In a significant development for policy analysis, academic Anthea Roberts from Canberra has integrated artificial intelligence (AI) into her research framework, enhancing decision-making processes across various sectors. In early 2023, Roberts received enthusiastic messages from fellow academic and tech expert Sam Bide, who was experimenting with ChatGPT. He had applied Roberts’ complex decision-making model, which she designed to assist policymakers and businesses, to different areas such as Indigenous policy and early childhood education.
Roberts quickly recognized the potential of merging her model with AI. “Suddenly, I had this model that could do this extraordinarily complex systems, interdisciplinary analysis,” she explained. The AI’s ability to navigate economic, security, and environmental perspectives allowed for a more holistic understanding of complex issues, something humans often struggle to achieve independently.
Dragonfly Thinking: A New Frontier in AI Applications
Together with fellow academic Miranda Forsyth, Roberts founded Dragonfly Thinking, which won Australia’s top AI start-up award in 2024 and gained acceptance into the CSIRO accelerate program aimed at commercializing research. The name “Dragonfly” is inspired by political psychologist Philip Tetlock, who found that experts often struggle with forecasting due to their limited perspectives. In contrast, dragonflies possess thousands of lenses, allowing them to synthesize information for a comprehensive view of their environment.
Since its inception, Dragonfly has collaborated with multiple federal departments, including the Department of Prime Minister and Cabinet, as well as private institutions such as major banks and superannuation funds. This collaboration reflects a growing recognition of the potential for AI to enhance policy formulation and strategic decision-making.
Roberts acknowledges concerns regarding AI, particularly fears that it may inhibit foundational skill development and expertise. Critics argue that large language models can produce errors and cannot truly learn. Yet, Roberts maintains that AI can augment human thinking. “I’d never really experienced cognitive strain before, but I’ve done harder critical thinking in the last two years,” she said, emphasizing that AI tools can facilitate deeper analysis and connection-making among users.
Real-World Applications and Challenges
Roberts’ journey includes a diverse background; she has taught at prestigious institutions such as Harvard and Columbia and authored three books. Her most recent work, Six Faces of Globalization: Who Wins, Who Loses, and Why It Matters, heightened her interest in stakeholder perspectives, leading to the development of a “red teaming” tool for Dragonfly. This tool simulates adversarial scenarios to help organizations assess vulnerabilities and strategize effectively.
In one instance, Roberts tested the red teaming tool by creating a hypothetical scenario involving an Australian academic with ties to China. The tool proposed strategies for undermining credibility, prompting Roberts to consider how her fictional counterpart might respond. “It was quite confronting!” she noted, highlighting the tool’s potential for serious strategic analysis.
Experts from various fields, including former chief investment officer Sue Brake and former BHP chief economist Huw McKay, have recognized the utility of Dragonfly’s AI in areas such as mergers, acquisitions, and national resilience assessments. They emphasize that while AI can enhance analysis, it should not replace human judgment. McKay stated, “AI tools without frameworks are just chatbots creating slop,” underscoring the importance of rigorous analytical frameworks.
Despite the promise of AI, Roberts faces challenges in promoting its adoption in Australia. A recent study conducted by the University of Melbourne and KPMG revealed that Australia ranks low in AI trust and optimism, with only 36 percent of Australians expressing willingness to trust AI. Furthermore, 78 percent of respondents expressed concerns about potential negative outcomes associated with AI use.
Roberts and Brake lament the slow adoption of AI technologies in Australia. Roberts stated, “I think it’s very clear that Australia is behind on AI,” suggesting that the country could benefit from a more proactive approach to embracing AI innovations.
As Dragonfly Thinking continues to develop and implement AI tools, it represents a significant shift in how policymakers and businesses approach complex challenges, demonstrating the potential for AI to enhance human decision-making rather than replace it. With ongoing advancements, Roberts’ work may pave the way for a more integrated and effective use of AI in various sectors, shaping the future of policy analysis in Australia and beyond.
