Abstract
Perkembangan pesat teknologi Artificial Intelligence (AI) telah mengubah lanskap pendidikan tinggi secara signifikan, terutama dengan hadirnya tutor virtual berbasis AI yang mampu memberikan umpan balik cepat, tepat, dan personal. Analisis Real merupakan salah satu mata kuliah yang paling menantang dalam program pendidikan matematika karena sifat abstraknya sering menyebabkan rendahnya self-efficacy mahasiswa. Penelitian ini bertujuan menganalisis pengaruh pemanfaatan AI sebagai tutor virtual terhadap self-efficacy mahasiswa Pendidikan Matematika Universitas Asahan yang menempuh mata kuliah Analisis Real. Penelitian menggunakan pendekatan kuantitatif dengan desain quasi-experimental one-group pretest-posttest, melibatkan 15 mahasiswa yang dipilih melalui purposive sampling. Instrumen angket self-efficacy dikembangkan berdasarkan tiga dimensi Bandura (magnitude, strength, generality), divalidasi melalui expert judgment dengan Cronbach’s Alpha sebesar 0,84. Hasil analisis deskriptif menunjukkan peningkatan rata-rata skor self-efficacy dari 65,67 (pre-test, kategori sedang) menjadi 79,93 (post-test, kategori tinggi), dengan peningkatan sebesar 21,7%. Uji Wilcoxon Signed-Rank Test menghasilkan nilai Z = −3,408 dengan p-value = 0,001 (p < 0,05), yang mengkonfirmasi terdapat perbedaan signifikan self-efficacy mahasiswa sebelum dan sesudah intervensi enam minggu menggunakan AI tutor virtual. The rapid development of Artificial Intelligence (AI) technology has significantly transformed higher education, particularly through AI-based virtual tutors capable of providing fast, precise, and personalized feedback. Real Analysis, one of the most challenging mathematics education courses, frequently causes low self-efficacy among students due to its abstract nature and demands for formal proof-writing. This study examines the effect of utilizing AI as a virtual tutor on self-efficacy of mathematics education students at Universitas Asahan. A quantitative quasi-experimental one-group pretest-posttest design was employed with 15 students selected via purposive sampling. The self-efficacy questionnaire was based on Bandura’s three dimensions: magnitude, strength, and generality, validated by expert judgment with Cronbach’s Alpha coefficient of 0.84. Descriptive analysis results showed an increase in mean self-efficacy scores from 65.67 (pre-test, moderate category) to 79.93 (post-test, high category), representing a 21.7% improvement. The Wilcoxon Signed-Rank Test yielded Z = −3.408 with p-value = 0.001 (p < 0.05), confirming a significant difference in student self-efficacy before and after the six-week AI virtual tutor intervention.