AI治理中的语言陷阱会误导责任归属，需重构术语体系。
Naming is framing: How cybersecurity's language problems are repeating in AI governance
- 用'对齐''幻觉'等术语隐含对抗性假设，弱化人类责任
- 类比网络安全的'最薄弱环节'叙事，放大技术恐慌
- 倡导从语言入手，构建透明、包容的AI治理框架
语言并非中立；它塑造理解、建构权力并影响治理。本文指出，'网络安全'与'人工智能'等术语不仅是语义瑕疵，更因遮蔽人类能动性、夸大技术预期、扭曲问责机制而带来重大治理风险。借鉴网络安全领域的语言困境，如'最薄弱环节'叙事，本文揭示当前AI话语正陷入'对齐''黑箱''幻觉'等隐喻的类似陷阱。这些术语将对抗性、神秘化或过度技术化的预设嵌入治理结构。为此，本文主张采取语言优先的AI治理路径：批判主导隐喻，凸显人类角色，共同构建精确、包容且具有反思性的术语体系。本文强调，语言改革不是治理的边缘议题，而是构建透明、公平且前瞻型监管框架的核心所在。
原文摘要 · Abstract (English)
Language is not neutral; it frames understanding, structures power, and shapes governance. This paper argues that misnomers like cybersecurity and artificial intelligence (AI) are more than semantic quirks; they carry significant governance risks by obscuring human agency, inflating expectations, and distorting accountability. Drawing on lessons from cybersecurity's linguistic pitfalls, such as the 'weakest link' narrative, this paper highlights how AI discourse is falling into similar traps with metaphors like 'alignment,' 'black box,' and 'hallucination.' These terms embed adversarial, mystifying, or overly technical assumptions into governance structures. In response, the paper advocates for a language-first approach to AI governance: one that interrogates dominant metaphors, foregrounds human roles, and co-develops a lexicon that is precise, inclusive, and reflexive. This paper contends that linguistic reform is not peripheral to governance but central to the construction of transparent, equitable, and anticipatory regulatory frameworks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。