AI와 인간 번역의 문체 재현과 독자 반응 연구: 추리소설 『그리고 아무도 없었다』사례를 중심으로

Abstract

Building on Kim Soonyoung's (2025) analysis, this study examines the extent to which the stylistic strategies employed by human translators can be reproduced by an AI system, and how differences between human and AI styles are perceived by readers. Focusing on three key scenes in Agatha Christie’s And Then There Were None where stylistic devices are most pronounced, the research compares two Korean human translations with a ChatGPT-generated version. It employs two research methods—a reader survey and expert interviews—through which the findings are cross-validated. The main results are as follows: (1) Sentence structure and rhythm were the most favored in the AI translation (52.3%), suggesting that its conciseness and consistency, rooted in statistical learning, enhance readability and convey a sense of urgency. (2) Punctuation and visual pacing were also rated highest in the AI translation (59.3%), indicating that regular and predictable punctuation supports a stable reading flow. (3) Repetition and emphasis still favored the AI translation; however, the human translation remained relatively competitive at 27.9%, demonstrating its ability to capture emotional nuances and aesthetic effects. Experts acknowledge that while AI can effectively reproduce the short breath and rhythm of mystery novels, it struggles to capture subtle emotional nuances, the aesthetic work of repetition, and cohesive discourse. By combining stylistic analysis with reader responses and interview-based expert insights, the study offers nuanced perspectives on the potential and limitations of AI in literary translation.