Abstract: [Purpose/Significance] Smart homes have enhanced the comfort of home life, but they also raise privacy concerns for users. Exploring the users' privacy risk perception and protective behavior in the scene of smart home usage is helpful to understand the new privacy problems faced by users in special situations and their behavior patterns, and help enhance the privacy protection design of smart home. [Design/Methodology] Semi-structured interviews were conducted with 37 participants, Grounded Theory was employed to analyze the interview data, and a theoretical model was constructed around "privacy risk perception of smart home users", with the storyline being described. [Findings/Conclusion] It is found that four types of antecedents affect smart home users' perception of privacy risk, including the type of privacy information, privacy experience, external assurance assessment and technological prospects forecast, as well as other information products experience. In the smart home usage scene, users' privacy protective behaviors include three categories: limiting the amount of information disclosure, managing the usage scenarios and avoiding the use of specific functions. Privacy risk perception of smart home users and their privacy protective behavior interact with each other, and the perceived value moderates the impact of users' privacy risk perception of smart home on their privacy protective behavior. There is no clear indication of a privacy paradox in the scene of smart home usage. [Originality/Value] The research focuses on the users' privacy risk perception and protective behavior in specific scenarios, which theoretically promotes the user privacy related research in human-AI interaction scenarios, and practically provides a reference for the privacy protection design of smart homes, and helps realize the long-term development goal of "human-centered artificial intelligence".

Keywords: Smart home, Privacy, Privacy risk perception, Human-AI interaction, Protective behavior