<![CDATA[The growing use of artificial intelligence (AI) among students has raised concerns about dependence that may weaken learning independence, a pattern also observed among students at SMK IPIEMS Surabaya who tend to rely on AI when completing tasks or answering questions in class. This study aims to examine the influence of self-efficacy and academic stress on AI dependence, with performance expectations as a mediating variable, among Office Management students at SMK IPIEMS Surabaya. An explanatory quantitative approach was used, involving 108 respondents selected through purposive sampling from a population of 150 students. Data were collected using a Likert-scale questionnaire and analyzed with Partial Least Squares Structural Equation Modeling (PLS-SEM) using SmartPLS 4. The results show that academic stress has a significant positive effect on AI dependence, while self-efficacy does not show a significant direct effect. Both self-efficacy and academic stress were found to increase performance expectations, which in turn play an important role in driving students' dependence on AI, including as a mediator for both exogenous variables. These findings confirm that students' expectations of AI's benefits serve as the main pathway toward dependence, so that managing academic workload and promoting wise AI literacy deserve attention from both schools and students. ABSTRAK Penggunaan kecerdasan buatan (AI) yang semakin masif di kalangan pelajar memunculkan kekhawatiran akan ketergantungan yang dapat melemahkan kemandirian belajar, termasuk pada siswa SMK IPIEMS Surabaya yang menunjukkan kecenderungan mengandalkan AI dalam menyelesaikan tugas maupun menjawab pertanyaan di kelas. Penelitian ini bertujuan menganalisis pengaruh self-efficacy dan academic stress terhadap AI dependence, dengan performance expectations sebagai variabel mediasi, pada siswa jurusan Manajemen Perkantoran SMK IPIEMS Surabaya. Penelitian menggunakan pendekatan kuantitatif eksplanatori dengan melibatkan 108 responden yang dipilih melalui purposive sampling dari populasi 150 siswa. Data dikumpulkan melalui kuesioner berskala Likert dan dianalisis menggunakan Partial Least Squares Structural Equation Modeling (PLS-SEM) dengan SmartPLS 4. Hasil penelitian menunjukkan bahwa academic stress berpengaruh positif signifikan terhadap AI dependence, sementara self-efficacy belum menunjukkan pengaruh langsung yang signifikan. Baik self-efficacy maupun academic stress terbukti meningkatkan performance expectations, yang selanjutnya berperan penting dalam mendorong ketergantungan siswa terhadap AI, termasuk sebagai mediator dari kedua variabel eksogen tersebut. Temuan ini menegaskan bahwa ekspektasi siswa terhadap manfaat AI menjadi jalur utama munculnya ketergantungan, sehingga pengelolaan beban akademik dan literasi penggunaan AI yang bijak perlu menjadi perhatian sekolah maupun siswa itu sendiri.]]>
