AI-Assisted Subtitling in Translator Training: Efficiency, Quality, and Student Perceptions in an Experimental Study

Abstract

Digitalization continues to reshape translation practices across domains, with audiovisual translation (AVT) being one of the fields most affected by automation and artificial intelligence (AI). This study investigates the pedagogical implications of AI-assisted subtitling in the context of translator education. Twelve undergraduate students from the Faculty of Philology at Akhmet Yasawi University (Turkistan, Kazakhstan) participated online in a pre–post experimental design. In the pre-test, students manually subtitled a one-minute English promotional clip; in the post-test, they used AI-generated subtitles which they post-edited for accuracy, style, and cultural appropriateness. Data were collected through task completion times, rubric-based quality scores, error analysis, and student reflections.

The results indicate substantial improvements in both efficiency and quality. Mean task completion time decreased from 745 seconds in the manual condition to 451 seconds in the AI-assisted condition, representing a 40% reduction (t(11) = 17.79, p < .001, Cohen’s d = 5.13). Quality scores improved significantly from 5.08/8 to 7.08/8 (t(11) = –8.12, p < .001, d = 2.35). Error analysis revealed a clear shift: manual subtitling produced numerous technical and segmentation errors, while AI subtitling largely eliminated these but required human correction of semantic and cultural nuances. Thematic analysis of reflections confirmed these trends: students valued the speed and technical precision of AI, but emphasized their indispensable role in ensuring idiomatic and culturally sensitive translations.

This study contributes to the growing body of literature on the hybrid ecology of AVT, where human and non-human actors collaborate within translation workflows. It suggests that AI-assisted subtitling can be fruitfully integrated into translator training as both a productivity aid and a pedagogical resource for critical reflection. At the same time, the findings highlight the limitations of automation and the continued necessity of human agency in audiovisual translation.

Keywords: audiovisual translation, AI-assisted subtitling, translator training, quality, student perceptions.