A Study on Generative AI-Assisted Adaptation Strategies for High School English Reading Texts
TL;DR
The findings reveal that both experts and ChatGPT rely most heavily on content-level adaptation, while experts engage least in syntactic adaptation and ChatGPT in lexical adaptation, and the importance of teacher–AI collaboration in text adaptation is highlighted.
Abstract
: With the increasing use of authentic materials in high school English reading instruction, adapting such texts to match students’ proficiency levels has become both essential and challenging. This study investigates the potential of ChatGPT in English reading text adaptation by comparing ChatGPT-adapted texts with expert-adapted texts based on ten sets of Gaokao reading materials (2022–2024). An analytical framework comprising four adaptation dimensions—lexical, syntactic, content, and cohesion—and 19 specific strategies was employed. The findings reveal that both experts and ChatGPT rely most heavily on content-level adaptation, while experts engage least in syntactic adaptation and ChatGPT in lexical adaptation. At the strategy level, experts tend to prioritize information omission and high-and low-frequency word adjustment, whereas ChatGPT shows a preference for explanation addition, paraphrasing, and reordering of information. ChatGPT demonstrates clear advantages in rapidly identifying grammatical errors and enhancing syntactic diversity; however, it also exhibits limitations, including insufficient awareness of test design conventions, over-adaptation of linguistic style, weak cohesion, and inadequate evaluation of informational relevance. These findings highlight the importance of teacher–AI collaboration in text adaptation. Teachers should assume the roles of leaders, evaluators, and optimizers to guide, assess, and refine AI-generated texts. This study contributes to understanding the application of generative AI in language education and provides practical implications for the development of reading materials.