[论文解读] Brokerage in the Black Box: Swing States, Strategic Ambiguity, and the Global Politics of AI Governance

Ha-Chi Tran|arXiv (Cornell University)|Jan 10, 2026

Ethics and Social Impacts of AI被引用 0

一句话总结

论文提出“技术性摇摆州”为具备科技能力与战略灵活性的中等强国，在AI不透明性背景下通过强调制度透明度而非技术可解释性来促成AI治理，韩国、新加坡和印度为案例研究。

ABSTRACT

The United States-China rivalry has placed frontier dual-use technologies, particularly Artificial Intelligence (AI), at the center of global power dynamics, as techno-nationalism, supply chain securitization, and competing standards deepen bifurcation within a weaponized interdependence that blurs civilian-military boundaries. Existing research, yet, mostly emphasizes superpower strategies and often overlooks the role of middle powers as crucial actors shaping the global techno-order. This study examines Technological Swing States (TSS), middle powers with both technological capacity and strategic flexibility, and their ability to navigate the frontier technologies' uncertainty and opacity to mediate great-power techno-competition regionally and globally. It reconceptualizes AI opacity not merely as a technical deficit, but as a structural feature and strategic resource, stemming from algorithmic complexity, political incentives that prioritize performance over explainability, and the limits of post-hoc interpretability. This structural opacity shifts authority from technical demands for explainability to institutional mechanisms, such as certification, auditing, and disclosure, converting technical constraints into strategic political opportunities. Drawing on case studies of South Korea, Singapore, and India, the paper theorizes how TSS exploit the interplay between opacity and institutional transparency through three strategies: (i) delay and hedging, (ii) selective alignment, and (iii) normative intermediation. These practices enable TSS to preserve strategic flexibility, build trust among diverse stakeholders, and broker convergence across competing governance regimes, thereby influencing institutional design, interstate bargaining, and policy outcomes in global AI governance.

研究动机与目标

• 解释为何AI不透明性是持续性的结构性特征，而非单纯的技术缺陷。
• 主张治理依赖的是制度与程序透明度，而非模型层面的透明度。
• 识别中等强国如何利用不透明性来影响全球AI治理与标准。
• 分析三个案例国如何在不透明性下维持战略灵活性并促成不同治理体制之间的共识。

提出的方法

• 开展对韩国、 新加坡和印度的比较性、基于案例的分析。
• 在概念层面应用网络理论中心性思想（度数、介数、特征向量）来解读战略定位，而非使用正式的定量模型。
• 主张不透明性源于算法复杂性与政治经济激励，而不仅仅是保密。
• 描述监管与治理机制（认证、审计、披露、监督）如何替代全面模型透明度。

实验结果

研究问题

• RQ1 AI不透明性如何作为结构性条件运作，而非技术短板？
• RQ2 技术性摇摆州在何种方式通过不透明性与制度透明度影响AI治理？
• RQ3 韩国、新加坡与印度如何在竞争治理体制之间促成共识？

主要发现

• 不透明性是由模型复杂性和激励所导致的结构性特征，优先考虑性能而非可解释性。
• AI问责从模型可解释性转向制度与程序透明度。
• 技术性摇摆州通过延迟、选择性对齐和规范性居间来维持战略灵活性。
• 认证、审计与阶段性监督等治理机制能够在不透明性下管理不确定性。
• 网络位置分析有助于解释接近美中核心的国家如何获得调解杠杆而不具强制力。