The Application Effectiveness and Improvement Strategies of Generative AI Feedback Systems in English Academic Writing Instruction for STEM Graduate Students

Abstract

High-quality academic writing is essential for effective dissemination of research findings in engineering disciplines, including electromagnetic waves, antennas, and propagation, where precise technical communication directly influences k nowledge t ransfer a nd i nternational c ollaboration. This study develops and validates a generative artificial i ntelligence ( AI) f eedback f ramework f or English academic writing instruction among STEM graduate students by integrating selfdetermination theory with process writing theory. The proposed system provides automated error diagnosis, language refinement, structural optimization, and personalized revision guidance through a collaborative AI–instructor workflow. Experimental results demonstrate that the framework significantly i mproves w riting q uality, learning motivation, and feedback efficiency w hile m aintaining r eliable p erformance i n g rammar c orrection and discipline-oriented language analysis. Complex reasoning tasks requiring expert judgment are delegated to instructors to preserve academic rigor. In addition, adaptive feedback strategies based on learner proficiency a nd c ontextual r equirements e nhance t he e ffectiveness o f t echnical w riting i nstruction and facilitate standardized expression of scientific i nformation. T he p roposed f ramework o ffers a practical approach for intelligent academic writing support and provides methodological insights for preparing high-quality engineering manuscripts and technical documentation in interdisciplinary research environments, particularly those requiring accurate communication of electromagnetic and related scientific knowledge.