NTU builds AI tool to simulate citizen response to economic policies

Researchers at Nanyang Technological University (NTU) in Singapore are developing an AI-driven tool to predict how citizens respond to government economic policies. The project aims to provide a more accurate simulation of household spending behavior, specifically for initiatives like cash handouts, helping policymakers refine strategies before implementation.

Led by Professor Hyeokkoo Eric Kwon, Nanyang Business School Provost’s Chair Professor in Information Systems, the project is one of 14 worldwide initiatives funded by OpenAI to promote societal resilience. OpenAI provided US$100,000 in funding for the effort. To build the system, the team is training AI agents on anonymized transaction data from more than one million users of a South Korean mobile budgeting app. Professor Kwon notes that AI can predict human behavior more effectively than traditional simulations based on deterministic rules because it recognizes complex patterns and possesses a vast knowledge of human behavior.

The Gap in Economic Modeling

Traditional economic simulations have long relied on qualitative interviews and surveys to gauge how the public might react to policy changes. However, these methods often reveal a significant gap between what people report they intend to do and how they actually spend their money. By shifting from survey-based data to large-scale, real-world transaction records, NTU researchers intend to bridge this divide. This approach allows the AI to simulate actual spending habits rather than relying on the self-reported intentions of a small sample group.

Implications for Public Policy

This technology could significantly reduce the risk of policy failure by allowing governments to simulate the precise economic impact of public spending. Policymakers could potentially determine whether funds are likely to be saved or spent, and exactly where those funds will flow within the economy, before committing public resources. Such precision reduces the guesswork involved in fiscal stimulus and allows for the testing of alternative policy scenarios in a virtual environment.

Future Outlook

While the current model leverages data from South Korea, the researchers aim to evolve the tool into an open-source resource applicable to any population. This move toward a population-agnostic tool suggests a broader application for global economic planning. Regarding data security, Professor Kwon emphasized the importance of anonymity, stating that there should never be a situation where the data can only be matched to one profile.