This study examines how the generative AI system ChatGPT reads poetry and constructs meaning, and aims to clarify the literary and critical implications of its interpretive processes. While recent advances in artificial intelligence have stimulated ac...
This study examines how the generative AI system ChatGPT reads poetry and constructs meaning, and aims to clarify the literary and critical implications of its interpretive processes. While recent advances in artificial intelligence have stimulated active discussion in the field of poetic creation, research that systematically analyzes AI’s interpretive mechanisms and limitations in the realm of poetic reading and appreciation remains comparatively scarce. Focusing on four poems-Kim Sowol’s “Mother, Sister”, Baek Seok’s “The Lark’s Song”, Yi Yuksa’s “Wilderness”, and Jang Seoknam’s “Narrow Spring”-this study analyzes the kinds of source data and logical structures that underlie ChatGPT’s readings, as well as the errors and distortions that emerge in the process.
The results show that ChatGPT reproduces established interpretations more accurately and provides more coherent analyses when it encounters works with high public familiarity and substantial prior scholarship. In contrast, for poems with limited critical tradition, the AI exhibits a tendency toward “information-combination interpretation,” reorganizing meaning around external cues introduced during prompting; moreover, hallucinations-such as fabricating nonexistent textual variants-occurred repeatedly. Compared to human criticism, which explores cultural context, emotional depth, and textual indeterminacy, ChatGPT’s interpretations are clear and consistent yet marked by the uniformity characteristic of machine-generated discourse.
Furthermore, when compared with the cultural context, emotional depth, and textural richness evident in human literary criticism, ChatGPT’s interpretations, though stable, cannot avoid the drawback of a uniformity akin to that of a mechanical fabric. Through this analysis, the present study reveals both the potential of employing artificial intelligence in the reading of poetry and its structural limitations. This, in turn, offers implications for how AI may be utilized and supplemented in literary interpretation.