Protecting Workers from AI Bias: Victoria Takes Action
This week, the Victorian government announced a groundbreaking initiative aimed at safeguarding workers against the “inappropriate and unfair uses” of artificial intelligence (AI) in the workplace. With a focus on ensuring that AI hiring decisions are “free from bias or discrimination,” the new laws are set to address pressing concerns regarding automated systems that might perpetuate discrimination.
The Shift Towards Regulating AI in Hiring
The Victorian government’s decision is a direct response to growing apprehensions about how AI technologies can affect hiring practices. By introducing laws to regulate AI, the state aims to counteract the risks posed by biased algorithms that might not only hamper diversity but also violate anti-discrimination laws. Business groups and some employment lawyers have criticized this move, labelling it unnecessary overreach, while unions see it as a proactive step toward worker protection.
The Danger of Bias in AI Hiring Systems
Evidence abounds regarding the potential for AI hiring tools to systematically disadvantage various groups, including women and individuals from diverse backgrounds. Research from the United States analyzing 3 million job applications revealed alarming instances of racial discrimination against Black and Asian applicants. When these biased systems are adopted by multiple employers, entire demographics may find themselves locked out of job opportunities.
My own studies further illuminate the impact of AI hiring systems on marginalized groups. Research suggests that candidates with disabilities, older workers, women, or individuals who speak English as a second language often face exclusion. This discrimination can stem from flawed data sets, biased algorithms, or even from the way organizations implement these technologies.
Do We Need New Laws in Victoria?
This push for regulation comes at a time when existing legislation falls short. The Victorian Equal Opportunity Act, drafted before the advent of AI technologies, effectively addresses discrimination in hiring practices but lacks specific provisions regarding the use of AI systems. Notably, there’s no obligation for employers to make reasonable adjustments for job applicants with disabilities.
More Read
As AI applications become increasingly integral to the hiring process, queries arise about accountability. Who is responsible if a biased decision is made by an automated system— the employer or the AI developer? The opacity of AI algorithms complicates this issue, making it challenging for victims of discrimination to seek recourse.
A Focus on Preventing AI Harms
The absence of federal regulation on AI has left states like Victoria to fill the void. The Victorian government’s initiative is a welcome attempt to establish standards, especially following the federal government’s retreat from implementing mandatory guidelines for high-risk AI systems. Such guidelines would require companies to take preemptive measures—like auditing their algorithms for bias—to mitigate potential harms.
Victoria’s proactive stance aligns with similar legislative efforts in New South Wales, which has also passed laws concerning AI regulation in the workplace. However, for a more cohesive approach, regulation at the federal level would be optimal to avoid a fragmented legal environment across states.
Implementing Safeguards Against AI Discrimination
The Victorian Equal Opportunity Act mandates that employers take “reasonable and proportionate measures” to eliminate all forms of discrimination, including those stemming from AI applications. Incorporating a human element in the hiring process—such as requiring a human to have final authority on significant automated hiring decisions—is one such measure being considered by the government.
Additionally, independent audits to assess for bias in AI hiring systems are crucial. Current regulations lack clear guidance on the necessary steps employers must take to ensure compliance with anti-discrimination laws, and therefore, clarity is indispensable.
There is also a pressing need to address the use of AI for emotion recognition in hiring processes, a practice that has been scientifically discredited due to its potential for racial, gender, and disability-based discrimination. The Victorian government’s indication to eliminate this practice mirrors the European Union’s stance, which restricts emotion detection AI to very limited scenarios.
The Path to Bias-Free AI Systems
Achieving a completely bias-free AI hiring system is an immensely challenging goal. Given that these systems are human-made and rely on human data, absolute neutrality may be unattainable. Furthermore, there’s no universally accepted methodology for detecting discrimination in AI systems, making oversight difficult.
However, steps can be taken to improve the existing framework. A comprehensive review of the Victorian Equal Opportunity Act is necessary to close gaps in protection and to strengthen measures aimed at preventing discriminatory practices before they occur.
Employers should avoid employing AI systems if there’s uncertainty about their neutrality. Empowering the equality regulator in Victoria with the necessary tools and resources to enforce legal protections against AI discrimination will be pivotal in creating a fair workplace environment.
As recent surveys indicate that only a fraction of Australians believe current regulations adequately protect against AI-related harms, it is clear that the need for action is urgent.
Inspired by: Source
