Starting this week, New York City job seekers will participate in a previously hidden part of the application process to learn if and how artificial intelligence is being used in hiring decisions. Become.
of the city Automated Hiring Decision Tool Enacted in 2021 and set to go into effect on July 5, the law positions New York City as a regulatory leader. Use of AI in recruitment. Other cities and states are expected to gradually follow suit.
This narrowly-coordinated law is designed to offset the potential abuses of AI that could significantly impact workers’ lives. Specifically, it will require companies that rely on AI tools to make hiring decisions to disclose this fact to their candidates. It will also conduct annual third-party “bias audits” of the technology and software employers use to expose how AI may discriminate against certain types of candidates. is also required.
“If an employer does use an Automated Employment Decision Tool (AEDT), the employer will commission an independent audit, publish a summary, and inform applicants and employees that they are using the AEDT. , we need to give applicants an opportunity to be considered,” and pursue an alternative selection process,” Pharrell Fritz’s employment attorney Domenic Camacho Moran told CBS Moneywatch. “We are only talking about tools that replace human decision-making.”
Audits aim to monitor sometimes controversial tools that the companies themselves don’t necessarily understand. AI screening tools can save companies time, but automated decision-making has also been criticized for reproducing stereotypes and, in some circumstances, disadvantaging women and people of color.
“That’s the whole risk, and it’s left unchecked, and sometimes humans can’t even explain what data points the algorithm is detecting,” said John Hausknecht, a human resources professor at Cornell University. That’s largely what’s behind this bill.” College of Industrial and Labor Relations. “Let’s track it, collect data, analyze it, report it, so we can change regulations over time.”
However, potential adopters have limited ability to opt out if they don’t like being evaluated by AI. The law stipulates that AI screening disclosures “must include instructions on how an individual may request an alternative selection process or other legally reasonable accommodation (where possible).” Employers do not actually have to use a separate screening process. .
replace human decisions
This law penalizes companies that do not comply. The first offense carries a fine of $500, and subsequent offenses carry fines of up to $1,500.
Importantly, the scope of the law is very narrow.
Niroy Ray, an employment attorney at Littler Mendelsohn, told CBS Moneywatch, “This is the first law specifically to call out and specifically regulate automated decision-making employment tools.” “This is only focused on the use of AI in hiring or promoting employees, not other employment lifecycle decisions.
Using AI tools alone is not enough to mandate disclosure. For the law to apply, AI tools must have a direct impact on hiring outcomes.
“We’re only talking about tools that replace human decision-making,” said Camacho Moran. “If you have an AI tool that runs 1,000 applications and says, ‘These are the top 20 candidates,’ that’s obviously a tool that fits his definition of AEDT.”
“On the one hand, if the AI is designed to put people in buckets so that these candidates have the right experience and have the right education, choose criteria. It is not a tool by definition.”
In other words, if AI flags candidates with relevant experience, but humans still review all applications and are the final decision-makers, the law is likely not to apply. says Camacho Moran.
What are the hurdles for SMEs?
Some critics of the law say its punitive nature and biased audit requirements make it a burden, especially for small business employers trying to leverage AI to streamline and improve their hiring processes. Some argue that there is.
Littler Mendelsohn’s Ray said, “This requirement is cumbersome as it only adds cost to the hiring and promotion process in New York City.” “Therefore, there is some risk of non-compliance in one way or another, as it was not clear what needs to be done to comply. And compliance certainly comes at a cost.”
For example, Emily M. Dickens, head of regulation at the Human Resources Management Association, opposes fines.
“It was a good faith attempt to put in place regulatory guardrails around this issue that could negatively impact some people if not used correctly,” she said. “But we should assume good intentions until something very egregious happens. This is the first law of its kind and will likely be similar in other jurisdictions, encouraging people to try to do the right thing.” I don’t want to start by punishing them.”
Given that many employers still struggle to recruit a diverse workforce and that a human-centric approach leaves out qualified candidates, she advocates for the responsible use of AI in the hiring process. increase.
“This process has been done by humans for many years, but it’s been a challenge to create more inclusive workplaces, access diverse talent, and meet the needs of companies struggling to find talent,” Dickens said. The issue of doing so has not yet been resolved,” he said. “We need guardrails, but we don’t need overregulation at the expense of workforce innovation.”
