The Institutionalization of Bilateral Cooperation, 1945–2020

All countries in the world conclude international agreements with other countries, at least from time to time. There is a myriad of reasons why this is the case. For example, states may want to trade more with each other and conclude trade agreements as a result. Or countries share access to a river such as the Danube that, for example, becomes the border between Bulgaria and Romania downstream. In these cases, too, states may want to define certain rules of how to use the river by concluding an international agreement. Or countries may strike agreements on cultural cooperation, which facilitate learning languages or student exchanges. Considering this myriad of reasons why countries may benefit from cooperation, it is not surprising that there is a staggering number of international agreements in the world today. In this project, we look specifically at "bilateral" agreements, which are agreements concluded between exactly two countries. Of course, there is also a considerable number of plurilateral agreements, i.e., agreements concluded by more than two states such as the Treaty establishing the European Union. Still, most international agreements are, in fact, bilateral, which is often neglected in "International Relations" (a subdiscipline of political science) which often focuses its attention on major international organizations such as the EU or United Nations (UN). It is precisely this gap that this project intends to close by drawing on approximately 30,000 bilateral agreements concluded since 1945 and registered with the UN Secretariat. In particular, this project focuses on "joint bodies", which are administrative structures set up relatively frequently in these agreements. A joint body could be a joint committee or joint commission and allows both governments to meet regularly to discuss the implementation of the agreement and even monitor each others behavior. This is because international agreements, usually, require states to take certain actions even after the agreements have been concluded. Put differently, just signing an agreement is not going to have too much of an impact if no one cares about it afterward. Although literally thousands of such joint bodies have been set up in the past, few researchers have looked at them and asked questions like why some are stronger or weaker (e.g., whether foreign ministers meet in joint bodies or only civil servants) or what effects they have on international relations, more broadly. To bring us closer to answering these questions, we will ask students to help us identify joint bodies in these agreements and then also program our own script in R (an open-source statistical software) that will help us to computationally determine their strength. Overall, this project will mark a major contribution and further our understanding of international relations, giving rise to results that are interesting for researchers and citizens alike.

This project applied Artificial Intelligence (AI) to study approximately 30,000 bilateral international agreements contained in the United Nations Treaty Series (UNTS). The UNTS is an extensive collection of international agreements assembled by the Secretariat of the United Nations (UN). Specifically, the project focused on the institutional mechanisms - the so-called "joint bodies" (JBs) - that states regularly include in their international agreements to facilitate future meetings. Essentially, joint bodies provide an organizational framework for states to discuss the implementation of the agreement in a well-thought-out and deliberate manner. For example, a bilateral international agreement may establish a "joint committee" or "joint commission" that commits the government of both states to send national delegates and meet at least once a year to thoroughly review the implementation of the agreement. Although literally thousands of such joint bodies have been created since 1945, we previously knew surprisingly little about how they were designed and what broader effects they entailed for international relations. This project, hosted at the University of Salzburg, used recent advances in AI to apply cutting-edge techniques and study the design and effects of joint bodies. The project compared the performance of various Large Language Models (LLMs) in terms of their ability to understand even complex international agreements and identify joint bodies. We used various LLMs to thoroughly explore these abilities but with an emphasis on open-weight models to make the research process more transparent and replicable. With the help of LLMs, we identified which international agreements set up joint bodies - and how sophisticated they were. For example, some joint bodies meet at least once a year, at senior levels (e.g., among foreign ministers), and are authorized through the international agreement to take binding decisions - whereas other joint bodies lack most or all of these features. What previously required time-intensive manual labeling, typically limiting existing research to a few hundred international agreements at most, could be studied across tens of thousands of agreements using AI. The project also contributed to our understanding of the effects of joint bodies on international relations. Overall, the project generated new evidence on the design and effects of joint bodies in bilateral international agreements and made a significant contribution by demonstrating how AI can be used to study international cooperation.