In New York City, the city council passed a bill in early November that restricts employers from using artificial intelligence to fill vacancies. The selection of applicants by algorithm is therefore only allowed if the software used has passed an annual bias audit.
A pre-selection of applicants with artificial intelligence must therefore be impartial. If an algorithm discriminates against people on the basis of gender, race or ethnicity and excludes applicants on the basis of this, according to the law, it may no longer be used in New York City in the future.
Right to human review
More and more employers in the USA – and increasingly also in Germany – are using algorithms when preselecting new employees. Written applications and language in job interviews are analyzed by artificial intelligence via video conference.
The law passed in New York City now requires an annual review of the software and is intended to force the providers of such tools to provide more insight into the complex algorithms. In addition, it should give applicants the opportunity to have the submitted documents checked by a person, reports Associated Press.
Concerns and criticism of the harshness of the law
However, there are also concerns about the effectiveness of the law. Some AI experts and digital rights activists fear that the measures did not go far enough and that federal regulators and legislators were setting the standards for review too low.
The approach to testing for bias is good – problematic are the weak and vague standards for it, explains Alexandra Givens, President of the Center for Democracy & Technology. The audits to check the software could therefore serve as a “fig leaf” for the providers.
Another problem, according to Givens, is that the proposal to monitor AI tools is only intended to protect against racial, ethnic and gender biases, while avoiding prejudice against people with disabilities and about their age. The law provides for fines against employers of up to US $ 1,500 per violation – although the conduct of the audits and the proof that the software meets the requirements of the city is left to the providers, Givens criticizes the watering down of the bill.
Manipulation of AI systems
Julia Stoyanovich, a professor of computer science, stated that the best part of the law is the disclosure requirement. People would now learn that they are being rated by a computer and “where their data is going”. “That will shed some light on the features that these tools use,” added Stoyanovich.
There is an increasing interest in job matching platforms and AI-supported job interviews in Europe as well. Applicants report that it is quite possible to impress AI systems and give four very specific tips.
New York City Council approved the bill by 38 votes to 4 on November 10, giving outgoing Mayor Bill De Blasio a month to act. De Blasio supported the law, his office explained whether he would block it or let it come into force with or without a signature, his office did not comment on this, the report said.
(bme)
– .
