Validation of a Questionnaire to Explore Artificial Intelligence Use Among Upper-Secondary and Higher Education Teachers

Abstract

Research on generative artificial intelligence (AI) in education has shifted from students to teachers, yet available teacher instruments rest on technology acceptance models, professional knowledge frameworks, or AI literacy, and none articulates dispositional conditions, teaching activities, and contextual consequences within a single functional logic. This study extends to teachers an instrument previously validated with students and grounded in Functional Behavior Analysis and the Interbehavioral Model of Didactic Performance. Participants were 233 Mexican upper-secondary and higher education teachers (mean age 47.22 years, SD = 10.44). The study is strictly psychometric in scope: it reports initial reliability and internal structure evidence, not a profile of teacher practice. Reliability was estimated with Cronbach’s alpha and McDonald’s omega with bootstrap confidence intervals, and the number of components was determined through Horn’s parallel analysis. Alpha coefficients ranged from 0.703 to 0.957 (omega, 0.755 to 0.962). The benefits dimension split into two uncorrelated components, pedagogical benefits (α = 0.953) and risks to academic integrity (α = 0.771). Domains of didactic performance proved unidimensional (63.0% of variance); the consequences dimension behaved as a formative index rather than a reflective scale; and the dispositional dimension failed to separate individual from contextual variables and required redesign. A second-order analysis yielded two orthogonal axes, favourable pedagogical orientation and risk recognition, on which four quadrants are described for illustrative purposes only. The instrument discriminates across self-reported frequency of AI use. Because components were extracted from Pearson correlations on ordinal items in a single convenience sample, the structure reported here is provisional and requires confirmation with polychoric matrices and robust estimators on an independent sample.