Development and Validation of Student Artificial Intelligence Self Efficacy Scale

Abstract

This study aimed to develop and validate the Student Artificial Intelligence Self-Efficacy Scale (SAISES) as an instrument for measuring university students’ self-efficacy in the use of artificial intelligence (AI) for academic purposes. The instrument was adapted from the General Self-Efficacy Scale developed by Sherer and Maddux (1982) and subsequently adapted into Indonesian by Suharsono and Istiqomah (2014). It encompasses three dimensions: level, strength, and generality. The adaptation process was conducted systematically following the American Educational Research Association (AERA, 2014) guidelines, including translation, synthesis, back-translation, and expert committee review. The study involved 317 university students from East Java, Indonesia (177 females and 140 males; aged 19–25 years) selected through area random sampling. Construct validity was examined using second-order Confirmatory Factor Analysis (CFA), while reliability was assessed through composite reliability. The results indicated that the measurement model achieved acceptable goodness-of-fit indices (CMIN/DF = 2.881, RMSEA = 0.077, CFI = 0.900, and SRMR = 0.051). Following the removal of one invalid item (Item 23), the remaining 22 items demonstrated satisfactory validity, with factor loadings exceeding 0.50. Discriminant validity was established based on Heterotrait–Monotrait Ratio (HTMT) values below 0.85. Furthermore, all dimensions exhibited composite reliability values above the recommended threshold of 0.70 (level = 0.899, strength = 0.873, and generality = 0.900). Among the three dimensions, level contributed the most strongly to the overall SAISES construct (loading = 0.907). In conclusion, the SAISES is a valid and reliable instrument for assessing AI self-efficacy among Indonesian university students. The scale can be utilized as an assessment tool to support the development of AI literacy interventions and educational programs in higher education settings.

Penelitian ini bertujuan mengembangkan dan memvalidasi Student Artificial Intelligence Self-Efficacy Scale (SAISES) sebagai instrumen pengukuran keyakinan diri mahasiswa dalam penggunaan kecerdasan buatan (AI) untuk keperluan akademik. Instrumen diadaptasi dari General Self-Efficacy Scale (Sherer & Maddux, 1982) yang telah diadaptasi ke bahasa Indonesia oleh Suharsono dan Istiqomah (2014), mencakup tiga dimensi: level, strength, dan generality. Proses adaptasi dilakukan secara sistematis mengacu pada panduan AERA (2014) melalui tahapan penerjemahan, sintesis, penerjemahan balik, dan tinjauan komite ahli. Penelitian melibatkan 317 mahasiswa di Jawa Timur (perempuan = 177, laki-laki = 140; rentang usia 19–25 tahun) dengan teknik area random sampling. Validitas konstruk diuji menggunakan Confirmatory Factor Analysis (CFA) dua tingkat (second-order), dan reliabilitas diukur melalui composite reliability. Hasil menunjukkan bahwa model memenuhi kriteria goodness of fit yang dapat diterima (CMIN/DF = 2,881; RMSEA = 0,077; CFI = 0,900; SRMR = 0,051). Setelah reduksi satu item yang tidak valid (item 23), 22 item terbukti valid dengan loading factor di atas 0,50. Validitas diskriminan terpenuhi berdasarkan nilai HTMT di bawah 0,85. Nilai composite reliability seluruh dimensi melampaui 0,70 (level = 0,899; strength = 0,873; generality = 0,900). Dimensi level memberikan kontribusi terbesar terhadap konstruk SAISES (loading = 0,907). Disimpulkan bahwa SAISES merupakan instrumen yang valid dan reliabel untuk mengukur self-efficacy penggunaan AI pada mahasiswa Indonesia, dan dapat digunakan sebagai alat asesmen dalam merancang intervensi literasi AI di perguruan tinggi.