The Potential Impact of Artificial Intelligence on Preventive Diplomacy from a Balance-of-Threat Perspective
Author
Oriesek, Daniel Felix
Metadata
Show full item recordCitation
Oriesek, Daniel Felix. 2022. The Potential Impact of Artificial Intelligence on Preventive Diplomacy from a Balance-of-Threat Perspective. Master's thesis, Harvard University Division of Continuing Education.Abstract
When the Peace of Westphalia was signed in the cities of Osnabrück and Münster in 1648, it not only ended the Thirty Years’ War but also, and more importantly, laid the foundation for the contemporary concept of territorial sovereignty and created a system of independent nation-states. These states mutually agreed not to interfere in each other’s domestic affairs while maintaining order by checking each other’s ambitions through the creation of a general equilibrium of power.An essential component in enabling such an equilibrium was the development of a method of influencing the decisions and behavior of the governments and peoples of nation-states so that issues of concern could be resolved by means of dialogue, negotiation, and other measures short of war or violence. This development represented the birth of modern diplomacy.
Over the ensuing centuries, the concepts of equilibrium and balance of power were further developed until Kenneth Waltz integrated and structured these ideas as part of his conception of structural realism. According to this theory, an anarchic self-help system and shifts in the relative distribution of capabilities mean that balances of power recurrently form in the international system and drive international relations (Waltz, 1979).
A few years later, Stephen M. Walt expanded this idea into what he referred to as balance of threat. Walt suggests that states balance not only against actual power but also against real or perceived threats. From Walt’s point of view, balance-of-power theory is not incorrect but rather incomplete. Walt acknowledges that power is a critical factor in the level of threat posed by a state but also suggests that a threat must also include other elements, such as geographic proximity, offensive capabilities, and perceived intentions (Walt, 1985, 1987).
As technical developments, particularly in the form of digitalization and artificial intelligence (AI), transform the way in which people work, and as computer processing power is approaching—if it has not already exceeded—its human equivalent, we need to consider how such factors will change the way diplomacy is conducted in the future.
Because the fields of both AI and diplomacy are relatively broad, this thesis focuses on the field of preventive diplomacy and, within this field, the actions that are directly linked to monitoring and maintaining a balance of threat. With regards to AI, this thesis adopts the system of categorization put forward by Nick Bostrom and focuses on the first category, artificial narrow intelligence (ANI), while also speculating as to what may lie ahead when we reach the second category, artificial general intelligence (AGI), also known as human-level AI (Bostrom, 2014).
This thesis considers the following question: “How might AI change the possibilities and conduct of preventive diplomacy in terms of managing balances of threat for more stability in the world?” Based on the individual assessments of a diverse pool of experts in the fields of diplomacy, security policy, foreign affairs, and AI, this thesis concludes that ANI will make significant contributions to the field of preventive diplomacy from the perspective of better managing balances of threat.
In sum, the experts agree that the potential benefits of AI outweigh the potential risks. The majority of experts believe there are four areas where ANI can make contributions: 1) monitoring, where ANI can contribute to providing a more complete and timely picture of potential changes in the balance of threat and could increasingly be used to develop a system of early indicators to warn of ongoing changes in the balance; 2) validation, where ANI can contribute to validating and improving the quality of information regarding potential changes in the balance of threat, thus increasing the trustworthiness of such information; 3) anticipation, where ANI can contribute to anticipating the reactions of players in a balance-of-threat system and thus could provide valuable decision-making support; and 4) solution finding, where ANI can contribute to detecting relevant players and their relations to one another in a given context as well as assist in detecting seemingly unrelated issues that could expand room for negotiations and solution finding.
Experts were divided as to whether ANI could make contributions in other areas, largely due to doubts that current-level ANI is sufficiently advanced for such tasks. The experts are also divided over whether ANI alone could improve the overall effectiveness of preventive diplomacy, as they fear that political forces and individual interests might work against such efforts.
The experts do not seem particularly concerned about the potential of AI to replace diplomats or security professionals within the next decade, but they are concerned that AI is potentially biased and could be manipulated. The experts see the greatest danger with AI as its potential to replace human decision making or the possibility that decision makers may become overly reliant on results generated by AI.
Therefore, the technical possibilities afforded by current ANI should be used to obtain better validated information more quickly, to better anticipate possible (re)actions by stakeholders, and to identify previously unknown paths to finding solutions. Relying on ANI for these purposes could significantly contribute to identifying potential problems in the balance of threat early on so that the balance can be restored diplomatically before arms races, unhappy alliances, or even preemptive wars ensue. Simultaneously, however, we must remain critical of ANI’s proposals and at least conduct a plausibility check before incorporating its recommendations into our decisions.
Terms of Use
This article is made available under the terms and conditions applicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAACitable link to this page
https://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37370754
Collections
- DCE Theses and Dissertations [1331]
Contact administrator regarding this item (to report mistakes or request changes)