GIST Proposes AI-Based Strategy for Elderly Health Prevention Campaigns

Effectiveness Decreases When Budget Is Spread Across Multiple Media Channels
AI Simulation Provides a "Policy Experimentation Environment" for Scenario Comparison

Gwangju Institute of Science and Technology (GIST) announced on April 30 that the research team led by Professor Deungjo Gong at the Graduate School of AI Policy and Strategy has presented a media allocation strategy for elderly health prevention campaigns that simultaneously considers both effectiveness and equity, utilizing AI-based simulations to address the challenges of a super-aged society.

(From left) Professor Gong Deuk-jo, AI Policy and Strategy Graduate School; Master’s student Yuna Kim; Principal Researcher Jihye Lee, Seoul National University (external researcher at GIST); Master’s student Jooyoung Park. Courtesy of GIST

This study empirically demonstrated the possibility of not only increasing preventive health service participation among the elderly but also mitigating gaps between groups caused by differences in digital accessibility, income, and social conditions. Furthermore, it is significant in that it proposes an AI-based decision-making framework that enables both effectiveness and equity to be considered from the policy design stage.

As Korean society rapidly enters a super-aged era, preventive health policies such as vaccination and health screenings are becoming increasingly important. However, existing campaigns are often designed based on overall average participation rates. As a result, despite rising participation, some elderly groups remain excluded from information due to differences in digital accessibility, literacy, living environment, and income level. This structure ultimately widens health disparities and can lead to increased social costs over the long term, highlighting the need for policy designs that address both efficiency and equity.

To address these issues, the research team employed agent-based modeling (ABM), which sets each individual as an "agent" (virtual individual) interacting within social relationships and simulates the spread of information and behaviors in a computer environment. By integrating data from the Korea Media Panel Survey and the Korea National Health and Nutrition Examination Survey, the team created a virtual environment with approximately 2,400 elderly agents reflecting media usage, digital competence, social relationships, and health behaviors. The simulation tracked how these agents encountered information via TV, digital, and print media, interacted with those around them, and ultimately engaged in preventive behaviors such as vaccination and health screening.

Notably, the research team devised 15 different policy scenarios—not a single strategy—to systematically compare and analyze how media allocation, budget levels, group-specific customization strategies, and messaging methods influenced outcomes.

Analysis Framework for Elderly Prevention Campaign Strategies Using Agent-Based Modeling (ABM). Provided by GIST

The analysis revealed that, contrary to intuition, dispersing the budget across multiple media channels resulted in a "dilution effect," which reduced the effectiveness of the campaign. Under the same budget conditions, using both TV and digital in a multichannel strategy actually produced lower average participation rates than focusing on a single channel. This was attributed to the diffusion of social amplification effects, where repeated exposure within one channel and sharing among peers more effectively led to behavioral changes.

The research team further segmented the elderly population into six distinct groups based on media usage patterns, digital affinity, and social activity levels, rather than treating them as a single homogeneous group. For example, one group exhibited high digital engagement but low TV viewing, while another relied heavily on TV but minimally used digital media. Each group accessed information differently. Reflecting these differences, the team designed a "group-customized channel allocation strategy," applying TV-centric, digital-centric, or mixed strategies as appropriate for each group.

Applying these customized strategies improved both effectiveness and equity. Among the group with the lowest initial vaccine participation, the rate increased from 86.8% to 90.9%. For health screenings, participation rose from 77.6% to 85.3%. Simultaneously, the participation gap (difference between highest and lowest groups) decreased by up to approximately 33%. This demonstrates the feasibility of policy designs that not only raise average participation but also uplift vulnerable groups.

This study introduced a "policy experimentation environment" that enables the comparison of various scenarios through AI simulation before actual policy implementation. This allows for the prediction of the effectiveness of resource allocation and targeting strategies prior to execution, and it has significant potential for application in other fields such as energy, urban planning, and public services.

Professor Deungjo Gong stated, "This research demonstrates that AI can contribute to solving complex societal issues, such as those faced by a super-aged society, beyond merely serving as a technical tool. Going forward, it will be important to design policies that leave no one behind, rather than simply pursuing policies with the highest average effect, and AI will become a key means of realizing these social values."

This study was conducted by Professor Deungjo Gong at the GIST Graduate School of AI Policy and Strategy (supervisor), Jihye Lee, Senior Researcher at Seoul National University (lead author), and Jooyoung Park and Yuna Kim, master's students at the GIST Graduate School of AI Policy and Strategy (co-authors). The research received support from the Ministry of Science and ICT and the Institute of Information & Communications Technology Planning & Evaluation (IITP) AI Graduate School Support Project, the Science and Technology Promotion Foundation for Special Zones' Regional Science and Technology Project for the Future, the National Land and Science Technology Promotion Agency's Urban Convergence Special Zone R&D Project, and the GIST-InnoCORE Project.

This content was produced with the assistance of AI translation services.