Neurocognitive effects of multimodal feedback in ai-supported writing: evidence from working memory measures

Abstract

Background: Artificial intelligence (AI)-supported writing tools have become increasingly prevalent in language education, yet little is known about their neurocognitive effects. In particular, whether multimodal feedback (textual, visual, and auditory) enhances or hinders learning by influencing working memory load remains underexplored.

Aim: This study aims to investigate the neurocognitive effects of multimodal feedback in AI-supported writing, with a specific focus on its influence on working memory load. It seeks to determine whether integrating textual, visual, and auditory feedback enhances EFL students’ writing performance measured in terms of accuracy, fluency, and lexical complexity while simultaneously reducing cognitive demands as evidenced by working memory performance indicators

Method: This quantitative study employed a quasi-experimental design with 80 EFL university students randomly assigned to two groups: (a) a control group receiving text-only AI feedback and (b) an experimental group receiving multimodal AI feedback. Data were collected through pre- and post-writing tasks evaluated with an analytic rubric (accuracy, fluency, lexical complexity) and working memory tests (digit span, 2-back, and reaction time). Statistical analyses included ANCOVA and effect size estimation (Cohen’s d).

Results: Findings revealed that the experimental group significantly outperformed the control group in writing accuracy (+6.8 points) and lexical complexity (+0.42 type-token ratio). Working memory tests showed lower cognitive load in the experimental group, as evidenced by improved n-back performance (p < .01) and faster reaction times. Cohen’s d indicated medium-to-large effects across both linguistic and neurocognitive measures.

Conclusion: The results suggest that multimodal AI feedback supports writing development while reducing cognitive load, providing neurocognitive evidence that adaptive feedback can optimize both performance and efficiency. This study contributes to bridging applied linguistics, cognitive psychology, and AI research, highlighting the need for cognitively sustainable AI-assisted language learning tools.