AI-driven data complementarity to support international researchers in climate, health, and agriculture

The potential for AI to support decision making in international development for climate, health, agriculture and associated domains is significant, but the bias within current AI models favour data and use cases from high-income countries means that there is limited potential to satisfy the needs and requirements of users, data stakeholders and communities in low- and middle-income countries. Elements of AI-driven data complementarity, where data from one system can be optimised with data from another to produce new and novel knowledge, offer an opportunity to not only increase the volume, but also the diversity of data. This is especially important for the integration of data from low- and middle-income country partners, who may have valuable data to contribute but lack infrastructure and systems to help make the data more widely available to the broader world, resulting in an imbalance of ideas and innovation. Significant gaps within the data available to AI models impact the way that innovative AI-driven technologies identify existing and emerging challenges and hinder the solutions that are subsequently developed. For instance, more than 450 organizations across 25 African countries maintain high-quality data-rich research repositories that, due to outdated infrastructure and limited resources, keep them from readily contributing to integrated global platforms including climate change guidance and recommendations for policy makers. Similar issues exist across low- and-middle income countries in Asia, Latin America, and the Middle East. In this session, we will discuss new tools an approaches from the Juno Evidence Alliance that try to ensure AI-based analyses and the large-language models (LLMS) integrate diversity of data in ethical, responsible ways.