Online learning with a ChatGPT-assisted revision approach to EFL graph writing

Students’ performance and perceptions

Bibliographic Data

ID22428201
AuthorsShu-Chiao Tsai (University of Science and Technology, corresponding author), S C Tsaï (0000-0002-0541-5593, National Kaohsiung University of Science and Technology)
Year2026
Pages1-24
Publication date2026-03-20
Peer ReviewedYes
Open AccessNo
TypeARTICLE
VenueInnovation in Language Learning and Teaching (JOURNAL)
Journal identifiersISSN: 1750-1229 • E-ISSN: 1750-1237
PublisherInforma UK Limited (PUBLISHER • GB)
DOI10.1080/17501229.2026.2645994
OpenAlexW7139135577
LanguageEN
References cited54

Purpose: This study focused on investigating the performance and perceptions of 139 non-English major EFL students completing graph writing with a five-step technology-enhanced learning performance (TELP) mode of the ChatGPT-assisted revision approach.Design/methodology/approach: The study adopted a pretest and post-test quasi-experimental design. A five-step TELP mode based on Chapelle’s (1998) principles for the development of multimedia CALL was implemented to create an online learning environment. Its implementation was conducted in two phases: helping students revise and evaluate their pre-writing scripts and directly revising their post-writing scripts. Students’ performance was evaluated by three types of free online computational assessments. A questionnaire on Google Forms was administered to elicit their experience and perceptions.Findings: Students with different English proficiencies employed different ways of using AI-based tools to complete the pre-writing; they also had different pre-writing outcomes. The linguistic features that ChatGPT generated on students’ pre-writing were classified into 12 categories, and their frequency was significantly correlated with students’ pre-writing performance evaluated by the other two computational assessments in this study. Students with low English proficiency adopted the strategy of L1 to L2 translation through using AI-based tools to achieve better pre-writing performance than high-proficiency students who did not use AI-based tools. Students’ performance for revision did not vary with their English proficiency and was significantly improved in delivering more professional vocabulary, expressing more lexical words, showing higher writing scores, and having a lower probability of writing errors, with effect sizes ranging from 0.14 to 0.82. Students thought that the ChatGPT-assisted revision approach on graph writing was convenient, rapid, and easy, and its feedback was rich, clear, well explained, easy to understand, and of good quality.Originality/value: This study contributes to the growing body of research on learning with the implementation of ChatGPT. Based on Chapelle’s suggested principles for the development of multimedia CALL (1998), the implementation of a five-step TELP mode of the ChatGPT-assisted revision approach in graph writing was proposed. The measurement results of evaluating students’ multiple writing features using different technological assessment tools were consistent. Implications for the implementation of ChatGPT in language learning and future studies are discussed.

Graph · Language acquisition · Online learning · Perception · Artificial Intelligence in Healthcare and Education · Second Language Acquisition and Learning · Text Readability and Simplification

Tools

Open DOI