This guest post is by Dr. Ruairidh J.Brown, the Head of Politics and International Relations at Forward College, where he teaches International Political Theory and International Relations. His research interests include phenomenology and hermeneutics, existentialism, the philosophy of amity and friendship, the philosophy of obligation, and developments in international political theory. Ruairidh was awarded the University of Nottingham’s Lord Dearing Award in 2019 for his pedagogical work in China. You can reach Dr. Brown at ruairidh.brown -at- forward-college.eu.
Editor’s Note (Pigeon, that is): I assess this is a useful and sober addition to the AI in the classroom debate. I’ll admit that I count myself as an AI skeptic, aligning with Ruairidh’s concerns below. Ruairidh’s article highlights the importance of expertise, as he had the necessary training and experience to evaluate ChatGPT’s efficacy in this exercise. Now on to the article!
I emailed my students last week to tell them the EU had plagiarised our Foreign Policy Analysis (FPA) seminar.
I was prompted by the story that European diplomats had used a ‘Starmer chatbot’ – an AI trained to respond like Kier Starmer – to practice and prepare for the ‘Brexit-reset negotiations’.
I was joking (in case you hadn’t guessed), but I was struck by how uncanny the EU preparation strategy was to some of my experimentations with AI this year.
I have been, admittedly, quite the AI sceptic – or perhaps even AI apathetic. A lot if this was due to how the need to adapt to AI has been framed in academia: overtly defensive, negative, and usually focused on the need to prevent students from cheating in coursework and exams. When it was suggested to me as a positive, it was usually in a way I found distasteful. In Political Theory, for example, as a means to summarise the works of Hobbes or Thucydides and thus save students the time and bother of reading the originals – something I found, personally as a political philosopher, blasphemous, and pedagogically, as a teacher, a sure way to doom future students to nothing beyond surface level understanding of the material.
Nevertheless, I have been increasingly curious about utilising AI in another module I teach, FPA. Unlike Political Theory, where a deep, nuanced grasp of philosophical ideas is essential, in FPA, a much more strategic and applied understanding is required. For example, students do not need a deep hermeneutic grasp of the concept of ‘harmony’, exploring its development across Confucian thought, as they would in political theory, but rather a grasp of how and what aspects of the idea influence and are leveraged in Beijing’s approach to other States.
You do not need to know, to paraphrase Kant, the thing ‘in-and-for-itself’, but rather, to quote Schmitt, ‘who is to be affected, combated, refuted, or negated by such a term’.
Scenario role plays are thus a perfect tool for FPA as they train students not to consider ideas in abstractum but as part of a strategy. I had, in addition, been keen to introduce some digital gaming models for FPA to make these classes more dynamic, and had been inspired by articles about using such games as Diplomacy. However, I found Diplomacy interface too clunky and a game far too complex to set up.
I thus decided to embrace AI. Afterall, I thought, surely the games I loved as a youngster, like Total War used some kind of AI for their non-player characters (NPCs)? Maybe I could embrace ChatGPT as an NPC for my seminar and see what it could bring to class.
I subsequently created two classes with ChatGPT involved. The first was on North Korea and nuclear weapons. The class was split into three groups of ‘US ambassadors’ who had to try and convince Pyongyang to abandon its nuclear program. North Korea was played by ChatGPT, and whom students would subsequently have to negotiate with.
The second was on economic diplomacy, with the case study of China’s relations with Central and Eastern Europe (ECC). One group would play China and would need to get as many ECC countries to join a new fictitious economic alliance as they could. If China could get, and keep, a majority by the end of the class, they won. ChatGPT played the role of a ‘random events generator’ with the face of Sky News, announcing events such as corruption scandals, civil unrest, and responses from the US in the form of social media posts by President Trump as he reacted to student actions.
These classes, especially the second one, exceeded my expectations.
Firstly, from a technical standpoint, it was incredibly easy to set up and offered incredible flexibility in design. I told ChatGPT what I wanted and how it was to be done, which it not only understood and executed exceedingly well, but even came back which greatly enhanced the class. It was a fraction of the time and effort spent working out Diplomacy.
Secondly, it provided a solid anchor to my seminars. As anyone who has used role plays in classes will know, some student roles require more heavy lifting than others. If a student allocated this role has a certain ‘out-the-box’ perspective, or simply does not do the prework, it can sink the whole class before it sets sail. You can, of course, mitigate this by giving such responsibilities to the same ‘reliable’ student, but this becomes obvious and can create resentment. By giving a key role, like Pyongyang, to an AI however, it allows you to create a reliable anchor and control mechanism on the class and keep it on course. I, for instance, asked it to be tough, which quickly resulted in students who were more unsure or had done less prework finding themselves on the backfoot to an ‘assertive’ North Korea.
Thirdly, it gave the classes more dynamism and kept them alive. We have all designed role plays where the issue is solved, or students get bored, in five minutes. The ‘Sky News’ interventions however kept the scenario in rapid flux, teaching ‘Beijing’ it harder to hold an alliance together than make it and give students in the process a more ‘realistic’ experience of constantly adapting to a ‘live situation’.
However, what the unexpected revelation was for me was the Trump interventions. ChatGPT did this so well, you would have thought it really was Trump watching and responding to our seminars, adding an incredible level of immersion (Trump, of course, works well for such tasks, as there is so much of his rhetoric online for the AI to build a convincing persona on). Indeed, so ‘Trump-like’ was it, students would respond and react to the AI. This resulted in subconflicts developing with the US, which added another level of complexity to the main game. The level of complexity, dynamism, and immersion (as well as, importantly, fun) that resulted from this was remarkable to me.
In short, therefore, inviting AI to my class was much easier than I imagined; gave a firm anchoring and control mechanism; and provided a dynamism and immersion beyond what I expected. Responding to a live situation also gave a sense of stakes and realism to the class. That EU negotiators are also using AI to game plan real negotiations was just an added cherry of realism on top, one I could not but help pointing out to my students.