技术乌托邦想象下智能翻译困境——过度适配批判与适度陌生化的跨文化交际价值

Techno-Utopian Imagination and the Dilemma of AI-Powered Translation—A Critique of Excessive Adaptation and the Cross-Cultural Communicative Value of Moderate Defamiliarization

摘要: 生成式智能翻译的普及催生了大众对跨语言沟通的技术乌托邦想象，译文流畅顺滑成为公众评判翻译质量的核心标尺，过度适配不断挤压翻译陌生化的生存空间，引发非文学文本跨文化阅读扁平化、同质化问题。现有研究多在文学翻译场景下讨论归化与陌生化，较少从社会科技想象的视角剖析智能翻译内部的价值矛盾。本文依托社会科技想象理论，界定翻译适配与翻译陌生化的概念边界，辨析适度陌生化与翻译讹误的本质区别，揭示“无限适配”幻象经由大众审美、算法反馈双向建构的生成机制，重估适度陌生化在认知提示、差异保存、反思激发层面的跨文化交际价值。研究认为，适配与陌生化并非二元对立，适配承担信息通达的工具价值，适度陌生化承载深度跨文化对话的认知价值。据此从观念澄清、实践策略与评价体系等维度提出二者动态平衡的实现路径，以期完善智能时代翻译价值评判体系。

Abstract: The popularity of generative AI-powered translation has given rise to public Techno-utopian Imagination of cross-linguistic communication. Fluency has become the dominant criterion for translation quality assessment. Excessive adaptation continuously squeezes the room for translational defamiliarization and leads to flattened and homogenized cross-cultural reading of non-literary texts. Existing studies mostly discuss domestication and defamiliarization within literary translation, yet few analyze value contradictions in AI translation from the perspective of Sociotechnical Imaginaries. Drawing on Sociotechnical Imagination Theory, this paper defines the conceptual boundaries of Translational Adaptation and Translational Defamiliarization, and distinguishes moderate defamiliarization from translational errors. It reveals how the fantasy of “unlimited adaptation” is co-constructed by public aesthetic preferences and algorithmic feedback, and re-evaluates the cross-cultural communicative values of moderate defamiliarization in cognitive cueing, diversity preservation and reflective stimulation. This study argues that adaptation and defamiliarization are not binary opposites: adaptation serves the instrumental value of information accessibility, while moderate defamiliarization realizes the cognitive value of in-depth cross-cultural dialogue. Pathways toward their dynamic balance are proposed from the dimensions of conceptual disenchantment, practical strategies and evaluation systems, so as to improve the translation value evaluation system for the AI-driven era.