From predicting public health risks to optimising resource allocation, Artificial Intelligence (AI) is increasingly adept at supporting policymaking. AI’s ability to process vast amounts of data and identify complex patterns almost instantly has enhanced evidence-based decision-making, helping policymakers make smarter decisions.
Associate Professor Eduardo Araral recently led a seminar at the Lee Kuan Yew School of Public Policy to showcase how AI models can empower and embolden governments to tackle real-world policy challenges. The session featured a series of presentations that demonstrated how AI can effectively support policy analysis, addressing issues from corruption detection in infrastructure projects to predictions of how US tariffs can impact Indonesia’s EV industry.
Translating AI Insights into Actionable Policy
Presenters offered a range of AI use cases, drawing on historical data to map out scenarios and potential considerations.
In one instance, an AI model was used to determine the most optimal way to allocate Singapore’s education spend, including which areas it should be deployed to. The team working on this combined linear regression models with current policy trends to determine which policies to fashion.
Another group shared their learnings from using AI for corruption detection in infrastructure project bidding in the Philippines, which was built on three key functions: “predictive analysis, descriptive insights, and general relatability.” This team also referenced other anti-corruption detectors common in infrastructure development, including the
European Union’s Arachne system, China’s
Zero Trust system, and Singapore’s
fraud detection system.
A key takeaway from the seminar was the importance of domain expertise when using AI for policymaking. Professor Araral asserted that no matter the AI proficiency level, only a strong foundation in the relevant subject matter can enable users to ask the right questions, critically evaluate the assumptions and limitations from AI models, and use the findings to inform real-world policies.
Strong communication is also key when leveraging AI for policymaking, Professor Araral shares, especially the ability to convey findings clearly and concisely. It may even be as important as the findings themselves, as was reflected in the segment about corruption detection. A network analysis mapped out the intricate relationships between individuals and organisations and the potential conflicts of interest to give decision-makers a bird’s eye view of the complex web of players in the bidding process.
The potential for effective AI use in policymaking is significant, but sizable hurdles stand in the way. A widening gap between AI’s technical complexities and policymakers’ understanding of the technology remains a pressing issue, putting pressure on developers to create intuitive and user-friendly software and applications for the sector.
Policymakers should also Engage with AI
The solution to making AI an effective tool for policymakers is not unidimensional. AI models developed for policymaking need to be built with the assumption that users only have a basic grasp of the technology. At the same time, policymakers ought to deepen their understanding of underlying methodologies and thought processes.
With that being said, Professor Araral also cautioned against building an overreliance on AI and stressed the importance of critically appraising the assumptions and limitations of AI models they use. Close collaborations and open exchanges between academia, AI developers, and the government can also play a crucial role in creating tools that meet the specific needs of different policymakers and their respective priorities.
Building AI literacy among policymakers involves developing a comprehensive understanding of AI's capabilities and limitations, ethical considerations, and the broader societal implications of its use. Training programs need to cater to different skill levels and aptitudes, and provide tailored support that includes both theoretical instruction and practical applications.
The APEC Policy Support Unit’s policy brief “
Artificial Intelligence in Economic Policymaking,” it notes that the use of AI in policymaking would be vastly more complex and on a much broader scale than using it in a commercial context: “While AI can have immense power in data analysis and logic, policy-relevant concepts such as fairness, justice and equity are inherently human.”
Equipping Policymakers for the Age of AI
Through lectures on AI concepts and methodologies and hands-on projects using real-world data, Professor Araral empowered future policymakers with an understanding of how AI can be leveraged in powerful, pivotal ways, especially when used to address policy challenges.
Fostering a culture of continuous learning within government institutions is crucial for keeping pace with the rapid advancements in AI. Critical thinking will continue to be vital - reiterating the need for simultaneous, symbiotic growth in how AI will be applied to policymaking and governance over time and the humans’ ability to power the tool.
Professor Araral highlighted that AI models primarily generate output based on the extrapolation of historical data to illustrate that all modelling happens in a vacuum. Invariably, AI can make mistakes, and the limitations and shortcomings of AI modeling must be kept in mind when using the tool.
Exercising the Transformative Potential of AI for the Public Good
Professor Araral re-emphasized policymakers’ need to be "smart consumers" of AI products and that AI is meant to augment, not replace human judgement. The consensus at his seminar was to avoid using AI if one was not familiar with how these models worked, reiterating the importance of proper training to familiarize policymakers with the technology.
Considering the potential upsides and challenges of AI in policymaking, it continues to assume an inextricable role in scenario mapping and policy planning. Uniting AI and human input is likely the best way to optimise the process of policymaking, and by extension, the effectiveness of those policies once enforced.