생성형 AI의 어휘문법적 선택 능력에 대한 문법교육적 연구 - 양태 표현을 중심으로

A Study on Generative AI’s Ability to Make Lexicogrammatical Choices from the Perspective of Grammar Education - Focusing on Modality

초록

This study assessed generative AI’s ability to make lexicogrammatical choices in modality and explored implications for Korean grammar education. Drawing on SFL, we designed two diagnostic tasks—generating modalities and correcting inappropriate ones—and applied them to seven versions in total across ChatGPT and Gemini. We examined whether the outputs met the required modality type, value, and positioning and whether their metalinguistic explanations were accurate. AI generally produced context-appropriate modality and corrections for obvious errors. However, while generating appropriate expressions, the AI exhibited metalinguistic errors, such as confusing congruent and metaphorical realization and oversimplifying subjectification/objectification and speaker/hearer positioning. It repeated conventional correction patterns and, in some cases, altered the proposition itself rather than merely adjusting modality. Based on these findings, this study proposes grammar education content to critically evaluate and revise AI outputs in line with their intended meanings.

키워드

생성형 AI; 양태 표현; 어휘문법적 선택; 체계기능언어학; 대인적 대기능; Generative AI; Modality; Lexicogrammatical choice; Systemic functional linguistics; Interpersonal metafunction
제목
생성형 AI의 어휘문법적 선택 능력에 대한 문법교육적 연구 - 양태 표현을 중심으로
제목 (타언어)
A Study on Generative AI’s Ability to Make Lexicogrammatical Choices from the Perspective of Grammar Education - Focusing on Modality
저자
신희성
DOI
10.24008/klle.2026..56.004
발행일
2026-08
유형
Y
저널명
한국어문교육
호
56
페이지
117 ~ 155