Overcoming the Intermediate Plateau in ESP: AI-Supported Self-Directed Learning for Ukrainian IT Students

Abstract

The intermediate plateau remains a persistent challenge in second language acquisition, particularly for ESP learners who lack sustained exposure beyond the classroom. This study investigated whether self-directed learning mediated by AI tools could support Ukrainian IT students in overcoming this plateau. Using a 12-week mixed-methods design, 66 participants engaged in structured autonomous learning cycles integrating ChatGPT, NotebookLM, and ELSA Speak for vocabulary development, linguistic analysis, domain-specific comprehension, pronunciation practice, and reflective production. Quantitative data from Cambridge proficiency assessment, an IT-specific language test, IELTS-aligned speaking evaluation, and weekly pronunciation analytics revealed statistically significant improvement across all measures (p < .001). The largest gains were observed in pronunciation accuracy and fluency, followed by domain-specific language competence and oral communicative performance. Thematic analysis of interviews and learner logs revealed three central findings: (1) AI interaction increased learner agency, motivation, and emotional safety; (2) adaptive, judgment-free feedback strengthened persistence and confidence; and (3) initial guidance in goal setting and strategy use was essential for sustained self-regulation. While participants encountered AI hallucinations, they developed improved digital resilience and critical verification practices. The findings demonstrate that structured AI-supported autonomy can mitigate linguistic stagnation while fostering metacognitive awareness, self-efficacy, and sustained engagement. The study advances understanding of AI-enhanced self-directed learning as a scalable model for ESP instruction and underscores the need for critical AI literacy in autonomous learning ecosystems.