It is estimated that almost a quarter of businesses are using automated or AI-driven tools in hiring and recruiting, among other hiring practices. The adoption of AI in hiring was particularly fueled by the pandemic, which saw employers forced to rapidly change their hiring practices to find alternatives to in-person interviews and assessments, meaning many turned to online solutions.
Not only are these tools more convenient and offer greater flexibility for both employers and candidates, but they can also reduce the time to fill open positions, improve candidate experience, and have the potential to increase diversity by reducing human biases. However, there have been several instances of algorithmic recruitment tools being misused or designed and developed without the appropriate safeguards, resulting in high-profile scandals and even lawsuits being filed.
As a result of the concerns raised by these risks, policymakers have demonstrated a clear commitment to regulating HR Tech tools, particularly in the US. New York is leading these efforts, having proposed three laws across the state and local levels to increase the safety and transparency of automated employment decision tools. At the local level, New York City Local Law 144 imposes bias audit requirements including the publication of a summary of results and notification requirements, while New York State Assembly Bill A07859 imposes notification requirements and Assembly Bill A00567 requires disparate impact analysis. In this blog post, we give an overview of the all-in-one approach of New York City and the two-part approach of New York State to regulating automated employment decision tools (AEDTs), starting by defining what AEDTs are in the context of these laws.
An AEDT is defined by New York City Local Law 144 as a computation process that uses machine learning, statistical modelling, data analytics, or artificial intelligence (AI) to issue a simplified output (e.g., a score, classification, or recommendation) that is used to substantially assist or replace discretionary decision making for employment decisions.
To substantially assist or replace discretionary decision-making means that the simplified output is solely relied on, is weighted more than any other criteria, or is used to overrule decisions from other factors. Further, machine learning, statistical modelling, data analytics and AI are defined as a group of mathematical techniques that generate a prediction or classification based on skills or aptitude, where a computer is used to identify the inputs, their weightings, and any other parameters for the model to improve the tool’s accuracy.
While A07859 uses the same definition for AEDT as Local Law 144 when setting out its obligations for notification requirements, A00567’s definition is notably different to A07859 and Local Law 144. Here, an AEDT is defined as a system used to filter employment candidates or prospective candidates to establish preferred candidates without the use of candidate-specific assessments by individual decision-makers. This includes personality tests, cognitive ability tests, resume scoring systems and any other systems that are underpinned by statistical theory or have parameters defined by techniques including inferential methodologies, linear regression, neural networks, decision trees, random forests, and other AI or machine learning algorithms.
Although the way that AEDTs are defined is vastly different between A00567 and A07859 and Local Law 144, the same kinds of tools fall within the scope of all three laws – tools used to evaluate candidates on pre-determined competencies or criteria that use a machine learning, AI, or data-driven approach to scoring. Depending on the technology they use, this can include game and image-based assessments, automated video interviews, and automated CV screening tools used to make employment decisions. However, it is important to note that New York City Local Law 144 applies to both screening candidates for employment and employees for promotion while the New York State laws apply only to the use of these tools for screening candidates for employment.
New York City Local Law 144, which was enforced from 5 July 2023, imposes two key requirements on employers and employment agencies. The first requirement is to commission an independent, impartial bias audit of the AEDT annually to assess the system for differential outputs based on sex/gender and race/ethnicity. This must be carried out by a third party with no employment relationship to the employer or employment agency or vendor that has commissioned the audit and that has no direct or indirect financial interest in the vendor or the employer or employment agency.
Employers and employment agencies are then required to make a summary of the results of this audit publicly available on the employment section of their website. This must detail the distribution date of the AEDT, the audit date, the data used for the audit, the number of individuals in an unknown category in the data, impact ratios based on the metrics put forward by the New York City Department of Consumer and Worker Protection (DCWP), and whether any groups were excluded from analysis due to small sample size.
The second key requirement of Local Law 144 is notification. Candidates and employees must be informed of the use of an automated employment decision tool to evaluate them. This must include the job qualifications and characteristics the tool uses to make its decision, the source of the data used for the evaluation, the AEDT data retention policy, and instructions on how to request an alternative selection procedure or accommodation under other relevant laws. This notification must be given at least 10 business days prior to the use of the tool.
Notice can be given to employees applying for a promotion. By providing notice in a written policy or procedure, providing notice in a job posting or via mail or e-mail. For candidates for employment, notice can be given on the employment section of the employer or the employment agency's website, in a job posting, or via mail or e-mail.
These two requirements result in separate violations of the law, where offenders can be fined both for failing to obtain an audit and publish the summary of results, and for not providing notice to candidates. Both requirements help to increase the transparency of the use of AEDTs by informing candidates about the use of the tool and allowing them to make informed decisions based on how the tool might perform for someone like them.
While New York City Local Law 144 imposes both bias audit and notification obligations, New York State has proposed 2 separate laws, one of which requires annual disparate impact analysis, which can be said to be equivalent to a bias audit, and another that imposes notification requirements.
Introduced on 9 January 2023, Assembly Bill A00567 seeks to require annual impartial disparate impact analysis of AEDTs to examine whether they result in differential outputs based on sex, race or ethnicity or other protected characteristics using the metric specified in the Uniform Guidelines on Employee Selection Procedures (i.e., the four-fifths rule of thumb). Like with New York City Local Law 144, this must be accompanied by a summary of results, which is to be provided to the employer or employment agency using the AEDT and made publicly available on their website. This must include a summary of the analysis and the distribution date of the tool. An additional requirement not seen in Local Law 144 is that the employer must also provide the Department of Labor with a summary of the most recent disparate impact analysis annually.
A07859 was introduced on 7 July 2023 and could be considered a follow-up to A00567, seeking to impose notification requirements for the use of AEDTs in New York. The Bill, therefore, acts like a second piece of the puzzle, with A00567 and A07859 together imposing the same requirements as Local Law 144.
Like with Local 144, employers and employment agencies must inform candidates that an AEDT will be used to evaluate them, the qualifications and characteristics that it will use to make its decisions, and information about the type of data collected, its source, and the data retention policy. These notices must be given at least ten business days before the use of the AEDT and should also provide instructions for candidates to request an alternative selection process or accommodation. The way in which such notices should be delivered has not yet been clarified, although this was only specified for Local Law 144 in the DCWP's rules to support enforcement.
As can be seen with the two New York State laws, New York City Local Law 144 is already having a ripple effect. This is not just confined to the state of the New York, though – New Jersey has also proposed a similar law with Assembly Bill 4909. Auditing and transparency requirements will soon become the norm for the use of HR Tech tools, with employers using AI-driven video interviews in Illinois and Maryland already facing notification requirements.
Acting early is the most effective way to get ahead of the incoming regulations and remain compliant. Schedule a call to find out how Holistic AI can help you with this.
Written by Airlie Hilliard, Senior Researcher at Holistic AI.
DISCLAIMER: This blog article is for informational purposes only. This blog article is not intended to, and does not, provide legal advice or a legal opinion. It is not a do-it-yourself guide to resolving legal issues or handling litigation. This blog article is not a substitute for experienced legal counsel and does not provide legal advice regarding any situation or employer.
Subscribe to our newsletter!
Join our mailing list to receive the latest news and updates.