Artificial Intelligence or Peer Discussion? Investigating Learning Outcomes in Medical Translation Tasks

With the growing integration of artificial intelligence (AI) in translation education, understanding its influence on learning outcomes has become increasingly important. This mixed-methods study examined the effects of different types of translation tasks on participants’ ability to learn and remember terminology. Participants consisted of 48 undergraduate students taking a medical translation course. Half collaborated in traditional group tasks, while the rest employed the large language model (LLM) tools DeepL and ChatGPT. Progress was measured quantitatively by having each participant take pre-, immediate post-, and delayed post-tests. In addition, semi-structured interviews were conducted with six of the participants to gather qualitative data. Participants showed significant improvement on the initial post-test regardless of the group they were assigned to. On the delayed post-test, however, the participants doing group work maintained this improvement, while the participants using LLM tools suffered a significant drop in test scores. These results indicated that either strategy could work in the short run, but using LLM tools was not an effective long-term strategy. This aligned with the interviews, in which group work participants highlighted the value of discussing tasks and supporting each other, in contrast to those using LLM tools, who pointed out an absence of peer interaction, despite how convenient they found the tools to be. These findings suggest that group work in a medical translation course could help students acquire and maintain new vocabulary, although using LLM tools could have short-term benefits as well.