listen to this post

On April 6, 2023, the New York City Department of Consumer and Workforce Protection (“DCWP”) enacted its final regulation (the “Final Rules”) in connection with the New York City Automated Employment Decision Tools Act (“AEDTL”). In connection with the Final Regulations, the DCWP also notified employers that it would further delay the application of the AEDTL from April 15, 2023 to July 5, 2023. The Final Regulations, among other things, expand the definition of “machine learning , statistical modeling , data analysis or artificial intelligence” as used in the AEDTL and clarify details about bias audits required by the AEDTL.

As we previously reported, the AEDTL regulations have been in a state of flux for months. The Final Regulations are the DCWP’s third attempt to formulate regulations for the AEDTL and have received an unusually high volume of public comment, requiring several highly attended public hearings. As a result, the application of the AEDTL was postponed twice; first from January 1, 2023 to April 15, 2023, and now through July 5, 2023. However, because Final Regulations have been issued up to this point, it is unlikely that the effective date of the AEDTL will change any further.

Overview of the AEDTL

Once executed, the AEDTL will restrict employers’ ability to use “automated employment decision tools” in hiring and promotion decisions in New York City. The AEDTL defines “automated employment decision tool” as “any computational process, derived from machine learning, statistical modeling, data analysis, or artificial intelligence, that emits simplified output, including a score, rating, or recommendation, used to substantially assist or replace discretionary decision-making with employment decisions that affect individuals”. The phrase “substantially assist or replace discretionary decision-making” means: (i) relying only on simplified output (score, tag, rating, rating, etc.), with no other factors considered; (ii) use a simplified output as one of a set of criteria where the simplified output is more heavily weighted than any other criterion in the set; or (iii) use a simplified output to overrule conclusions derived from other factors, including human decision making.

Employers may not use automated employment decision tools unless: (i) the tool has been subject to a bias audit carried out in the previous year in accordance with AEDTL requirements; and (ii) the employer has posted a summary of the results of the tool’s most recent bias audit, as well as the distribution date of the tool to which such audit applies, on its publicly available website. The audit for bias must be performed by an “independent auditor”, who cannot have been involved in the use, development or distribution of the tool, cannot be employed by an employer seeking to use the tool, or a vendor who developed or distributed it. , and must not have a material direct or indirect financial interest in the employer seeking to use the tool or the vendor that distributed it. Employers must use historical data (i.e., data collected during the use of the tool by the employer) to conduct the audit. However, if insufficient historical data is available to conduct a statistically significant audit, employers may use non-historical test data, provided the employer explains why historical data was not used and how the test data was used. were generated and obtained.

Under the AEDTL, employers who use automated job decision tools must also disclose the following information to applicants at least ten business days before the tool is used: (i) the fact that an automated job decision tool will be used in connection with the evaluation or evaluation of any applicant residing in New York City; (ii) the qualifications and job characteristics that the automated employment decision tool will use to evaluate the candidate; and (iii) instructions on how an individual can apply for an alternative selection process or reasonable accommodation, if available. The AEDTL does not require employers to provide an alternate selection process, although employers are required to provide reasonable accommodation if required by the Americans with Disabilities Act and analogous state and local laws.

Finally, the AEDTL requires employers to undertake the following additional disclosure steps: (i) provide information in the employment section of their website in a clear and visible manner about their automated employment decision tool data retention policy, the type of data collected for the tool, and the source of the data; (ii) post instructions in the employment section of its website in a clear and conspicuous manner on how to make a written request for such information and, if a written request is received, provide such information within 30 days; and (iii) if such a request is denied, explain why disclosing such information would violate applicable law or interfere with a police investigation.

Employers who violate the AEDTL may be subject to civil fines ranging from $500 to $1,500 per day that the employer fails to comply with the law. The AEDTL does not expressly permit or prohibit a private right of action, but states that it should not be construed as “limiting any right of any applicant or employee for an employment decision to bring a civil action in any court of competent jurisdiction”.

Impact of the Final Regulation

The Final Regulation imposed several discrete changes to the requirements of the AEDTL and clarified the obligations of employers. More specifically, the Final Regulation:

  • Expand the definition of “machine learning, statistical modeling, data analysis, or artificial intelligence” to mean “a group of computer-based mathematical techniques: (i) that generate a prediction, i.e., an expected result for an observation, such as an assessment of a candidate’s suitability or likelihood of success, or that generate a rating, ie, an assignment of an observation to a group, such as categorizations based on skill sets or aptitudes; and (ii) for which a computer identifies, at least in part, the inputs, the relative importance assigned to those inputs and, if applicable, other parameters for the models in order to improve prediction or classification accuracy”;
  • Requiring bias audits to indicate the number of individuals the tool assessed that are not included in the calculations because they fall into an unknown category, and requiring that number to be included in the summary of results;
  • Allow auditors to exclude categories comprising less than 2% of the data used for a bias audit from impact rate calculations;
  • Clarify examples of a biased audit;
  • Clarify when an employer can rely on a bias audit conducted using the historical data of other employers;
  • Provide examples of when an employer may rely on a biased audit conducted with historical data, test data or historical data from other employers; It is
  • Clarify that the number of applicants in a category and the category score, if applicable, must be included in the summary of results.

Next steps

The AEDTL and the Final Regulation are complex and this blog only provides an overview. Fortunately, employers who use or are considering using automated job decision tools now have an additional three months to comply. During this period, employers must: (i) identify the automated employment decision tools they currently use that may be subject to the AEDTL; (ii) begin collecting historical data or, if sufficient historical data is not available, identify appropriate test data; (iii) identify an appropriate independent auditor and obtain a biased audit; and (iv) plan for compliance with the AEDTL reporting requirements. Given the many nuances of the AEDTL and Final Regulations and the potentially significant penalties at stake, employers are strongly encouraged to coordinate with attorneys in their compliance efforts.

We will continue to monitor any new developments and provide updates as they become available.