A Comparative Study of Chatbot-Based Feedback and Traditional Teacher Feedback in Correcting Focused Grammatical Errors of ESL University Students

Abstract

This research examines the comparative efficacy of chatbot-generated feedback through Replika and conventional teacher feedback in the correction of targeted grammatical errors among Pakistani ESL university students. Taking a mixed-method, quasi-experimental approach, the study was carried out for five weeks on 60 volunteers who were randomly assigned to experimental and control groups. Grammatical mistakes in students' writing were analyzed and classified with Ferris's (2006) taxonomy, including both treatable and untreatable error categories. Although error categories were all reviewed, verb tense/form and prepositions were the most common in the pretest and were chosen for specific feedback. SPSS (version 26) was used to analyze the quantitative data, and semi-structured interviews were analyzed using Braun and Clarke's (2006) thematic analysis. The research was conceptualized on the basis of Activity Theory (Engeström, 1987), which positioned the application of tools; AI-driven feedback through Replika and conventional written corrective feedback as teachers of the learning process in the classroom context. Findings indicated that the control group receiving teacher feedback performed significantly better than the experimental group in reducing grammatical errors. Nonetheless, students in the experiment group indicated positive attitudes towards AI feedback, enumerating its immediacy, clarity, and convenience despite some technical constraints. The study indicates that although teacher feedback is still better in enhancing grammatical correctness, a combination of AI-based tools with conventional approaches could yield a wider and more effective feedback experience for ESL learners.