As Generative AI technologies rapidly proliferate in the digital transformation era, understanding user technology acceptance mechanisms has become critical. This study empirically investigates a generative AI technology acceptance model by integratin...
As Generative AI technologies rapidly proliferate in the digital transformation era, understanding user technology acceptance mechanisms has become critical. This study empirically investigates a generative AI technology acceptance model by integrating privacy calculus theory and dual trust theory, utilizing a large-scale personal data breach incident as a natural experimental setting. To overcome the causal inference limitations inherent in single cross-sectional studies, we collected repeated cross-sectional data before and after the incident and employed Coarsened Exact Matching (CEM) combined with Difference-in-Differences (DiD) methodology to verify causal effects at the group level.
The analysis reveals that perceived security and information quality serve as key antecedents of dual trust formation. We demonstrate a sequential trust structure in which cognitive trust evolves into affective trust, subsequently shaping brand image. Relative advantage and perceived enjoyment exert significant positive effects on trust formation, with a notable finding that enjoyment transfers into perceptions of functional superiority.
The central discovery of this study concerns the dual role of privacy concerns. Following the major data breach, female users (particularly MZ generation residing in metropolitan areas) exhibited a positive moderating effect where privacy concerns strengthened on-device AI purchase intentions by 48.7%. Conversely, male users demonstrated a utility-seeking pattern, selecting cloud AI's advanced functionality and usefulness despite recognizing privacy risks when strong brand image exists. These findings establish that privacy concerns function not merely as inhibitors but play a dual role in either redirecting acceptance paths toward preferred technologies or suppressing adoption, contingent upon gender and technology type.
These results suggest the integrative extension of privacy calculus theory and gender schema theory, demonstrating that female risk-aversion tendencies and male utility-seeking propensities become distinctly differentiated following external shocks such as data breaches. Furthermore, we provide differentiated theoretical contributions by empirically establishing that trust serves as an antecedent to brand image in novel technology environments like generative AI, contrasting with existing research.
Methodologically, this study presents a novel approach to repeated cross-sectional research by applying the reproducibility concept from natural sciences to social sciences through the integration of CEM and DiD. Practically, we propose dual marketing strategies based on gender target segments, the buffering role of brand image during crisis situations, and the necessity of visualizing security and information quality. From a policy perspective, we suggest expanding social science R&D support, introducing incentives to encourage voluntary corporate investment in information protection, and strengthening nationwide AI literacy education, thereby contributing to the establishment of a trust-based AI ecosystem and the cultivation of informed AI consumption culture.