Preliminary Development of the Generative AI Instructional Integration Experience Scale: A Synthetic Persona-Based Pilot Examination of Item Structure

This study preliminarily developed a scale for measuring teachers’ generative AI instructional integration experience and conducted a pilot examination of its item structure using expert content validation and two independent synthetic persona-based simulated response datasets. Based on prior studies on TPACK, AI self-efficacy, teacher AI competence, AI acceptance, and ethical use of generative AI, 40 initial items were developed across four dimensions: AI lesson preparation, AI lesson implementation, AI assessment and feedback, and AI use reflection. Twelve experts reviewed item relevance, clarity, representativeness, redundancy, and factor fit, resulting in 28 preliminary items. For exploratory factor analysis, 549 valid responses were obtained after excluding 11 inattentive cases from 560 Nemotron-Personas-KOREA-based simulated responses. Maximum likelihood extraction and direct oblimin rotation retained 20 items. For cross-validation within synthetic persona-based data, a separate synthetic persona sample(N=600), generated independently from the EFA sample, was used for confirmatory factor analysis. The one-factor model was inadequate, whereas the correlated four-factor model showed acceptable fit and the second-order model also showed acceptable fit. The findings provide preliminary evidence for the item structure of a multidimensional construct and demonstrate a methodological example for early-stage item refinement in the AI era. Because the simulated responses were generated from synthetic personas rather than actual teachers, the results should be interpreted as a supplementary pilot examination of item structure, not as final validation evidence for the teacher population.