By Haozhi Pan, Professor at the School of International and Public Affairs, Shanghai Jiao Tong University.
Taming an Ocean of Information
Imagine trying to solve a global puzzle where the pieces are scattered across decades, written in hundreds of languages, and buried in thousands of dusty archives, diplomatic cables, and satellite data. This was the monumental challenge facing scientists seeking to understand what makes water cooperation between nations succeed or fail. Now, our study tries to crack the code, not with more diplomats, but with Artificial Intelligence. By harnessing the power of Large Language Models (LLMs), over 70 years of “messy” historical text are transformed into a predictive roadmap for global peace.
The study draws on more than 2,000 documented cases of transboundary water conflict and cooperation worldwide between 1951 and 2019. These cases range from high-profile disputes, such as tensions over the Nile or Indus rivers, to quieter examples of long-term collaboration, like joint river basin organizations in Europe and West Africa. Traditionally, scholars have studied such cases through qualitative analysis or statistical summaries. What makes this study different is its forward-looking ambition. Instead of only explaining the past, it asks what kinds of cooperation are most likely to prevent future conflicts under climate change.
The team employed a technique called Retrieval-Augmented Generation (RAG). They essentially trained their LLM system on a vast library of interdisciplinary literature, from political science theories of cooperation to numerical models of hydrology and climate science. Then, it is unleashed on the messy historical texts. The AI could read a short, ambiguous description of a 1960s water meeting, understand its context through the lens of international relations theory, and codify it in a way that allowed for systematic comparison with satellite-derived water stress data from the same region and period. For the first time, the nuanced “story” of diplomacy could be quantitatively linked to environmental and socioeconomic drivers.

Six Patterns Forged from Data
From this large body of evidence, the study identifies six distinct modes of water-related cooperation that have repeatedly appeared in real-world practice.
Some are formal and legalistic, such as cross-border basin agreements, which establish treaties, rules, and dispute-resolution mechanisms for shared rivers or aquifers. Others are more flexible, such as collaborative planning and adaptation, where countries jointly plan for floods, droughts, and climate uncertainty. There are also more technical forms of cooperation, including joint water allocation models, shared data and monitoring systems, transboundary water quality standards, and coordinated hydropower operations.
Each of these modes reflects a different theory of how cooperation works. Treaties emphasize rules and stability; planning processes emphasize learning and participation; data-sharing emphasizes transparency and trust. Importantly, the study does not treat these as abstract categories. Each mode is grounded in concrete historical examples, from the Mekong River Commission’s data exchange to the Senegal River Basin Organization’s integrated management approach.
Why combinations matter more than single solutions
One of the study’s most important findings is that no single cooperation mode is a silver bullet. Instead, the strongest reductions in conflict risk occur when multiple modes are combined. In high-risk settings, combining cross-border basin agreements, collaborative planning, and joint water allocation is associated with roughly one fewer conflict over five years. This insight matters because many international efforts still focus on one-off solutions. signing a treaty, building a dam, or launching a data platform.

Looking ahead to a warmer, riskier world
The study goes beyond historical analysis by linking its findings to future climate scenarios, using the widely adopted Shared Socioeconomic Pathways (SSPs). Under high-emission, high-stress scenarios, water-related conflicts are projected to rise sharply in parts of South America, Africa, and Asia. Yet the study also shows that targeted cooperation could offset more than half of these projected conflicts in regions such as Europe, North America, and parts of East and Southeast Asia.
The picture is less optimistic for some low-income, highly water-stressed regions, where institutional capacity and diplomatic leverage are limited. In these contexts, even well-designed cooperation frameworks struggle to deliver the same benefits. This finding highlights a crucial equity issue: the places most vulnerable to climate-driven water stress are often the least equipped to implement effective cooperation on their own.
Haozhi Pan, Professor at the School of International and Public Affairs, Shanghai Jiao Tong University.






