ABSTRACT This experimental study investigates how pairing learners by English proficiency shapes task engagement and text quality in L2 digital collaborative writing (CW). We captured both self‐perceived and observed task engagement in a digital CW task through multiple data sources: a self‐report questionnaire, multimodal interaction logs, and collaborative artifacts. A cohort of thirty‐six undergraduates from a top‐tier Chinese university was grouped into high‐, low‐, and mixed‐proficiency dyads. Using a synchronous writing tool with video conferencing, we collected multimodal interaction logs and artifacts, alongside a post‐task questionnaire on cognitive, social, and emotional engagement. Meanwhile, observed task engagement was measured through multimodal interaction data focusing on language‐related episodes (LREs), idea units, instances of responsiveness, and expressions of enjoyment. Findings revealed discrepancies between self‐perceived and observed task engagement, especially in cognitive and social dimensions. High‐proficiency dyads reported lower cognitive engagement yet produced more LREs than low‐ and mixed‐proficiency dyads. Mixed‐proficiency dyads reported higher social engagement but showed fewer instances of responsiveness than low‐proficiency dyads. Text complexity was significantly lower for mixed‐proficiency dyads than for both high‐ and low‐proficiency dyads.
Wang et al. (Sun,) studied this question.