Creative Self-Efficacy, System Assurance, and Continued-Use Intention for AI-Enabled Design Tools: A Dual-Path Model Among Design Students

Abstract

Design students increasingly use generative AI tools, yet post-adoption research has not differentiated the beliefs associated with continued use in open-ended creative work. This study proposes a dual-path model organizing post-adoption beliefs by source and function: AI-assisted creative self-efficacy refers to the user, system assurance to perceived reliability and competence of the technology, and perceived usefulness to task-relevant benefit. Data were collected from 329 students in design-related programs across Chinese universities. After excluding 29 invariant response patterns, the primary sample was 300, with sensitivity analyses using full and stricter-screened samples. Confirmatory factor analysis and robust maximum likelihood structural equation modeling supported the proposed structure, scaled χ 2 (72) = 146.97, robust CFI = .948, robust TLI = .934, robust RMSEA = .068, 90% CI [.052, .084], and SRMR = .054. Creative self-efficacy was associated with usefulness (β = .546), ease of use with system assurance (β = .539), and system assurance with usefulness (β = .241). Indirect associations were statistically nonzero but should be interpreted cautiously because of the final usefulness–intention segment. Perceived usefulness and continued-use intention could not be satisfactorily separated empirically (latent r = .939; HTMT = .924), and constraining their correlation to unity did not significantly worsen fit. A parallel-antecedent specification showed comparable fit by AIC. The principal contribution therefore concerns the upstream source–function organization of capability, interaction ease, perceived reliability and competence, and usefulness; H4 and related indirect associations are interpreted as statistically nonzero but conceptually limited rather than evidence of a distinct behavioral stage.