Logo

Authors

JUNE, LEE EUI

Issue Date

2026-10

Publisher

Elsevier Ltd

Citation

논문 Computers and Education, v.251, no., pp.-

Journal Title

Computers and Education

Volume

251

DOI

10.1016/j.compedu.2026.105653

ISSN

0360-1315

Abstract

Implementing personalized learning in traditional classrooms, where one teacher manages many students, is challenging. However, advancements in AI technology have made it increasingly feasible to curate learning materials tailored to learners’ preferences and proficiency levels. This study aims to develop 2xAI, an AI-powered L2 reading system embedded with a diffusion AI model that generates personalized learning materials for middle school English learners in Korea. To ensure pedagogical validity, we analyzed the linguistic quality of the AI-generated texts using computational (Coh-Metrix) and expert evaluations. Coh-Metrix focused on ten key indices of lexical sophistication, syntactic complexity, and text easability, compared to middle school English textbooks. Thirteen experienced teachers and textbook authors then evaluated the language accuracy, content relevance, and pedagogical appropriateness of the AI-generated texts. The results showed that intermediate-level AI-generated passages closely matched the linguistic profiles of textbook passages. Expert ratings indicated comparable quality and contextual suitability. The system was pilot-tested in middle school English classrooms (N = 57). Log data showed students’ preferences for AI-generated texts and their learning progress. Survey results indicated high satisfaction with and perceived usefulness of the AI-assisted reading experience. Overall, these findings suggest that the AI-powered L2 reading system like 2xAI can effectively facilitate personalized, adaptive English reading instruction in formal educational settings. © 2026