Learner Agency Priorities and Speaking Performance in GenAI‐Mediated Project‐Based Learning (PBL): Evidence From Q Methodology

ABSTRACT

Project‐based learning (PBL) is widely used in foreign language education to organise learning around sustained collaboration and tangible outcomes, yet little is known about how learners prioritise agency‐related behaviours in PBL speaking tasks when generative artificial intelligence (GenAI) tools are available during project work. To address this gap, this study used Q methodology with 49 Chinese undergraduates in a Chinese university English course to examine two questions: (1) What distinct learner agency priorities emerge in GenAI‐mediated PBL? and (2) How are these priorities associated with differences in learners’ speaking performance? Participants completed Q‐sorts that required them to rank agency‐related statements and by‐person factor analysis was used to identify shared learner agency viewpoints. Persuasive speech ratings were then compared across the resulting viewpoint groups to examine whether different agency‐priority patterns were associated with variation in speaking performance. Three viewpoints were extracted, foregrounding strategy use with GenAI, personal beliefs and motivation and teamwork in GenAI‐mediated project work. Performance differences were most evident between the teamwork‐oriented and strategy‐oriented viewpoints, with learners aligned with the former achieving higher communicative achievement and overall scores. This pattern highlights the importance of collaborative agency in GenAI‐mediated PBL for developing discourse‐level speaking performance, including audience awareness and coherence. The findings clarify how learner agency is patterned in GenAI‐mediated PBL speaking projects and suggest that the relevance of GenAI depends less on tool presence itself than on how its use is organised within collaborative participation. By linking Q‐derived viewpoint profiles to an external performance measure, the study shows more precisely how differences in agency priorities within a shared project design relate to variation in speaking outcomes. Future research could further examine these patterns with process‐based evidence of classroom interaction and GenAI use.