Sociocultural Mediation Through Large Language Models: Enhancing Learner Agency and Identity in Chinese EFL Speaking Contexts

Sociocultural Mediation Through Large Language Models: Enhancing Learner Agency and Identity in Chinese EFL Speaking Contexts This dataset supports a mixed-methods study investigating the role of large language models (LLMs) in mediating learner agency and identity construction in Chinese EFL speaking contexts, grounded in Sociocultural Theory (SCT). Study Details- Intervention: 8-week LLM-based speaking practice (DeepSeek/Kimi Chat) delivered via WeChat to 55 intermediate Chinese EFL university students (CEFR B1-B2).- Data Types: 1. Qualitative: Anonymized interview transcripts (n=12) and human-curated thematic coding sheets. 2. Quantitative: Online Student Engagement Scale (OSE; n=28), Voice Recording Task (VRT) proficiency scores (pre/post/inter-rater reliability; n=12), and weekly comfort/app usage feedback (n=55). 3. Visualizations: Final high-resolution figures of thematic frequencies, correlation matrices, and proficiency gains.- Ethical Compliance: Anonymized data with ethical approval from the College of Foreign Languages and Cultures,University A (Approval Number: CFLC_2025010_00X). All participants provided informed consent for data sharing. Dataset StructureThe dataset is organized into 3 core folders for transparency and replicability:1. Raw Data: Original, unprocessed data collected during the study.2. Processed Data: Human-curated cleaned, coded, and merged data (including statistical appendices).3. Figures: Final figures used in the study’s results section. ReplicabilityThis dataset enables full replication of the study’s mixed-methods findings. No automated scripts are required—all steps can be replicated using standard office software (Microsoft Excel, Word) and statistical tools (SPSS, R, Python) for quantitative analysis. A detailed README.txt file provides step-by-step guidance for replication. ## LicenseCC BY 4.0 (Attribution 4.0 International)