Speaker
Description
As generative artificial intelligence (AI) becomes increasingly integrated into higher education, developing AI competence requires more than technical proficiency. Students need to use AI critically, make informed decisions, and maintain ownership of their learning and writing. Guided by Student Engagement Theory (Fredricks et al., 2004) and Control-Value Theory (Pekrun, 2006), this mixed-method action research examined the effectiveness of an agency-oriented AI scaffolding framework in supporting first-year students’ paragraph writing and responsible AI use. The study involved 116 English-major freshmen enrolled in an intensive second-semester writing course at a Vietnamese university. The intervention integrated Generative AI and Automated Writing Evaluation (AWE) tools into weekly drafting cycles under explicit ethical guidelines. Rather than relying uncritically on AI-generated feedback, students were encouraged to evaluate AI suggestions, justify their revisions, and take responsibility for their final writing decisions. Data were collected through pre- and post-intervention writing assessments completed without AI assistance, a learner autonomy questionnaire informed by the two theoretical frameworks, and semi-structured interviews. The findings showed improved performance in paragraph unity, coherence, vocabulary use, and grammatical accuracy. Questionnaire responses indicated that students viewed AI as a tool that supported independent learning, sustained engagement, and greater confidence in the writing process. Interview data further revealed that structured AI scaffolding encouraged students to evaluate AI feedback critically, make informed revision decisions, and develop stronger ownership of their writing. The study demonstrates that AI is most effective when used as a scaffold for learning rather than a substitute for student thinking. It offers a practical framework for promoting AI competence, learner agency, and ethical assessment in AI-enhanced university writing classrooms.
