Generative AI as a Virtual More-Knowledgeable Other: A Mixed-Methods Investigation of DeepSeek's Impact on Grammatical Accuracy and Technology Acceptance in Chinese ESL Writing
Social Reports · 2026
DOI: 10.66638/ep1sq081 ↗Kurz gefasst
Mixed-Methods-Studie in der Zeitschrift Social Reports, 2026. Verfasst von Ke Yang. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 33 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Despite sustained curriculum reforms, intermediate Chinese English-as-a-Second-Language (ESL) undergraduates continue to exhibit a stable grammatical error profile characterized by the mismanagement of articles, tense-aspect systems, and prepositional collocations, attributable largely to persistent L1 transfer. While automated writing evaluation platforms have traditionally targeted surface-level errors, the emergence of domestic large language models—notably DeepSeek—offers a theoretically rich alternative capable of providing context-sensitive, concept-reinforcing feedback. Grounded in Sociocultural Theory, the Process Writing Approach, and the Technology Acceptance Model (TAM), this explanatory-sequential mixed-methods study investigated the pedagogical efficacy and user acceptance of DeepSeek over an eight-week quasi-experimental intervention. Participants comprised 72 intermediate (CEFR B1–B2) English-major undergraduates recruited from Peking, Tsinghua, and Shanghai Universities (n₂₁ = 35 experimental; n₂₂ = 37 control). Analysis of covariance (ANCOVA) revealed that learners receiving DeepSeek-mediated feedback achieved significantly greater post-test reductions in grammatical error density compared to a control group receiving traditional instructor commentary, with a large effect size (Cohen's d ≈ 2.34) that warrants careful interpretive caution given the presence of a task-modality confound. Path analysis confirmed that perceived usefulness and perceived ease of use positively predicted behavioral intention, which subsequently mediated actual system use and grammatical uptake. Thematic analysis of semi-structured interviews revealed three dominant learner profiles—dialogic prompters, transactional processors, and algorithmically over-accommodated users—illuminating how the depth of human-AI interaction moderates learning outcomes. The findings offer actionable benchmarks for curriculum designers and policymakers seeking scalable, firewall-compliant generative AI solutions for L2 writing instruction in the Chinese mainland higher education context.
Zitieren
APA
Yang, K. (2026). Generative AI as a Virtual More-Knowledgeable Other: A Mixed-Methods Investigation of DeepSeek's Impact on Grammatical Accuracy and Technology Acceptance in Chinese ESL Writing. Social Reports. https://doi.org/10.66638/ep1sq081
Vancouver
Yang K. Generative AI as a Virtual More-Knowledgeable Other: A Mixed-Methods Investigation of DeepSeek's Impact on Grammatical Accuracy and Technology Acceptance in Chinese ESL Writing. Social Reports. 2026. doi:10.66638/ep1sq081
BibTeX
@article{Yang2026,
author = {Yang, Ke},
title = {Generative AI as a Virtual More-Knowledgeable Other: A Mixed-Methods Investigation of DeepSeek's Impact on Grammatical Accuracy and Technology Acceptance in Chinese ESL Writing},
journal = {Social Reports},
year = {2026},
doi = {10.66638/ep1sq081}
}
EUVIMED