditors must be independent
Through the final rule, the DCWP clarified that bias audits must be conducted by truly independent auditors. The original proposal was open to interpretation as to who could properly function as an auditor. For example, it was unclear whether the company that developed the AEDT could have another business unit conduct an “independent” audit of the tool. The DCWP made it clear in its final rule that a biased audit must be an unbiased evaluation by an independent auditor. It also defines an independent auditor as an individual or group capable of making an “objective and impartial judgment on all matters within the scope of the biased audit.”[8] To maintain impartiality, auditors are not permitted to be involved in the use, development, or distribution of AEDTs. Have an employment relationship with an employer who intends to use the AEDT for the duration of the bias audit. Or have a financial interest in her use of the AEDT for the duration of the audit.[9]
b. Audits should calculate how the AEDT affects decisions related to people in each EEO protection category.
In addition to providing guidance on auditors, the DCWP states that when an AEDT selects candidates for employment or promotion, or sorts candidates into groups, biased audits should at least: I specified.
- For each EEO protection category, we calculate the selection rate (that is, the rate at which individuals of the race, gender, and ethnicity categories are selected to progress through the recruitment process).
- For each EEO-protected category, calculate the influence rate (that is, the category’s selection rate divided by the selection rate (the rate at which individuals in the category are selected to progress in the hiring process)).
- In the calculations in paragraphs 1) and 2), please calculate separately the effect of AEDT on:
- “Gender Category (for example, Influence rate of selection of male and female candidates),”
- Racial/ethnicity categories (e.g., influence rate of selection of Latino vs. Black (not classified as Latino) candidates), and
- Cross-categories of gender, ethnicity, and race (for example, selection influence rates for Latino male candidates and non-Latino black female candidates).
- If the AEDT classifies candidates for employment or promotion into designated groups, ensure that the calculation is performed for each classification group.and
- Indicates the number of individuals evaluated by AEDT that are not included in the required calculation above because they fall into an unknown category.[10]
In particular, the AEDT method does not address whether non-binary individuals should be included in gender category influence rates, and if so, how to properly calculate influence rates for these individuals. For many years, New York City made it illegal to discriminate against individuals based on a person’s gender identity, self-image, appearance, behavior, or expression. There seems to be It is unclear if DCWP will clarify this open issue.
3. What kind of notice is required for candidates and employees?
As previously reported, candidates for job postings or employees seeking promotion should receive notice from their employer that the AEDT will be used as a recruitment tool. The final rule mandates that employers update their websites at least 10 business days prior to using the AEDT, provide notice of the use of the AEDT in job postings, and require candidates to use the AEDT in U.S. mail or email. Provide logistics for providing notice, including notifying used.[11]
The final rule also clarifies that if an AEDT is used for hiring or promotion, the employer or staffing agency must conduct a bias audit within one year of using the tool.[12] For example, if an employer wants to use an AEDT to screen resumes and conduct interviews for vacancies, the employer must ensure that a bias audit was conducted at least one year before the employer’s planned use of the AEDT. must be verified. It is trivial that the AEDT is only used for initial résumé screening and not for making final employment decisions. Regulation still requires employers to enforce audit bias. Additionally, before an employer can use her AEDT, the employer must publish the date and results of the most recent bias audit in the employment section of the website.[13]
* * *
Therefore, with such influential regulation on the horizon, it is important that employers begin investigating and adjusting their internal and external HR processes and practices. Below are some of the key issues employers should address before the law takes effect.
- Contact your organization’s Human Resources representative to discuss and review the details of the hiring process from start to finish to determine if you would like to use the AEDT.
- If you currently use an AEDT for any aspect of employment or promotion, contact your AEDT vendor to assess whether the AEDT in use is aligned with the requirements of the AEDT Act. please give me.
- If your organization uses an AEDT, hire an independent auditor to conduct a bias audit as soon as possible. This should be done well before July 5th. By conducting this audit sooner or later, he will give the organization ample time to adjust the AEDT should it show any indication. of bias.
- For large organizations, the employment section of the company’s website must be updated by July 5 to inform candidates that they will be using the AEDT for recruitment and promotion. Job postings are posted regularly, so we recommend that you do not wait until the job posting is published.
- To ensure compliance, hire an attorney to review the AEDT and overall employment policies.
The AEDT law marks a change in New York City that other states are likely to replicate. With the overwhelming focus on artificial intelligence, it’s imperative that employers and HR professionals keep abreast of these changes.
[1] look New York City Department of Consumer and Worker Protection, Notice of Adoption of Final Rule, Available at https://rules.cityofnewyork.us/wp-content/uploads/2023/04/DCWP-NOA-for-Use-of-Automated -employment-decision-making-tools-2.pdf.
[2] look New York City Department of Consumer and Worker Protection, New Laws and Rules, at https://www.nyc.gov/site/dca/about/new-laws-rules.page#:~:text=December%202022 %20Update available, page %20for%20updates.
[3] https://www.strock.com/news-and-insights/are-you-ready-for-nycs-anti-bias-ai-law
[4] look New York City Department of Consumer and Worker Protection, New Laws and Rules, at https://www.nyc.gov/site/dca/about/new-laws-rules.page#:~:text=December%202022 %20Update available, page %20for%20updates.
[5] Furthermore, the final method is that machine learning, statistical modeling, data analysis, or artificial intelligence combine groups of computer-based mathematical processes to produce expected results or predictions, or to produce assignments of observations to groups. It is clear what is meant (i.e., classification), “such as a classification based on skill set or If so, identify other parameters in the model to improve classification accuracy.
[6] look New York City Department of Consumer and Worker Protection, New York City Regulations Title 6 Chapter 5 Updates, Subchapter T, § 5-300, available at https://rules.cityofnewyork.us/wp-content/uploads/ 2023/04/DCWP-NOA-for-Use-of-Automated-Employment-Decisionmaking-Tools-2.pdf (“Updated AEDT Act”).
[7] Updated AEDT law with § 5-300.
[8] Updated AEDT law with § 5-300.
[9] Updated AEDT law with § 5-300.
[10] Updated AEDT Act with § 5-301.
[11] Updated AEDT Act with § 5-304(b).
[12] Updated AEDT Act with § 5-301(a).
[13] Updated AEDT Act with § 5-303(a).
