Study proposes personalized learning and AI partnership to give students greater control over their learning in the age of AI

A new study in ECNU Review of Education examines CHATTING, a ChatGPT‑assisted writing system designed for students with dyslexia. Conducted with 101 Hong Kong secondary students, the research found the tool increased motivation and engagement, particularly among learners with dyslexia. However, writing quality declined and plagiarism rates rose. The authors recommend teacher‑guided AI integration to maximize benefits while addressing risks such as over‑reliance, weak question‑asking skills, and ethical concerns in AI‑supported learning.

Students with dyslexia experience persistent challenges in writing, from limited vocabulary to difficulty organizing ideas. These hurdles are compounded in Hong Kong’s large‑classroom settings, where providing individualized writing support is often impractical. Although AI‑driven tools such as ChatGPT have begun to reshape language learning, few have been developed specifically for learners with dyslexia or rigorously tested for their impact on both motivation and writing performance.

A research team led by Fung K. Y., Fung K. C., Lee L. H., Lui R. T. L., Qu H., Song S., and Sin K. F. sought to address this gap. They designed CHATTING, a ChatGPT‑assisted writing system equipped with inclusive features such as adjustable speech rate, speech‑to‑text capabilities, and multi‑language support in Traditional Chinese, Cantonese, and English. The system was intended to be both accessible and adaptable, allowing students to interact with AI in ways that fit their learning needs.

The study aimed to measure CHATTING’s influence on learning engagement—including behavioral, emotional, cognitive, and intrinsic motivation—and to assess its effects on Chinese and English writing quality. It also explored how students formulated questions for the AI, the types of plagiarism that emerged, and the language barriers encountered during use.

The researchers recruited 101 secondary students, including both learners with and without dyslexia. Participants were randomly assigned to an experimental group, which used CHATTING, or a control group, which received traditional writing instruction. The intervention lasted four days: two days for pre‑writing tests and two days for post‑writing tests.

During the study, students completed writing tasks in both Chinese and English. These were scored on content, language, organization, and other features. Engagement and motivation were measured using questionnaires based on Self‑Determination Theory. Additional feedback on usability and functionality was gathered through surveys, while open‑ended interviews captured students’ personal experiences. Plagiarism detection software (Copyleaks) identified copied material, and statistical analysis using analysis of covariance (ANCOVA) assessed changes in engagement.

The data revealed striking differences between students with and without dyslexia. Learners with dyslexia saw significant gains in emotional engagement, which rose by 16.57%, and in intrinsic motivation, which increased by 8.71%. Their peers without dyslexia experienced moderate growth in emotional engagement of 8.65% and in cognitive engagement of 10.95%.

Feedback indicated that students with dyslexia valued CHATTING more highly across multiple dimensions, including helpfulness, relevance, readability, consistency, and conciseness. Many reported that the tool helped them generate ideas quickly, boosted their confidence, and provided a more interactive learning experience through question‑and‑answer exchanges.

Despite these positives, writing performance declined in both groups. Although word counts increased, overall scores in Chinese and English writing dropped after using CHATTING. Plagiarism was a notable concern: it was most severe in English for students with dyslexia and in Chinese for those without dyslexia.

The study also highlighted the importance of question‑asking skills. Students who could formulate specific, open‑ended questions tended to receive more relevant and useful AI responses. Others, struggling to understand AI‑generated content, resorted to copying text directly, raising both comprehension and integrity issues.
Fung et al. highlight, “The results present a nuanced picture of AI in education. On the one hand, tools like CHATTING can significantly enhance engagement and motivation, particularly for students who face traditional learning barriers. On the other hand, without guided instruction and clear expectations, such tools may unintentionally undermine writing quality, encourage plagiarism, or limit the development of independent writing skills.”

The authors argue that the key lies in teacher‑guided integration of AI into existing curricula. Educators can help students use AI as a scaffold—offering support without replacing critical processes like idea development, drafting, and revision. This approach can ensure that AI complements rather than compromises learning.

The research team acknowledges several limitations. The sample size was relatively small and lacked diversity, making it difficult to generalize findings to other contexts. The intervention was brief—just two days of AI‑assisted writing—so long‑term effects remain unknown. Additionally, some AI outputs were overly long, off‑topic, or culturally mismatched, posing challenges for second‑language learners.

The study suggests that educators should integrate AI writing platforms as supplementary tools, pairing them with explicit training in question‑asking and critical evaluation of AI output. Developers could improve such systems by incorporating plagiarism‑prevention features, adaptive difficulty settings, and clearer explanations tailored to learners’ needs. Policymakers might consider establishing ethical guidelines for AI use in education, addressing issues such as plagiarism, over‑reliance, and inclusivity.