Complementing but Not Replacing: Comparing the Impacts of GPT-4 and Native-Speaker Interaction on Chinese L2 Writing Outcomes DOI Creative Commons
Zhengang Shan, Zefeng Song, Jiang Xu

et al.

Behavioral Sciences, Journal Year: 2025, Volume and Issue: 15(4), P. 540 - 540

Published: April 17, 2025

This study explored the efficacy of large language models (LLMs), namely GPT-4, in supporting second (L2) writing comparison with interaction a human partner pre-writing phase. A within-subject behavioral experiment was conducted 23 Chinese L2 learners who were exposed to three conditions: “without interaction”, “interaction GPT-4”, and partner”. They then completed an task. It found that yielded significantly improved results compared both GPT-4 case without terms overall scores, organization, language. Additionally, types enhanced participants’ topic familiarity confidence reduced task’s perceived difficulty interaction. Interestingly, GPT-4” condition, positively correlated better outcomes, whereas partner” content scores; however, scores negatively associated confidence. suggests LLMs should be used complement not replace partners

Language: Английский

Complementing but Not Replacing: Comparing the Impacts of GPT-4 and Native-Speaker Interaction on Chinese L2 Writing Outcomes DOI Creative Commons
Zhengang Shan, Zefeng Song, Jiang Xu

et al.

Behavioral Sciences, Journal Year: 2025, Volume and Issue: 15(4), P. 540 - 540

Published: April 17, 2025

This study explored the efficacy of large language models (LLMs), namely GPT-4, in supporting second (L2) writing comparison with interaction a human partner pre-writing phase. A within-subject behavioral experiment was conducted 23 Chinese L2 learners who were exposed to three conditions: “without interaction”, “interaction GPT-4”, and partner”. They then completed an task. It found that yielded significantly improved results compared both GPT-4 case without terms overall scores, organization, language. Additionally, types enhanced participants’ topic familiarity confidence reduced task’s perceived difficulty interaction. Interestingly, GPT-4” condition, positively correlated better outcomes, whereas partner” content scores; however, scores negatively associated confidence. suggests LLMs should be used complement not replace partners

Language: Английский

Citations

0