Fostering Higher-Order Thinking Skills in ESL Classrooms Through Ai-Supported Writing Tasks Jurayeva Zuhra

Аннотация

The rapid diffusion of generative artificial-intelligence (AI) tools such as ChatGPT is reshaping language- learning ecologies. While early adopters have focused on gains in fluency and accuracy, a more compelling pedagogical question is whether AI can cultivate higher-order thinking skills (HOTS)—analysis, evaluation and creative synthesis—in English as a Second Language (ESL) learners. The present study examines an intervention in which undergraduate ESL students completed a sequence of argumentative and reflective writing tasks mediated by an AI co-writer that provided dynamic scaffolding, metacognitive prompts and automated discourse analysis. Grounded in Bloom’s revised taxonomy and socio-constructivist learning theory, the mixed-methods design combined quasi-experimental pre- and post-testing with thematic analysis of learner journals. Quantitative results showed statistically significant improvements (p < 0.05) in students’ HOTS rubric scores compared with a control group engaged in traditional peer-review cycles. Qualitative data revealed heightened metalinguistic awareness and strategic risk-taking in idea development. The findings suggest that, when carefully orchestrated, AI-supported writing can transcend mere linguistic assistance and become a catalyst for deeper cognitive processing. Pedagogical implications and design principles for AI-enhanced assessments that preserve academic integrity are discussed. Keywords: Higher-order thinking; AI-supported writing; ESL pedagogy; ChatGPT; Bloom’s taxonomy; academic integrity; mixed-methods. evaluating evidence and producing novel syntheses. Introduction: Educators have long aspired to move Contemporary scholarship urges a re-examination of second-language writing instruction beyond sentence- these taxonomic categories in light of generative AI, level accuracy toward the cultivation of analytical and positing that strategic prompting can prompt learners creative reasoning. Yet empirical evidence indicates to operate at “create” and “evaluate” levels more that ESL classrooms often remain anchored in lower- consistently. Nevertheless, there is limited classroom- level cognitive activities, partly because teachers must based research that operationalises HOTS explicitly devote considerable time to error correction. The within AI-supported task design for ESL populations. emergence of large language models (LLMs) offers Addressing this gap, the present study investigates unprecedented opportunities to redistribute this whether structured human–AI collaboration during cognitive load. Recent studies conducted in writing tasks can measurably enhance HOTS while undergraduate writing courses demonstrate that maintaining language-development objectives. students who receive AI-mediated formative feedback outperform peers on measures of evaluative reasoning This investigation adopted an explanatory-sequential and argumentation quality. However, scepticism mixed-methods design that combined a quasi- persists regarding over-reliance on algorithmic text experimental component with qualitative process generation and the potential erosion of original tracing to elucidate how artificial-intelligence thought. mediation affects higher-order thinking during second- language writing. The study unfolded over a fifteen- Higher-order thinking skills, situated at the apex of week semester in two intact sections of “English for Bloom’s taxonomy, encompass analyzing patterns, International Journal of Pedagogics 18 https://theusajournals.com/index.php/ijp International Journal of Pedagogics (ISSN: 2771-2281) Academic Purposes II” at a public university in Central The adjusted post-test means revealed that Asia. Prior to group assignment, students (N = 54) experimental-group students achieved higher overall completed the TOEFL ITP to ensure baseline HOTS scores (M = 22.3, SD = 2.1) than control peers (M comparability (M = 487, SD = 21); Levene’s test = 18.7, SD = 2.8), yielding a medium effect size (η² = confirmed homogeneity of variance (p = 0.64). The 0.21). Dimension-level analysis indicated the greatest experimental cohort (n = 27) engaged in AI-supported gains in evidence evaluation and creative integration, writing cycles, whereas the control cohort (n = 27) aligning with earlier findings on human–AI followed a traditional peer-review model. collaborative writing. Journal analysis uncovered three dominant themes. First, students reported a “dialogic Instructional materials were fully parallel across push,” noting that AI prompts forced them to articulate conditions: three genre modules—argumentative warrants for their claims rather than accept surface- essay, problem–solution report and reflective blog— level paraphrases. Second, learners described each allotted four weeks. In the experimental group, heightened confidence to experiment with unfamiliar students interacted with ChatGPT-4 on a closed disciplinary vocabulary because immediate AI feedback intranet at three scaffo