Synthesizing eye-tracking research in L2 listening assessment: a human-AI collaborative systematic review

Eye-tracking technology has been increasingly used to investigate the complex cognitive processes of L2 learners during listening assessments. Despite this growing body of research, the field still lacks a comprehensive synthesis of its diverse research foci, methodologies, and technical specifications. This study presents a systematic review of eye-tracking research in L2 listening assessment, conducted using a human-AI collaborative methodological approach. Following the SMART and PRISMA frameworks, a total of 19 studies were retrieved and coded for six dimensions: (1) study characteristics, (2) research themes, (3) technical specifications, (4) methodological characteristics, (5) listening tests, and (6) eye-tracking metrics and inferred cognitive mechanisms. The analysis identified six major research themes: multimodal listening, strategy use, test item design, test-taker factors, test method effects, and test equity. Across studies, we found notable variability in technical specifications, methodological characteristics, and listening test features, as well as inadequate reporting of critical technical and methodological parameters. Fixation- and visit-related eye-tracking metrics were most commonly used and were drawn on to infer three key cognitive mechanisms: visual attention allocation, cognitive load, and strategy use. We also demonstrated how a custom ChatGPT, configured with a predefined coding scheme, a coded example, and task instructions, was developed and used in collaboration with human researchers to streamline the laborious data-coding process. Overall, this study offers a comprehensive account of the current landscape of eye-tracking research in L2 listening assessment and provides methodological insights for GenAI-assisted research synthesis.