Construction and validation of the Emotional Trust in Artificial Intelligence Scale (CEIA)

Emotional trust in Artificial Intelligence (AI) has become a crucial component of human-AI interaction, especially in academia, where generative systems are used for academic, communicational, and socio-emotional purposes. While research has advanced in cognitive trust, there is a lack of validated instruments that rigorously assess the affective dimension of trust, which involves processes of perceived authenticity, emotional validation, affective recognition, and the establishment of symbolic bonds with artificial agents.

To develop and psychometrically validate the Emotional Confidence in Artificial Intelligence Scale (CEIA) in university students, examining its factor structure, reliability, convergent and discriminant validity, as well as its measurement invariance by sex.

Six hundred and fibe Peruvian university students, aged 16 to 88 years (M = 22.93, SD = 7.35), participated in the study. Item distributions, item-total correlations, and descriptive statistics were analyzed. Internal structure was assessed using Exploratory Factor Analysis (EFA) with polychoric correlations and Confirmatory Factor Analysis (CFA). Reliability was estimated using Cronbach's alpha and McDonald's omega. Convergent and discriminant validity were examined using AVE, CR, and HTMT. Sex invariance was assessed using multigroup CFA with the WLSMV estimator.

The EFA supported a multidimensional structure consistent with the theoretical model. The CFA confirmed a four-factor solution-Perceived Authenticity, Emotional Validation, Affective Recognition, and Symbolic Bonding-with adequate fit indices (CFI = 0.944; TLI = 0.937; RMSEA = 0.073; SRMR = 0.033). The dimensions showed high internal consistency (α = 0.91-0.94; ω = 0.91-0.94) and satisfactory evidence of convergent and discriminant validity. Invariance analysis demonstrated metric invariance and partial scalar invariance between men and women.

The CEIA is a valid, reliable, and partially sex-invariant instrument for assessing emotional trust in AI among university students. It constitutes a robust tool for research, technological design, and the ethical development of AI systems with affective capabilities.