2026-05-11 ペンシルベニア州立大学（Penn State）
＜関連情報＞
- https://www.psu.edu/news/research/story/qa-ai-democratizing-global-health-or-reinforcing-old-inequities
- https://journals.plos.org/globalpublichealth/article?id=10.1371/journal.pgph.0006220
感染症モデリングにおける権力バランスの再調整：包括的かつ文脈に即したアプローチに向けて Rebalancing power in infectious disease modelling: Toward inclusive and contextual approaches
Justice Moses K. Aheto ,Megan Auzenbergs,Matthew J. Ferrari,Allison Portnoy,Chigozie Edson Utazi,Romain Glèlè Kakaï,Ezra Gayawan,James M. Azam,Justice Nonvignon
PLOS Global Public Health Published: April 3, 2026
DOI:https://doi.org/10.1371/journal.pgph.0006220
Why now? A critical time for global health equity
Over the past several decades, infectious disease modelling has become a central tool in global health decision‑making, shaping financing decisions, vaccination strategies, and disease control policies [1]; for measles alone, our review identified over 400 modelling studies published since 2000 [2]. However, many of the modelling analyses that have guided these decisions originate in high‑income countries (HICs), even when they intend to inform policy in low- and middle-income countries (LMICs) [3]. With the rapid expansion of Large Language Model (LLM)‑enabled modelling, concerns are intensified about analyses produced without adequate contextual understanding. Models developed at a distance can rely on assumptions that fail to reflect local epidemiology or realities, carrying real‑world consequences for feasibility, equity, and impact.
LLMs, machine learning and other Artificial Intelligence (AI) tools are increasingly being applied in infectious disease modelling, offering rapid data processing and automated model generation—though this is an emerging area, their outputs still require careful validation and contextual interpretation. However, this raises an important question: if anyone can now generate a model using AI, how do we ensure ethics, relevance and local ownership? Recent studies on the utility of LLMs in infectious disease modelling illustrate both promise and limits: Kraemer et al. outline AI’s potential for faster surveillance and forecasting while stressing accountability [4], and Kwok et al. show that AI tools can design models but still require expert validation [5]. Tripathi et al. further emphasise that the benefits of LLMs depend on rigorous validation, transparent processes, and ethical safeguards, stressing that LLMs should complement—not replace—traditional modelling approaches and expertise [6]. Building on this, a roadmap created by Chen et al. highlights that equitable adoption of LLMs in LMICs requires attention to five dimensions—People, Products, Platforms, Processes, and Policies—to avoid reinforcing existing disparities and ensure inclusivity in global health modelling [7].
