Integrating a SAMR-Based Generative AI Framework to Enhance ESL Learners' Essay Writing Skills and Critical Thinking: A Classroom-Based Study

https://doi.org/10.5281/zenodo.21307764

Abstract

The use of generative artificial intelligence (AI) tools in English as a Second Language (ESL) academic writing is becoming commonplace, but their use in the classroom is not always systematic. This study investigated the perceptions of 34 undergraduate ESL students of the integration of generative AI, specifically ChatGPT, into the teaching of essay writing in four theoretically informed dimensions: AI acceptance, writing self-efficacy, critical thinking, and integration of generative AI in the context of SAMR. The composite reliabilities were acceptable to good (α = .63-.86; 15-item overall scale α = .83), except for a two-item acceptance subscale (α = .31) using a descriptive-correlational survey design with a 15-item, five-point Likert questionnaire. Students' writing self-efficacy (M = 3.91), perceived critical-thinking engagement (M = 3.91), perceived SAMR-based integration (M = 3.87), and acceptance score (M = 3.65) were all high, and 76.5% planned to continue using AI for future academic writing. The SAMR-based integration was highly correlated with critical thinking (r = .77) and moderately-to-strongly correlated with acceptance (r = .63), and an independent-samples t-test showed that there was a significant difference between the self-efficacy of male and female students (p = .003). Interestingly, intention to continue using AI was significantly related to critical-thinking engagement (r = .39, p = .025), but not to acceptance alone (r = .22, p = .22). The results indicate that the depth of critical engagement, not just ease of use, is the strongest predictor of continued use of AI, providing empirical evidence for a staged approach to AI pedagogy, informed by SAMR, over blanket permission or prohibition policies. Keywords: SAMR Model, generative AI, Essay Writing Skills, Critical Thinking, Writing Instruction.