Why are so many people angry about New York's AI hiring law?

An overwhelming number of people were outraged by the artificial intelligence and hiring laws that took effect in New York City last week. This law is the first AI law in the United States, so the way it is implemented will provide lessons and guidance for other cities developing AI policy and debate. Like New York, other U.S. states are also considering incorporating AI hiring provisions into the European Artificial Intelligence Act.

The use of artificial intelligence in recruiting has sparked criticism due to the presence of automation and the way it can reinforce existing racial and gender biases. AI systems have been shown to favor white, male, and able-bodied candidates when evaluating their facial expressions and language.

This issue deserves attention because most companies have used artificial intelligence at least once in the recruitment process. Charlotte Burrows, chairwoman of the U.S. Equal Employment Opportunity Commission, said at a January 2023 meeting that as many as 80% of companies use some form of automated tools to make hiring decisions.

New York City’s Automated Employment Decision Tool Act, which took effect on July 5, stipulates that employers who use artificial intelligence in recruitment must truthfully inform candidates that they are doing so. To prove that their systems are not racist or sexist, they need annual independent audits. Job seekers can ask potential employers for information about the collection and analysis of data involved in this technology. Violations are subject to fines of up to $1,500.

(Source: STEPHANIE ARNETT/MITTR | GETTY)

Supporters of the law say it's a good start, even if it's not perfect, for regulating artificial intelligence and mitigating some of the harms and risks that come with its use. Companies are being asked to take a deeper look at the algorithms they use to determine whether the technology is inadvertently unfairly discriminating against women or people of color.

This is a rare but successful case, and from the perspective of U.S. artificial intelligence regulatory policy, we may see more relevant local regulations. Sounds promising, right?

But this law has been hugely controversial. Public interest groups and civil rights advocates say the bill is neither enforceable nor broad enough, while businesses that must comply argue it is impractical and burdensome.

Organizations such as the Center for Democracy & Technology and the Surveillance Technology Oversight Project (S.T.O.P.) argue that the law is “inadequately inclusive” and risks missing the use of many automated systems in recruitment. Including systems that use artificial intelligence to screen thousands of candidates.

Given that the relevant audit industry is currently immature, the more important aspects of the results of independent audits that are uncertain are its more important aspects. The BSA - an influential tech trade group whose members include Adobe, Microsoft and IBM - submitted comments to New York City in January 2023 criticizing the law, arguing that third-party audits were "unfeasible."

S.T.O.P. Executive Director Albert Fox Cahn said: "The key question is how auditors will obtain company information and to what extent they can actually interrogate the company's operations way. Although we employ financial auditors, we lack a set of universally accepted accounting principles, let alone tax laws and auditing rules."

According to Kahn, the law could lead to a false sense of security about artificial intelligence and the hiring process. He said: "This is a fig leaf used only to demonstrate that protections exist, and in practice I don't think any company will be held accountable as a result of this being enshrined in law."

Importantly, mandatory audits must assess whether the output of an AI system is biased against specific groups of people, using a metric known as an "impact ratio" to determine whether the technology's "selection rate" is due to It varies from group to group.

Audits do not need to try to determine how an algorithm makes a decision, and the law sidesteps the issue of “explainability” in complex forms of machine learning such as deep learning. As you can imagine, these omissions have become a hot topic of debate among AI experts.

In the United States, pending federal legislation, we may see more local laws regulating artificial intelligence of this kind, most of which target one specific application of the technology. By engaging with these local legal controversies, we can shed light on how definitions of AI tools, security mechanisms, and enforcement will evolve in the coming decades. New Jersey and California are already considering similar laws.

Support: Ren

The above is the detailed content of Why are so many people angry about New York's AI hiring law?. For more information, please follow other related articles on the PHP Chinese website!