AI-Generated Captions as a Memory Support Tool in Video-Mediated Recall Tasks

Articles in Press

Document Type : Research Paper

Author

Department of Literature and Foreign Languages, Faculty of English Language, Payame Noor University, Iran

10.22111/ijals.2026.51519.2511

Abstract

This study investigates the use of AI-generated captions as a memory support tool for interpreting, focusing on their impact on grammatical and lexical errors in a video-mediated recall task (i.e., participants watched a video and then verbally translated the content without access to the source text). Thirty-four post-graduate translation students were divided into an experimental group exposed to AI-generated English captions and a control group exposed to video with visuals only. Participants performed a bidirectional recall-based translation task (Persian-to-English and English-to-Persian). Results indicated that the experimental group exhibited significantly fewer grammatical errors (p = .03, d = 0.65) and lexical errors (p = .04). The reduction in grammatical errors ranged from 1.1 to 1.3 points depending on condition, while lexical error reduction ranged from 0.7 to 1.0 points. Both groups performed better at slower speeds, and translation direction affected output quality. The interaction between group and speech speed was not statistically significant, suggesting that caption benefits did not differ by speed in this sample. These findings demonstrate that AI captioning can serve as an effective supplementary tool for interpreters. Future research should isolate the effects of visual input and employ higher-accuracy captions (≥95%).

Keywords

Iranian Journal of Applied Language Studies

Articles in Press, Accepted Manuscript
Available Online from 18 July 2026

Files

Share

How to cite

Statistics