Human–AI Collaborative Feedback and L2 Writing Performance: The Mediating Effects of Task Motivation, Cognitive Load, and Revision Behavior

While generative artificial intelligence (GAI) is increasingly integrated into second language (L2) writing, the cognitive, affective, and behavioral mechanisms underlying the effectiveness of human–AI collaborative feedback remain insufficiently understood. The present study aims to investigate the impact of human-AI collaborative feedback on L2 writing performance and to verify the mediating effects of task motivation, cognitive load, and revision behavior. A quasi-experimental design was adopted with 66 Chinese EFL learners who completed an AI-assisted writing task. Data were triangulated from self-reported questionnaires, corpus-based automated linguistic indices (syntactic and lexical complexity), and manual coding of textual revision behaviors. The results indicated that human–AI collaborative feedback exerted significant positive direct effects on both syntactic and lexical complexity. Task motivation significantly predicted syntactic complexity but not lexical complexity, suggesting an asymmetric influence across different dimensions of language production. Furthermore, task motivation and revision behavior significantly mediate human–AI collaborative feedback, as well as both syntactic and lexical complexity. The findings were expected to provide empirical implications for constructing a learner-centered human-AI collaborative writing pedagogy and optimizing the instructional integration of generative AI feedback.