A bill introduced in the Connecticut State Senate, known as Connecticut Senate Bill 435, would establish some of the strictest state-level regulations on artificial intelligence (AI) across the country. The bill, introduced by the Senate’s Labor and Public Employees Committee, would put regulations in place on how AI can be used for workplace decisions and applications.
The bill would require companies to inform applicants and employees if AI is being used in their employment process. It would also mandate that any AI based decisions be audited and reported to the state, requiring evaluations on potential bias. If passed and signed into law, the bill would go into effect on Oct. 1. The bill has 31 co-sponsors including state senators Gary Winfield and James Maroney, who represent parts of West Haven, as well as Connecticut gubernatorial candidate and state representative Josh Elliott. Maroney has previously introduced multiple bills regulating AI surrounding issues such as chatbots, online safety, personal data and education on AI. A vote was held on a similar bill introduced by Maroney last year and was passed in the Senate but not voted on in the House. Connecticut governor Ned Lamont also opposed the bill.
Part of the bill would also require mandated discussions on AI during any bargaining agreements between unions and corporations.
Some within the Connecticut Department of Labor (CTDOL) have expressed opposition to this bill, such as their commissioner Dante Bartolomeo.
“AI systems should be monitored with human review of system input, output, decision-making logic, errors, accuracy, and appropriateness, yet CTDOL does not directly employ any AI subject matter experts,” said Bartolomeo. “Similarly, CTDOL does not directly employ any human resources subject matter experts. Therefore, while CTDOL does have employment law experts, our agency does not have the expertise to comply with this legislation’s requirements regarding automated employment-related decision systems.”
The opposition is targeted towards section eight of the bill which would require the CTDOL to employ independent auditors and cause a financial reconstruction in the department. The Connecticut Hospital Association also expressed opposition to the bill, fearing that it would financially burn out Connecticut hospitals and put hospital human resources technology under through having to be frequently audited. They also fear that employment could slow down for hospitals under the bill in what they describe is a large market for employment.
Some experts support increased regulation of AI. Vahid Behzadan, a coordinator for the university’s M.S. in Artificial Intelligence program and associate professor in computer and data science said, “We’re at the point where AI systems are making consequential decisions about people’s livelihoods, who gets hired, who gets fired, who gets promoted, how shifts are scheduled.”
“One of the issues that has been very well documented in these sort of AI systems are the bias problems. Numerous reports from researchers show that AI hiring tools can encode and amplify and implode existing biases unintentionally. A resume screener trained on hiring data from a company that has historically underrepresented certain groups will end up learning to replicate that pattern which is because that’s what the data represents.”
One of the main intentions of the bill is to improve transparency between employers and applicants. Susanne Murphy, a professor in practice for the Communication, Film and Media Studies department, and who also put together a new special topics course called “AI Machines of Loving Grace,” said, “We’re talking about hiring and firing decisions, we’re talking about meaningful life decisions, and the more transparency exists about how they are made to me is a good value. I think people have the right to know how the decisions about them are being made.”
Ardiana Sula, another coordinator for the university’s M.S. in Artificial Intelligence program and associate professor in computer and data science said, “It’s going to put some regulations on the AI companies like how to build their AI but it’s also going to start the conversation on how to build responsible AI.”
However, Sula believes that to achieve ethical AI standards, there would need to be regulations put in place nationwide to ensure there is transparency everywhere.
Behzadan also discussed what could change if this bill went into effect for the university. “The Common Application Portals (for staff and teachers) used already employ AI so our university would already become subject to this,” he said. “In recruiting student employees, portals like Handshake and similar portals already employ a good deal of AI. Our university is probably not fully exempt from this.”
Murphy also however discussed the positives to AI for job interviews, discussing how Big Interview, which is part of the class Principles of Communications, can make it easier for students to prepare for interviews and rehearse them or record them at the time of their choice.
