ChatGPT ‐Supported Literary Translation: Tense Accuracy and Dynamic Emotion Profiles in the Language Classroom

ABSTRACT Background Generative artificial intelligence can facilitate task completion, but whether it improves learning or consistently shapes task‐related emotions remains unclear. Objectives This mixed‐methods classroom study compared literary translation with ChatGPT access, translation with dictionary access and optional teacher assistance and regular instruction, focusing on tense accuracy and task‐related emotional experiences. Methods The study involved 42 English‐major undergraduates. All participants completed a pretest and a posttest of English tense use. The 29 students in the translation conditions completed six translation tasks and provided 556 repeated emotion‐bar records with concurrent written explanations; 10 of these students also participated in interviews. Robust analysis of covariance, within‐participant models, exploratory configurational measures and thematic analysis were used. Results Mean scores increased in all three conditions, but adjusted posttest scores and gain scores did not differ significantly among them. No statistically significant between‐condition differences were detected in individual‐emotion percentages or configurational measures. None of the six emotion percentages varied significantly across task sessions, although curiosity was lower at the fourth than at the first measurement point. Profile mixedness was lower in Task 5 (Session 6) than in Task 2 (Session 3). Qualitative findings linked emotions mainly to text content and difficulty, progress, time pressure and classroom conditions. ChatGPT was valued for its speed and immediacy but raised concerns about reliability, effort and ownership. Conclusion The study did not detect a statistically significant advantage for ChatGPT support in tense accuracy or emotional experience. The exploratory configurational analysis nevertheless identified a task‐session difference not apparent in the separate emotion percentages, illustrating the value of examining learning outcomes, repeated emotion reports and students' appraisals together when evaluating AI‐supported language tasks.