ChatGPT在文学翻译中“创造性叛逆”及其成因——以《杀死一只知更鸟》为例

Creative Treason of ChatGPT in Literary Translation and Its Causes—A Case Study of “To Kill a Mockingbird”

摘要: 本文以“创造性叛逆”理论为支撑，选取《杀死一只知更鸟》为研究文本，通过对比ChatGPT (GPT-4)译文与权威人工译本，系统剖析ChatGPT在文学翻译中“创造性叛逆”的三大表现维度：语言层面的句式重构与语气调适、文化层面的负载词适配与意象转换、叙事层面的节奏优化与视角微调。研究发现，ChatGPT的创造性叛逆源于“语料训练偏差”“文学性理解局限”“目标语流畅性优先”的算法逻辑，其叛逆表现兼具积极价值(提升阅读流畅度、降低文化隔阂)与局限性(弱化文学个性、消解文化独特性)。本文结论可为AI文学翻译的优化方向提供参考，丰富创造性叛逆理论在机器翻译领域的应用研究。

Abstract: Supported by the theory of “creative treason,” this paper selects “To Kill a Mockingbird” as the research text. By comparing the translations of ChatGPT (GPT-4) with authoritative human translations, the study systematically analyzes three primary dimensions of creative treason in ChatGPT’s literary translation: syntactic reconstruction and tone adjustment at the linguistic level; adaptation of culture-loaded words and image transformation at the cultural level; and rhythm optimization and perspective fine-tuning at the narrative level. The findings reveal that ChatGPT’s creative treason originates from “corpus training biases,” “limitations in understanding literariness,” and an algorithmic logic that “prioritizes target language fluency.” This treason manifests both positive value (enhancing reading fluency and reducing cultural barriers) and limitations (diluting literary individuality and dissolving cultural uniqueness). The conclusions of this paper offer a reference for the optimization of AI literary translation and enrich the application of creative treason theory within the domain of machine translation.