NSW Gov Employers Shouldn't Use AI to Hire Decision Making

AI News


NSW government employers are instructed not to use artificial intelligence to screen or make decisions for candidates applying for roles against bias or potential hallucination concerns.

NSW Gov Employers Shouldn't Use AI to Hire Decision Making


The Public Service Commission's NSW Bureau has issued guidelines that AI should not be used as “not to be used” in recruitment-based decisions, with a particular emphasis on avoiding the use of “highly professional roles.” ”

According to the use of the commissioner's AI in recruitment guidance [pdf]AI can be used appropriately for tasks such as developing recruitment documents, preparing materials and communications, summarizing candidate responses and interviewer notes, handling applicant inquiries, especially when volume is high, and scheduling interviews.

“Institutions that choose to use AI during the recruitment process should ensure that AI use is legal, clearly and transparently documented, and that its risks are properly identified and mitigated.”

“Critical, AI should not replace humans when making decisions.”

Important concerns about AI in employment decisions focus primarily on bias.

These include automation biases that allow employment teams to unconsciously rely on AI outputs excessively. Intrinsic bias built into AI tools. Video interview technology that makes inferences based on candidates' emotions, personality, or appearance. and unreliable evaluation criteria.

Additionally, the report highlights another important risk: errors or manufactured information known as “hastisation.”

From a technology standpoint, the report said, “Purchase of ready-made products looks attractive, but it's not always appropriate in all institutional environments.”

“The product may work in one agency, but there may be unexpected risks in another agency,” the report added.

“When procuring AI products, carefully check marketing and contract materials. [NSW AI Assessment Framework] Requirements.

“This is especially important as AI products may be reluctant to provide information on the decision-making of the algorithms used by them, and therefore AI products may only be tested in a limited way.”



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *