The Future of International Relations Theory?
If the world is getting more complex, will theorists rely on new tools to explain it?
[WARNING: If you are not interested in international relations theory, do not even start to read this, just move on.]
Yesterday I wrapped up my Seminar in International Relations (IR) theory class at The Fletcher School. In my day I have taught other IR theory seminars — one based on classic texts and one focused on power, for example. This was the first time I have taught that the general survey course in a good long while. And there is something that came up in the last class that keeps gnawing at me.
The theme of the last session was "The End of International Relations Theory?”1 IR scholars probably remember a flurry of articles and special issues about this topic from a decade ago. Several of the articles I assigned proffered longstanding complaints about the state of IR research: too many IR theory “camps” that do not talk to each other, too heavy a focus on “simplistic” hypothesis testing at the expense of theory-building, or a discipline weighted down by its origins in political science.
Some of the complaints, however, were more novel in nature, suggesting that the world had changed so much in recent years that extant theory could no longer cut it as an explanatory tool. For those interested in environmental politics, the most significant shift in anyone’s lifetime is the shift from the Holocene to the Anthropocene. This radical increase in the complex interaction between humans and their environment, however, has not been matched by significant shifts within international relations theory. For those interested in globalization, the growing complexity of interdependence leads to a situation in which the global political economy has taken on the qualities of a complex adaptive system — an entity that renders most IR theories moot. And those who believe the world is currently experiencing a global polycrisis argue that the planet is at risk from the causal entanglement of crises in multiple global systems that are newly interlinked. These kind of tightly coupled, complex adaptive systems are beyond the explanatory power of standard international relations paradigms.
Pick your theoretical driver, but what these three arguments have in common is the notion that the international relations system is more than the sum of its parts. Now this, in and of itself, is nothing new. Many IR scholars argue that the international realm is different because of anarchy.2 For these same scholars, however, the difference is not transformative. Realists believe that the anarchic ordering principle suggests continuity. Open economy politics scholars believe that systemic outcomes can be determined by aggregating the power and preferences of individual actors. Marxism and constructivism tell more teleological stories, but those family of theories still implicitly posit the existence of a stable end state.
Scholars who focus on rising levels of systemic risks however, argue that the density of global interactions has grown so complex that the very idea of equilibrium loses its meaning. The system itself develops emergent properties distinct from the component parts of the system. If these arguments are correct, world politics is shifting from homeostasis to hysteresis.
This leads to a rather provocative question: if international relations is shifting from a world defined by anarchy to a world defined by complexity, what are the best theoretical tools that can be used to model it? Formal modelling seems inadequate to this task. Complexity theory might be of some use, but it might not be terribly predictive either.3
To put it plainly: will the next generation of IR scholarship need to rely on artificial intelligence as a theoretical aid?
The hard-working staff here at Drezner’s World has previously expressed skepticism about the viability of human reliance on AI. But trying to explain complex adaptive systems might be an area where generative AI has a comparative advantage over human forms of cognition. Even simple forms of artificial intelligence can read and interpret network structures, providing useful insight to human IR scholars. And those same scholars are beginning to argue that the world is more and more networked than at any time in human history. A theorist who knows how to exploit AI to the fullest might be able to develop new and interesting theories of he world.
Can AI alone be a competent theorist? Nah. Humans are necessary, because they have an important comparative advantage As I wrote earlier this year: “By their very programming, AIs are prone to overconfidence bias — they always think they have a good bead about what they are analyzing. The AI’s lack of metacognition is always its greatest weakness.” Human scholars need to provide the right test data, subsequently evaluate an AI’s speculative findings, and decide which AI hypotheses would be worth pursuing.
Full disclosure: I don’t really think I am correct. I can see AI as a gateway to an awful lot of bad theorizing. At the same time, I cannot dismiss this idea entirely. And now all of you get to consider whether my supposition has any conceptual legs.
For those IR nerds reading this, here are the assigned readings: Christine Sylvester, “Experiencing the End and Afterlives of International Relations/Theory,” European Journal of International Relations 19 (September 2013): 609-626; John J. Mearsheimer and Stephen M. Walt, “Leaving Theory Behind: Why Simplistic Hypothesis Testing is Bad for International Relations,” European Journal of International Relations 19 (June 2013): 427-457; Cameron Harrington, “The Ends of the World: International Relations and the Anthropocene,” Millennium 44 (June 2016): 478-498; Justin Rosenberg, “International Relations in the Prison of Political Science,” International Relations 30 (June 2016): 127-153; Thomas Oatley, “Toward a political economy of complex interdependence,” European Journal of International Relations 25 (December 2019): 957-978; Michael Lawrence et al, “Global Polycrisis: The Causal Mechanisms of Crisis Entanglement,” Global Sustainability 7 (January 2024): 1-16.
#NotAllIRTheorists, however.
To be fair, that is one possible conclusion to draw: IR theories should simply be mote modest in their predictive expectations.
Why not use AI agents to simulate the interactions of nations and back test against specific events? This seems doable today, but difficult even a year ago.
I am a simple retired high school teacher (served mostly farmworker kids in the San Joaquin Valley) but in a former life was a state university political science grad student in the Age of Reagan. My last seminar was in IR, spring 1984(!). I remember that we read Machiavelli and Thucydides, from whom I used the Melian Dialogue to lever my discussion about justice and dominance as it applied to U.S. intervention in Nicaragua. I had migrated ideologically from a military brat’s Cold War true believer perspective to something of a reformed drunk’s self-righteous neo-Sandalista one.
Before that I remember Rational Actor vs Bureaucratic Politics models vying to explain. Some other names still float in the wastewater of my brain: Waltz, Art, and even Wallerstein. As much as it entertained me for a time, my 32 years in the public high school classroom was more fruitful and rewarding.