Working Memory, Attention Control, and Vocabulary Retention in AI (ChatGPT)-Assisted Foreign Language Learning: A Structural Cognitive Modelling Approach

Abstract

This study examined how working memory, attention control, and frequency of ChatGPT-4 use are structurally associated with vocabulary retention in foreign language learning. A quantitative cross-sectional survey design was employed, with data collected from 1002 EFL learners via stratified random sampling. Validated self-report instruments measured working memory, attention control, frequency of ChatGPT use, and vocabulary retention (immediate recall, delayed retention, semantic integration, and productive use). Structural equation modelling was used to test the proposed model. The results showed that working memory was strongly associated with attention control and exerted a direct effect on vocabulary retention across all dimensions. Attention control explained a substantial share of the relationship between working memory and retention, indicating that regulatory allocation of attention, rather than memory capacity alone, governs whether lexical information is stabilised during ChatGPT-assisted learning. The frequency of ChatGPT use conditioned these cognitive pathways by strengthening links between working memory and attention control, and between attention control and vocabulary retention, at higher levels of engagement. Frequency did not predict retention independently, indicating that repeated use supports learning only to the extent that it reinforces cognitive regulation rather than increasing exposure. Vocabulary learning with AI relies more on cognitive regulation and engagement than exposure.

Ähnliche Arbeiten