Exploring the impact of diverse feedback sources on learners’ performance, motivation, and preference in a translation course: Tutor, peer, and GPT insight

Background Feedback, both human and computer-assisted, plays a critical role in scaffolding learners’ development, especially within the Zone of Proximal Development (ZPD) through interaction with more knowledgeable others. While recent studies have examined the role of Generative AI (GenAI) in supporting second language learning, its potential in translation pedagogy remains underexplored. Objectives This paper examines how GenAI feedback influences learners’ performance, motivation, and preferences in an ESL translation course. Based on the sociocultural theory, it emphasizes that all feedback sources should meet learners’ developmental needs. Methods This study adopted a mixed-methods quasi-experimental design, incorporating pre- and post-tests, a composite questionnaire, and semi-structured interviews. A total of 84 undergraduate students were randomly assigned to three ESL translation classes. At the initial stage, all participants received tutor feedback alongside unguided peer feedback. In the subsequent intervention phase, the Control Group (CG) continued with unguided peer feedback, the Experimental Group 1 (EG1) shifted to guided peer feedback, while the Experimental Group 2 (EG2) received AI-generated feedback in place of peer feedback. Tutor feedback remained consistent across all groups. This design facilitated both within-group and between-group comparisons, allowing the study to isolate the effects of feedback structure (guided vs. unguided peer feedback) and feedback source (peer/tutor vs. GenAI feedback). Results and conclusion The study found that guided peer feedback (EG1) led to the highest performance, significantly outperforming both unguided peer feedback (CG) and GenAI feedback (EG2). Although EG2’s performance lagged behind EG1, its learners showed the greatest increase in motivation. Feedback preferences ranked as: tutor > guided peer > GenAI > unguided peer. Learners appreciated GenAI’s accuracy, but noted issues such as limited creativity and emotional connection. The study suggests that human feedback, particularly guided peer and tutor feedback, provided stronger emotional support and adaptive scaffolding, fostering deeper cognitive engagement and a more empathetic environment compared to GenAI, which is correlated with higher performance and preference gains. While qualitative responses praised GenAI for accurate and comprehensive feedback, learners identified challenges such as information overload, limited creativity, and lack of emotional connection. The study suggests a hybrid feedback strategy integrating GenAI with human interactions to enhance translation pedagogy.