Student-AI collaboration in literacy: Evaluating ChatGPT-driven text personalization

This study explores the impact of ChatGPT-Student Interaction-Based Texts (CSIBT) on the reading comprehension and motivation of primary school students. The present study was grounded in the Human-AI Collaborative Learning Model. To this end, a mixed-methods design was employed, combining a quasi-experimental pretest-posttest control group with qualitative interviews. The experimental group engaged with personalized reading materials co-created with ChatGPT, while the control group followed the standard curriculum. Quantitative data were collected through the administration of validated reading comprehension and motivation scales, and subsequently analyzed using the analysis of variance (ANOVA) and the analysis of co-variance (ANCOVA) statistical procedures. The validity of the normality assumptions was confirmed through the implementation of the Shapiro-Wilk test. The findings indicated that students in the CSIBT group exhibited significantly more pronounced enhancements in both reading comprehension and motivation compared to the control group. The qualitative findings corroborated these results, underscoring an increase in student engagement, enjoyment, and perceived relevance of the texts. These findings suggest that AI-supported personalized texts, when integrated through student interaction, can effectively enhance reading outcomes in early education. The study offers practical implications for the development of learner-centered and AI-assisted literacy instruction.