The question at the first WebDebate of 2023 was: Will AI take over diplomatic reporting? Four speakers joined us to try and answer this question: Dr Jovan Kurbalija, director of DiploFoundation; Ambassador Umej Bhatia, the head of Singapore’s mission to Geneva and Vienna; Irena Zubcevic, senior advisor to the Permanent Mission of Croatia to the UN in New York; and Liz Galvez, former senior British diplomat and Diplo senior fellow and lecturer. More than 200 participants from around the world joined the debate and contributed to a lively discussion in the chat. This is the ‘human’ summary of WebDebate #56. A part of the real-time AI-generated debate summary was presented by the head of the Diplo AI team, Jovan Njegic, at the showcase that followed the debate. The human summary you are reading features key points from each speaker, the answer to one of the Q&A questions, and a summary of key points. See the AI-generated reports from the web debate here. Setting the scene, Kurbalija described diplomacy as a piece of machinery and placed diplomatic reporting at its core. Historic examples underscoring the centrality of reporting include the Armana archives and Renaissance missions. The first ministry of foreign affairs, established by Cardinal Richelieu in 17th century France, was built around the diplomatic archive. To this day, a considerable amount of diplomats’ time is spent on reporting. Kurbalija argued that some of the reporting work of diplomats cannot be automated, but that exploring which aspects can be taken on by AI is crucial for the future of the profession and diplomatic training. Jovan Kurbalija: What part of the diplomatic garden should be kept away from the AI avalanche? Amb. Bhatia highlighted three elements of the work of a multilateral mission: 1) reporting from multilateral meetings; 2) bilateral exchanges and lobbying, 3) analytical reporting. All three areas can be assisted by AI tools, yet ‘diplomats in the loop’ will always be required. First, when it comes to reporting national statements from multilateral meetings, established positions are often restated. In these instances, AI tools can be used for recording, transcribing, and summarising. This could save smaller missions and international organisations a lot of time and resources. The ‘diplomats in the loop’ are required to check for accuracy and address misunderstandings, mistranslations, potential double-meanings, and the implications of certain phrases. Second, regarding bilateral work and alliance building, AI tools might be helpful in fine-tuning talking points, acting as an aide memoir, and suggesting alternative formulations or phrasing. Third, in analytical reporting, AI may serve for background research and collecting data. But this area of diplomatic work crucially relies on human judgement and interpretation.
Watch WebDebate #56: Will AI take over diplomatic reporting? here.
Ambassador Bhatia: Texts written by AI can mimic authority without any substance. Contextual and situational judgement is crucial and cannot be provided by AI.
Drawing on her experience as former focal point at the UN Department of Economic and Social Affairs for voluntary national reviews on sustainable development goals (SDG) implementation, Irena Zubcevic highlighted ways in which AI tools can support the SDG review process. Since working in silos is not useful for SDG implementation, AI could help to analyse data, search and combine various data basis, identify trends, and help understand the impact of policy actions on different sectors. AI, however, would not be capable, in any meaningful way, of (a) interpret trends; (b) making decisions on how to utilise interlinkages and trade-offs between SDG implementation in different sectors; (c) correcting the course of SDG implementation. In a sense, AI lacks the creative thinking for these kinds of tasks. SDG reporting is often done with limited human capacities. Developing countries in particular could benefit from AI tools to support their SDG review process. User-friendly, and freely or cheaply available AI tools are a needed for that.
Irena Zubcevic: Can AI be used for diplomatic reporting? Yes. Can AI do better than humans? It depends.
Former British senior diplomat Liz Galvez introduced a distinction between factual and analytical reporting. Galvez’s experience reporting from the UN General Assembly general debate is an example of factual reporting. It is mainly done for the records of the ministry to have a synthesis of what was said, when, and by whom. The verbatim record of this already exists, but diplomats produce a more digestible form for the purposes of their own ministry. Doing this can be a tedious job. Speeches are repetitive, overrun, and it’s very unlikely that something new will be said that was not signaled in advance.
This type of reporting, at least in part, is an ideal candidate for AI automation. Human diplomats will be needed to provide commentary, context, and judgement on relevancy regarding strategic interests, and highlight unusual or noteworthy incidents. Similarly, standard monthly economic or political roundups, based on media and other publicly available materials, are also candidates for AI-automated diplomatic reporting.
Diplomats would only be have to add succinct commentary and provide political judgment. AI cannot meaningfully answer questions such as: What should happen next? How is the country’s national interest or policy affected? It also remains unclear how the veracity of sources is ensured and how an AI tool can deal with contradictory public information or opposing views. Diplomats also use their own observations, information from their trusted networks, and off-the-record exchanges for their reporting. It is clear that these inputs to reports cannot be replaced by AI.
Liz Galvez: I can’t see any circumstances where AI would ever be able to produce that kind of analysis because it is based on human interaction and connections and on political judgement.
Speakers agreed that AI will be able to help in producing the zero draft of a resolution, which forms the basis for further negotiation. An AI draft might be good enough to only leave a few words or phrases up for further negotiations. Further, for diplomats stuck in tense and exhausting negotiations, AI tools can suggest alternative text formulations and alternative wording. Trust is crucial. Parties to the resolution will need reassurances regarding algorithmic biases. Further, an AI tool might need to be trained not only on previous successful resolutions but also on failed resolutions. AI tools can also help with background research such as finding relevant related resolutions and creating an overview of the previous positions of countries on topics .
