Data Sharing Determinants in GenAI-applications: A Systematic Review and Structural Equation Modelling Approach

Abstract

This paper investigates the determinants influencing individuals' data sharing decisions through a two-pronged approach combining systematic literature review and empirical analysis. First, a systematic review of 53 studies across multiple fields and technologies identifies and synthesizes key determinants into a comprehensive framework, addressing the absence of unified definitions and closing a critical literature gap. Second, an online survey (n=357) examines relationships between privacy concerns, perceived risks and benefits, trust, prior disclosure behavior, and willingness to share personal data specifically in the GenAI context. Using partial least squares structural equation modeling, the study reveals trust significantly reduces privacy concerns, consistent with existing literature. However, trust's effects on data sharing willingness and perceived risks proved statistically insignificant. Surprisingly, both perceived risks and benefits showed negative (though insignificant) relationships with sharing willingness, contradicting traditional assumptions. These unexpected findings are attributed to ChatGPT's still novel technological environment, potential survey framing effects, and the unique nature of AI-driven platforms that may reshape conventional decision-making processes. The study provides evidence of the privacy paradox in ChatGPT contexts, where stated privacy concerns diverge from actual sharing behaviors. This research contributes a comprehensive framework for understanding data sharing determinants, offers insights into privacy decision-making in emerging AI technologies, and provides practical implications for businesses deploying large language models. Despite limitations in sampling methods and external validity, this work advances theoretical understanding and practical applications while establishing foundations for future research in AI-mediated data disclosure behaviors.

Keywords