Many government agencies are using generative artificial intelligence to support their goals. However, opinions on GenAI’s potential varies among senior public servants – some believe it can transform policy work, while others are cautious due to perceived risks.
Past challenges with technology implementation, including learnings from policy failures identified from the Royal Commission into the Robodebt Scheme, are influencing how senior public servants perceive advancements in the application of GenAI.
The use of GenAI tools has been cautious to address risks and concerns. GenAI outputs can be biased and inaccurate, which can affect policy quality. Publicly available GenAI tools can be like a ‘black box,’ making it hard for the government to ensure transparency. These tools can also reinforce existing biases in the data.
There is widespread agreement, therefore, that the use of GenAI in policy work requires adequate human oversight.
Firstly, there is a need to formalise the intended role of GenAI in policy work. A statement regarding why governments are investing in GenAI is critical to building understanding of its intended contribution to high-quality policy work. Through a GenAI strategy, measures of value can be further identified, monitored, and assessed.
Secondly, it is important to enhance critical AI literacy amongst the senior executive and workforce so that it aligns with other public service values.
There remains a need to maintain a pipeline of expertise and skills in policy craft, including the features that reflect distinctively human forms of analysis and reasoning, while also understanding these tools have an impact on how public servants complete work-related tasks.
The widespread use of these tools is expected to contribute to shifts in how individuals and teams work. Pathways for career development need to be recalibrated to consider whether and how GenAI can augment policy work.
GenAI in policy work requires adequate human oversight.
Thirdly, we emphasise the necessity of ensuring that GenAI adoption occurs in line with regulatory and governance arrangements that account for the breadth of agencies involved in these functions.
Current GenAI tools predict content based on patterns observed within the historical data; they provide plausible answers to questions based on available data. Some complex policy issues, however, require solutions that cannot emerge from what has been done before.
Further research is necessary to understand whether GenAI tools can provide public servants with tools and opportunities to refine and transform policy craft; or a ‘black box’ that might function in ways that are counter to transparency and accountability.
The strength of the GenAI strategy and the implementation of plans is essential to ensuring policy work remains primed to improve to meet the needs of Australian citizens.
Authors: James Connor, Jade Hart, Helen Dickinson, Kate Henne, Vanessa McDermott, and Shibaab Rahman
Leave a Reply