New York’s Local Law 144 of 2021 will finally go into effect on July 5, 2023, after several delays. As we previously discussed, the law requires employers to provide candidates for employment and promotion with notice about the use of an AI system, offer them an opt out, and audit any such systems for bias. The law is intended to benefit job applicants and may provide useful guidance for employers who wish to use AI to help eliminate workplace bias.
Although the law was effective as of January 1, 2023, the New York City Department of Consumer and Worker Protection, just recently published their final rules on April 6, 2023. This has left companies with a reduced timeframe to comply with the new rules.
The final rules included:
- Clarifying definitions for Automated employment decision tools (“AEDT”). This included providing a definition for “substantially assist or replace discretionary decision making for making employment decisions” to cover decisions that rely on a simplified output or overrule human decision-making. The rules also clarified the definition of “machine learning, statistical analysis, modeling, data analytics, or artificial intelligence” to be a group of mathematical, computer-based techniques that generate predictions or classifications and for which the computer at least in part determined the inputs, priorities, or other parameters to make the classification.
- Bias audit examples. The final rules offer several illustrative examples and methods for calculating bias scores. Those scores must be included in published results.
- Historical data. The final rules also offer examples of when companies may rely on a bias audit conducted with historical data, test data, or historical data from other companies
Businesses should remember that non-compliance does have consequences. Forgoing a bias audit or not providing notice are two separate violations and subject to civil penalties. Each day of noncompliance is considered a separate violation that could result in civil penalties.
Putting It into Practice: Businesses that use automated decision making for hiring and promotional decisions should start considering now whether they need to modify these tools. Additionally, businesses should start looking for ways to complete an independent bias audit to ensure that they remain in compliance.
Kathryn Smith also contributed to this article.