AI-assisted Post-editing and the Development of Critical Technology Awareness:A Study of Chinese-Japanese Political Text Translation Pedagogy

Abstract

Generative artificial intelligence and neural machine translation are changing the conditions under which translation is taught and practiced. For graduate students in translation, the challenge is no longer simply how to use machine translation and large language models, but how to identify, explain, and revise semantic shifts, terminological inconsistency, and register problems in AI-generated output. This study reports a one-group pre-test/post-test quasi-experiment conducted with 15 first-year MTI students in a Japanese translation program at a Chinese university. The instructional intervention focused on Chinese-Japanese post-editing of political texts. Three recent Chinese speeches from United Nations Security Council contexts were used as parallel materials: one for the pre-test, one for the post-test, and one for classroom training. The study drew on anonymized three-rater scoring, MQM-informed error annotation, an AI critical literacy questionnaire, post-editing logs, and student reflections. Wilcoxon signed-rank tests showed significant gains in overall post-editing quality, machine-translation error identification, and AI critical literacy. The strongest improvements appeared in revision rationale and evidence use, terminology and proper-name handling, and political-diplomatic register. Student logs and interview comments further suggest a shift from accepting fluent AI output to checking, justifying, and critically revising it. The paper argues that post-editing pedagogy in the AI era should move beyond tool operation and be designed as a competence loop linking technology use, error diagnosis, discourse reconstruction, and critical reflection.