Trust and Perceived Usefulness in GenAI-Assisted Assessment Adoption Among Design Teachers in Higher Education: Distinct Judgments and Their Associations with Behavioral Intention

Abstract

Generative artificial intelligence (GenAI) is increasingly being considered for assessment-related work, but educators need to judge whether it is reliable enough to serve as professional input and whether it is useful enough to justify its incorporation into assessment practice. This study examined whether trust (TR) and perceived usefulness (PU) are empirically distinct judgments and whether their independent associations with behavioral intention (BI) to use GenAI for assessment differ. Cross-sectional survey data from 352 design teachers at universities and higher vocational colleges in China were analyzed using confirmatory factor analysis, covariance-based structural equation modeling, comparisons of alternative structural specifications, and 5000-sample bootstrap tests. Measurement evidence supported treating TR and PU as separate latent judgments. In the correlated parallel model (Model C2), PU was positively associated with BI, whereas the TR–BI association was nonsignificant; a formal test of the unstandardized path coefficients showed that the PU–BI coefficient was significantly larger than the TR–BI coefficient. Because Model C2 was structurally just-identified, its global fit was not interpreted as independent evidence of structural robustness. Sensitivity analyses accounting for prior GenAI-assisted assessment use yielded the same substantive path pattern. The findings concern acceptance judgments and behavioral intention rather than actual adoption or assessment outcomes. They suggest that reliability and perceived task value are related but nonredundant considerations when educators consider GenAI support while retaining final assessment responsibility.