The impact of large language models on collaborative learning processes and outcomes in higher education for English translation.

Collaborative learning is an important teaching method in English translation in Chinese higher education and is affected by the development of Large Language Models (LLMs). Previous studies mainly focus on the application of LLMs by individual students, but there is a lack of information about the influence of these models on the group dynamics when they are used in collaborative work. To fill this gap, our research examines the effect of an LLM, as a third participant, on the collaborative learning procedures and results. A comparative experimental design is conducted, involving 372 undergraduate and graduate students majoring in English in China, who are randomly divided into two groups: a control group working together without the LLM and an experimental group where the LLM is included in a human-human-AI collaboration team. The translation quality, questionnaire data and process records of both groups are analyzed by a mixed-method approach to make a comparison. The findings show that there is no significant difference in the final translation quality between the two groups, but the experimental group reports a significantly higher perceived efficiency and task satisfaction. A qualitative analysis clarifies this result, indicating that the participation of the LLM changes the collaborative procedure. This change involves a transition from a generative to an evaluative negotiation in the decision-making process, and also leads to the formation of informal roles. The study concludes that in collaborative learning, LLMs affect the process rather than enhance the content directly. These findings are of great importance for the design of effective human-ai English translation teaching in the Chinese higher education environment.