AI could enable 'discrimination', 'surveillance' in APS: CPSU

These concerns are outlined in a submission from CPSU national secretary Melissa Donnelly to the standing committee on employment, education and training, which is currently inquiring into the digital transformation of workplaces, driven by recent advances in AI and machine learning.

On Wednesday, July 3, the committee held its second public hearing where it heard from Professor Toby Walsh, chief scientist at the University of New South Wales AI Institute, that the rapid growth of AI was unprecedented.

"It is not fully appreciated that we are in the middle of the largest gold rush in the history of capitalism," Professor Walsh told the committee.

"[ChatGPT developer] OpenAI is the fastest growing company in the history of capitalism."

The rapid roll out of AI and machine learning tools in the workplace would inevitably create "winners and losers," University of Melbourne economics Professor Jeff Borland said.

"What technology does is that it benefits some workers and disadvantages others. Some see jobs destroyed, others find skills complemented by new technologies."

These concerns were reflected in the CPSU submission, which noted that entry-level roles in the public service were most likely to be affected.

"The advent of computers and the internet had a similar impact on APS roles, leading to a reduction of APS 1 and 2 jobs," Ms Donnelly said.

"It is important the APS maintains strong pathways for entry level employees to work and build their expertise."

Despite this predicted widespread disruption, experts were also sceptical of the power of AI, with Professor Beth Webster from Swinburne University of Technology noting the technology as it stands was good at identifying patterns and correlations, but struggled to interpret the cause of events.

"AI is great because it gets rid of routine jobs, including in the service industry, but it is not going to get rid of analytical tasks; the need to understand what it all means," she said.

Already, the roll-out of automation across the public service has a chequered record, with notable failings including the Robodebt debacle and a 2021 trial of AI to screen candidates for roles within the Australian Tax Office unable to improve recruitment efficiency.

Macquarie University professor Deborah Richards warned that a major limitation of AI was the inherent biases of the technology, in part due to the lack of representation of women in the ICT industry.

Ms Donnelly shared these misgivings about the unchecked application of AI in the APS opening up the potential for discrimination.

"Without proper controls, AI could discriminate against workers exercising their rights, such as taking personal leave, or those with protected attributes, leading to missed opportunities in rostering, recruitment, and promotion, and increased workplace surveillance."

However, the public sector is unlikely to be immune from the AI wave sweeping the workforce, with the Assistant Public Service Minister Patrick Gorman telling a Public Administration Australia ACT event in June that there was "no other option" than embracing AI tools.