Against rising global loneliness, AI companions promise connection, yet accumulating evidence suggests that, for some users and contexts, intensive companion-style use can correlate with increased loneliness and reduced offline socialisation. This position paper challenges the dominant "AI as companion" paradigm by proposing a shift: from AI that simulates relationships with humans to AI that supports relationships between humans. We introduce Relational AI Translation, positioning AI as cultural-relational infrastructure that scaffolds human connection across cultural, generational, and geographical divides. Using first-generation East Asian migrants as a theoretically productive critical case, we outline a multi-agent architecture instantiating three translation operations: emotion-intent decoding, contextual reframing, and relational scaffolding. We articulate design provocations around measurement, safety architecture, and the tension between technological intervention and structural justice, and explicitly frame success as graduation toward renewed human-to-human support rather than sustained engagement with the system.

翻译：在全球孤独感日益加剧的背景下，AI伴侣承诺提供情感连接，然而累积证据表明，对于某些用户和情境而言，密集的陪伴式使用可能加剧孤独感并减少线下社交活动。本立场论文对主导性的"AI作为伴侣"范式提出挑战，主张实现范式转换：从模拟人际关系的AI转向支持人际关系的AI。我们提出关系性AI翻译的概念，将AI定位为能够跨越文化、代际和地理鸿沟支撑人类连接的文化-关系性基础设施。以第一代东亚移民作为具有理论生产性的关键案例，我们勾勒出一种多智能体架构，该架构实现三种翻译操作：情感-意图解码、语境重构和关系支撑。我们围绕测量机制、安全架构以及技术干预与结构正义之间的张力提出设计思辨，并明确将成功定义为逐步实现重新激活人类间相互支持，而非维持对系统的持续参与。