Abstract:

This study investigates artificial intelligence (AI) integration patterns among students in English for Specific Purposes (ESP) courses at four European universities participating in the ELITE-AI Erasmus+ project. Through systematic analysis comparing complementary student and teacher perspectives, the research documents critical tensions between reported satisfaction and actual learning behaviors that challenge assumptions about AI integration effectiveness.

The investigation employed a descriptive quantitative design combining structured self-assessment instruments with teacher observation protocols. Student self-assessments were administered to more than 100 undergraduate participants following AI-integrated classroom sessions across four institutions in Croatia, Romania, Slovenia and Italy, representing education, mobility, criminal justice and law, and tourism fields. Concurrently, instructors completed systematic observation sheets documenting classroom dynamics during AI integration. This dual-perspective approach enabled triangulation between student perceptions and teacher observations of actual classroom behaviors. Data collection measured effectiveness of AI tool use, active participation patterns including help-seeking behaviors, independence in AI use decisions, critical evaluation of AI-generated content, and integration of AI into learning processes.

Student self-assessment results revealed overwhelmingly positive satisfaction (91.1% reporting AI integration helpful), yet behavioral patterns showed concerning discrepancies. Regarding autonomy, 66.3% used AI only because teachers instructed them, whereas just 32.7% demonstrated proactive engagement. Help-seeking patterns shifted notably: 82.2% utilized AI for clarification at least occasionally, while 30.7% never asked teachers for help, suggesting students increasingly rely on AI as first resource. Critical evaluation revealed that while 63.4% engaged in verification behaviors, 28.7% accepted AI outputs uncritically without questioning accuracy. Integration patterns showed 59.4% perceiving AI as useful classroom tool, but 39.6% using AI for isolated tasks without connecting to overall learning.

The investigation reveals fundamental tensions undermining pedagogical objectives. Teacher-directed AI engagement contradicts learner-centered instruction principles, potentially reinforcing dependency rather than cultivating autonomy. Students consulting AI instead of teachers may diminish valuable interactions serving multiple pedagogical functions beyond information transmission. The substantial proportion accepting AI outputs uncritically demonstrates digital literacy gaps, indicating urgent need for explicit AI literacy instruction including hallucination recognition and verification strategies. Task-specific application without broader integration indicates perceived usefulness alone does not guarantee meaningful outcomes.

This paper proposes four essential components for effective ESP AI integration: scaffolded autonomy development, critical AI literacy instruction as curricular priority, balanced human-AI interaction maintaining teacher-student relationships, and holistic pedagogical design connecting AI interactions to comprehensive objectives. The dual-perspective methodology reveals that successful AI integration requires substantially more sophisticated implementation than simple tool provision combined with positive attitudes.

Keywords:

AI integration, English for Specific Purposes, learner autonomy, critical AI literacy, self-regulated foreign language acquisition, classroom dynamics, pedagogical scaffolding.