Adaptation of the Human–Computer Trust Scale into Turkish in the Context of Generative Artificial Intelligence
International Conference on Educational Technology and Online Learning 2026, Bremen, Almanya, 17 - 20 Ağustos 2026, (Özet Bildiri)
- Yayın Türü: Bildiri / Özet Bildiri
- Basıldığı Şehir: Bremen
- Basıldığı Ülke: Almanya
- Recep Tayyip Erdoğan Üniversitesi Adresli: Evet
Özet
Trust has become a critical construct in human–artificial intelligence interaction as generative artificial intelligence increasingly produces, recommends, and mediates content in educational and everyday contexts. Unlike conventional artificial intelligence systems, generative AI requires users to evaluate the reliability, accuracy, transparency, usefulness, and ethical acceptability of system-generated outputs. Although recent studies in the Turkish context have mostly focused on artificial intelligence literacy, attitudes, awareness, and anxiety, there is still a clear need for valid and reliable instruments that specifically measure trust in artificial intelligence and generative AI. Therefore, this study aimed to adapt the revised Human–Computer Trust Scale into Turkish in the context of generative artificial intelligence and to examine its psychometric properties.
This study was designed as a scale adaptation study. The original scale, developed by Gulati et al. and later extended and cross-culturally validated by Beltrão et al. (2025), consists of 14 items rated on a five-point Likert scale. The translation process included forward translation, back-translation, and expert review. The Turkish form was evaluated by two Turkish language experts and eight instructional technology experts for linguistic clarity, content validity, conceptual equivalence, and contextual appropriateness. Prior to final administration, face-to-face interviews with ten undergraduate students were conducted to assess item comprehensibility and dimensional alignment. Based on expert and pilot feedback, necessary revisions were made. The final form was administered online to [N = 696] participants during the 2025–2026 academic year. Confirmatory factor analysis, convergent validity, discriminant validity, composite reliability, and Cronbach’s alpha analyses were conducted.
The confirmatory factor analysis indicated that item m3 was removed due to a low factor loading, while items m12 and m7 were removed because of modification indices. The final model consisted of 11 items and five factors: perceived risk, benevolence, competence, structural assurance, and trust. The model showed acceptable to good fit values [χ²/df = 3.429; GFI = .969; AGFI = .940; NFI = .946; CFI = .961; TLI = .936; IFI = .961; RMSEA = .059; RMR = .040; SRMR = .045]. Factor loadings ranged from .545 to .853. AVE values ranged between .43 and .61, composite reliability coefficients ranged between .60 and .76, and Cronbach’s alpha coefficients ranged between .60 and .75. The overall internal consistency coefficient was α = .66. HTMT values were below the commonly accepted threshold, supporting discriminant validity.
The findings indicate that the Turkish form of the scale provides acceptable validity and preliminary evidence of reliability for measuring trust in generative AI systems. The adapted scale may contribute to future research on human–AI interaction, generative AI adoption, educational technology use, and trust-related variables in Turkish educational contexts. Future studies are recommended to test the scale with larger and more diverse samples and to examine its relationships with variables such as AI literacy, technology acceptance, perceived usefulness, and ethical AI use.