In an era marked by ever-increasing technological advancements, New York City has taken a pioneering step in ensuring accountability and transparency in the use of automated decision systems, particularly within the realm of employment. The New York City Local Law 144-21, often referred to colloquially as the NYC Bias Audit Law, stands as a significant milestone in the ongoing pursuit of fairness and impartiality. Enacted in 2021, this comprehensive legislation came into effect on July 5, 2023, promising to usher in a new era in the use of automated employment decision tools (AEDTs) in the city.
At the core of the NYC Bias Audit Law is the definition of AEDTs. These are computational processes that derive from machine learning, statistical modeling, data analytics, or artificial intelligence. Their primary function is to generate simplified outputs, such as scores, classifications, or recommendations. These outputs are designed to assist or, in some cases, even replace discretionary human decision-making, thus heralding a significant shift in the landscape of talent management. However, it is important to note that not all automated tools are subject to this law. For instance, text translation or transcription tools are exempt from compliance. The crux of this legislation pertains to tools that substantially influence or entirely replace human judgment in hiring and promotion decisions.
Applicability of NYC Local Law 144
NYC Local Law 144 applies to employers and employment agencies that operate within New York City and utilize AEDTs for evaluating candidates for employment or employees for promotion. It is crucial to understand that these organizations bear the responsibility for compliance, even if they rely on third-party vendors to provide the tools.
The NYC Bias Audit: An Impatial Assessment
The cornerstone of Local Law 144 is the requirement for an annual bias audit. These audits are independent, impartial evaluations conducted by third-party entities to assess whether AEDTs result in disparate impact against individuals based on race, ethnicity, and gender. Disparate impact refers to the disproportionate negative outcomes for specific demographic groups.
The protected categories for assessment encompass male, female, and optionally other categories for gender, as well as Hispanic or Latino, White, Black or African American, Native Hawaiian or Pacific Islander, Asian, Native American or Alaska Native, and two or more races for race and ethnicity. The audit methodology varies based on whether the AEDT operates as a regression system with continuous scores or a classification system with binary outputs. In both cases, the audit is conducted to calculate impact ratios, which are used to compare the selection rates for different demographic categories, thus ensuring fairness in decision-making processes.
Data Requirements for Bias Audits
To conduct a comprehensive bias audit, data is indispensable. Ideally, this data should be historical, collected from real-life tool usage. In cases where historical data is insufficient or the tool has not been used, test data can be utilized, provided full disclosure is made. Organizations also have the option to rely on audits conducted by other employers or employment agencies if they have contributed to the aggregated data. While the law permits auditors to exclude groups with a small sample size (representing less than 2% of the audit data), scoring or selection rates must still be calculated for these groups, ensuring a rigorous evaluation process.
Transparency and Notification Requirements
Transparency and notification are at the core of the NYC Bias Audit Law. Employers and employment agencies are mandated to provide a summary of the bias audit results on their websites before deploying AEDTs. This summary must include details on data sources, the number of applicants in each category, exclusions due to missing demographic data, distribution dates, and impact ratios for standalone and intersectional groups. These summaries must be updated annually and kept online for six months after the tool is retired. In addition to transparency, organizations must notify candidates and employees at least ten working days before using an AEDT. The notice should contain information about the tool’s use, data sources, accommodation request procedures, and data retention policies.
Enforcement and Penalties
The enforcement of New York City’s Local Law 144 falls under the jurisdiction of the Department of Consumer and Worker Protection (DCWP). In cases of suspected violations, individuals should report them to the DCWP through contact at 331 or by visiting the DCWP website. Penalties for non-compliance range from $500 to $1500 per violation, signifying that fines can accumulate rapidly. Moreover, the damage to an organization’s reputation resulting from a failure to conduct an independent bias audit, publish a summary of results, or provide adequate notice can be even more detrimental.
Conclusion
The New York City Bias Audit Law, set to be enforced from July 5, 2023, represents a revolutionary step towards ensuring fairness and accountability in the use of automated employment decision tools. By mandating impartial bias audits, transparency, and notification requirements, the law aims to safeguard individuals from disparate impacts based on gender, race, and ethnicity, ultimately reshaping the landscape of talent management in the digital age. Compliance with this legislation is not merely a legal requirement; it is a moral imperative as organizations strive to maintain equity in their hiring and promotion processes.
Introduction