Comparing AI chatbots and web-based synchronous collaborative writing tools, leveraging group cohesion, motivation, and performance.
TL;DR
From a Group Cohesion Theory (GCT) perspective, AI appeared to strengthen task cohesion more than social cohesion, improving coordination and performance, though it did not always deepen peer interaction or shared meaning-making.
Abstract
This study examines the effects of generative AI (GenAI) chatbots versus traditional web-based resources (TWRs) on group cohesion, motivation, and writing performance during face-to-face synchronous collaborative writing (SCW) tasks among L2 learners. Despite growing adoption of AI in language education, key gaps remain: limited psychological research on AI-supported collaborative writing, scarce attention to motivation and self-efficacy in group contexts, and absence of direct comparisons between AI chatbots and conventional tools in real-time, in-class settings. Sixty-three English-major sophomore students at a Chinese university were randomly assigned to two classes: an experimental cohort (n = 33) using Kimi AI chatbot and a control cohort (n = 30) using TWRs (e.g., dictionaries, search engines). Both groups worked in teams of 5-6 on a shared cloud document to revise a 500-word descriptive essay. Mixed-methods data included screen recordings, writing assessments, typing speed measures, pre/post motivation, perception questionnaires, and individual notes on team cohesion. Quantitative results showed the AI group completed writing and proofreading significantly faster (EWT: M = 24:20 vs. 41:10; EPT: M = 9:33 vs. 15:25), achieved higher essay scores (M = 90.28 vs. 81.11, p = .001, d = 2.43), and reported substantially greater post-task motivation (ANCOVA, p < .001, η2 = 0.888). Qualitative analysis of the cohesion notes suggested that AI helped streamline communication, reduce information overload, and support engagement, but it also introduced challenges, as its rapid and uniform responses could reduce discussion, limit exploration of alternatives, and encourage over-reliance on the tool. From a Group Cohesion Theory (GCT) perspective, AI appeared to strengthen task cohesion more than social cohesion, improving coordination and performance, though it did not always deepen peer interaction or shared meaning-making.